Saturday, September 09, 2006

Recent Poll Results: Do They Show Anything?

By now, most hardcore bloggers will have heard about the Decima poll, which supposedly shows us lots and lots of interesting things. To that, I simply say, hogwash. Let's take a look at some of the potential conclusions that some people might have reached from reading those results. (Note: this blog is partially a reproduction and expansion of comments made on

1) Bob Rae is, at least in terms of popular support, the front runner in the leadership race.

This is a poll using a sample; it's an estimate of support among the population, not the actual value. Using standard hypothesis test values, there is no statistically significant difference between the results of Rae, Dion, and Ignatieff. If you tested them, you’d find that they’re exactly the same.

You can just take numbers from polls and believe that they’re the population values. If you do, you have no knowledge of statistics, and you shouldn’t be claiming to interpret documents of a statistical nature.

2) At least in Ontario, Rae is way ahead of Ignatieff.

11 points WOULD be statistically significant if the sample size were 1,000, but it’s a subset of the earlier, it’s it’s probably more in the 350-400 range. Once again, huge confidence intervals. A 400 person sample has the capacity for wide divergence from the population value, so you shouldn't overinterpret the results.

3) But Rae did poll ahead of Ignatieff. It's not a sure-fire sign that he's the leader, but it might be an indication that he has a lot of popular support.

Ok. Let’s assume that the difference was indeed bigger. Let’s assume that Bob Rae is way ahead in the race, at least as far as your average voter is concerned. What’s this based on? Is this based on a reasoned analysis of the issues or the candidates? No. This is entirely based on name recognition. So much of what happens in politics is based on name recognition, and certainly Bob Rae has that.

For most Canadians, the only political figures who are household names are the Prime Minister, their premier, and occasionally the leader of the largest opposition party.

If you look through poll results asking people to identify the names of prominent politicians, time and time again (over the span of the past decade), nearly 20% of Canadians can’t even identify the PM correctly, almost half can’t name the Finance Minister, and roughly the same number can’t correctly identify any cabinet minister.

And now some have asserted that one of 10 candidates in the race to become the head of the opposition party has somehow become a household name because of a few months of articles in the Globe and the National Post? Come on.

In political circles, Ignatieff has become widely known and somewhat popular. Among the general population, he has certainly gained ground in terms of recognition. But there are literally millions of people in Canada who have never heard of Michael Ignatieff. That may be an unfortunate sign of political disconnect among ordinary Canadians, but for the time being, it's a fact.

4) Fine. But it does show, at the very least, that Bob Rae isn't unelectable in Ontario, as some have previously claimed.

Even this is false. Let's suppose that Bob Rae had, instead of collecting 26% of support, managed to poll in the 55-60% range. At that point, I'm willing to concede that Bob Rae is not unelectable in Ontario. With only 26% support, however, I don't think it shows his electability at all.

Receiving only 26% support may make you the front-runner, but it will never win you an election. To win an election, you have to receive as much of the remaining 74% of voters as possible. That means getting the undecideds and the moderates. I think that while Bob Rae might be attractive to a certain group (some old guard Liberals, as well as leftist Liberals and NDPers who haven't realized he's not the same old Bob Rae from days of yore), he won't have the capacity to build significant support for himself in Ontario, especially not in the 905 ridings around Toronto--and to a lesser extent the 519 and 705 ridings to the south-west and north--which are absolutely necessary to Liberal victory.

I think that someone like Stéphane Dion or Michael Ignatieff, both of whom polled marginally lower than Bob Rae among people definitely voting Liberal and considering voting Liberal, would still do better once the undecideds are included back in the mix.

So, at the end of the day, I think this poll shows very, very little. People from all camps should refrain from overanalyzing it.


Post a Comment

<< Home