Reboot Alberta

Monday, September 07, 2009

Can Pre-Election Opinion Polls Do More Harm Than Good?

I don't have much faith and even less trust in opinion polls. Not because I think they are wrong or a waste of time. I just think they are too trapped in time and context and only a superficial snap shot of "opinion." So they are a pretty meaningless predictor of actual behaviours.

So I was interested to read this piece on polling in yesterday's Edmonton Journal. Ipsos Reid is one of the best brand names for polling around. It was good to see the discussion on the role and limitations on opinion polling.

It is getting harder to get folks to respond to polls these days, and when they do to ensure you have a truly random sample and that participants are telling you the truth. One of my favourite bumper stickers from many years ago was "Save Democracy, Lie to a Pollster."

I think this attitude about intentionally lying to pollsters is more prevalent than many realize. Also, the random sample may be demographically pure but there are so many more phone calls that have to be made to get people to actually participate. As a result we get a self-selection skew in the randomness of sampling. This self-selection skewing is especially true in on-line polling techniques.

I know lots of people who admit lying to pollsters, intentionally. Others give normative but untrue answers that do not reflect reality either. For example less than 60% of Canadians voted in the last federal election but polls indicate a much larger number say they did. Are they lying? some are. Did they forget that they did not vote but intended to vote and re making a mistake? Are they giving what they know is the proper (normative) answer even though they know it is not the truth?

Mr Bricker of Ipsos Reid also notes in the article that the order and syntax of questions will make a big difference in the answers polls generate. So with all this, I take opinion poss with a grain of salt. Those unscientific "surveys" you see in websites of traditional media are actually dangerous. This is because they have a air of unwarranted credibility about they because of the authority of the newspaper, radio or television broadcaster who is hosting them. They often get hijacked by special interests or competing interests, like political parties. The results mislead an unsuspecting public and can have a significant impact on the actual beliefs of many well intentioned but ill-informed people. Look how many Americans still think 911 terrorists came through Canada, even the US Cabinet Secretary involved made a recent comment to that effect. Mistaken initial beliefs are had to change, regardless of the amount and credibility of the subsequent evidence to the contrary.

That said, let the polls proceed. We just need to ensure we have some general literacy in our society about what opinion polls do, what they mean and don't mean and what they "prove" - if anything. I often do commentary and analysis on political opinion polls in the blog. I think the real value they have is, over time and with many sources on similar questions and issues, they can collectively provide a sense of trend or direction of public sentiment. But unless we are into an actual election, asking a hypothetical "how would you vote tomorrow if an election was called" generates pretty meaningless data.

Elections matter and campaigns create consequences that generates real results that truly matter to the good of the country. Poll away but don't let them have any sway until the reality of an actual election is happening, then, and only then should people pay them some heed.

1 comment:

  1. Anonymous8:43 pm

    I find this particular interesting given the recent polls that both the Wildrose Alliance Party and the PC Party are pointing to as some sort of success and/or measure of their support in Calgary-Glenmore.

    Of course the polls conducted by the WAP are going to favour them and the polls conducted by the PC's are going to favour them.

    I think focusing on these polls with any sort of definitive expectation is faulty thinking. There are far too many variables to be able to make any sort of definitive conclusions based strictly on these polls.

    These are just some of the variables I can think of with respect to the recent WAP and PC polls.

    1. The Dycap polls were paid for by the WAP.

    2. The Ipsos Reid poll was paid for by the PC's.

    3. The Dycap polls were only conducted in Calgary-Glenmore.

    4. The Ipsos Reid poll was conducted city-wide.

    5. There isn't a long enough election/by-election history of the AAP/WAP to try and chart trends accurately.

    6. People do lie to pollsters...whether to be polite or just to get the pollsters off the phone. (For example, in 2004, according to polling numbers, a particular candidate should've won by approximately 500 votes. On Election Day, it became very clear that 1200 people who said they were going to vote for that candidate did not show up to vote.)

    7. Lots of people plan to vote on election day but other stuff comes up so they don't end up making it out to vote.

    8. By-elections campaigns/results are the most unpredictable.

    9. Voter turnout in by-elections tends to decrease significantly from general elections.

    10. There is no incumbent in the Calgary-Glenmore by-election race.

    In this particular instance, I think these polls are interesting in terms of building institutional knowledge and historical data, but certainly not for trying to predetermine a win or election results/predictions.


Anonymous comments are discouraged. If you have something to say, the rest of us have to know who you are