Is it really back to 2008?

Looking at the recent polls, some might argue that polls (and seats projections) look a lot like the results of the 2008 election. Yes the Conservatives would get 3-4 more seats, but overall, it's awefully similar. However, people should be carefull before saying that. The fact is that during the 2008 campaign, polls and projections were placing the Tories way lower than 143 seats. Here is some evidence.

DemocraticSpace, the first website to provide good seats projections, had the Liberals at 92 seats and the Conservatives at 126 only, in its final projections. They were based on an average of polls. Myself, with a simplified model of the one currently used, had the CPC at 131 and the Liberals at 82 seats.

The "problem" was that the polls didn't indicate such a big lead for the Tories. Look at this graph, taken from this wiki page listing all the polls from 2008.


As you can see, the trend was clearly indicating a lower score for the Tories. Yes of course, some polls (like the last one from AR) were pretty right, at least nationally. But I think we should all agree: overall, the CPC performed better than expected. What happened? Well, I'm sure sure some will argue this is the "fail-CTV-interview" of Dion, but I don't buy that. If mistakes or campaign glitches had such an effect, Harper and the Conservatives should be way lower this time! (joke). I believe more than it was a differential in turnout, namely that the Conservatives voters showed up to cast ballot, when Liberals voters stayed at home. Remember than in almost every poll, you only have a maximum 25% of respondents saying they won't vote (and quite often, it's even as low as 15%), while in real life, it was closer to 40%. This is a major problem for polls accuracy. On top of that, as I already explained, having accurate national votes intentions is nice, but for seats projections, we need accurate polls in every province. Even by averaging most recent polls, we are still left with MOE of almost 2-points in Ontario, which can make a big difference.

So now, let's look at the same graph for 2011, from this wiki page.


By the way, I believe tis graph has a problem. It's a regression-based average and each poll is weighted according to its MOE. But it's considering every Nanos poll has a new one with MOE of 3-points, with is wrong since it's a rolling average. So all that to say that this graph is a little too biased towards the number from Nanos.

With that being said, the big difference with 2008 is that the Conservatives are polled consistently around 39-40%. If the same effect as in 2008 should happen, Harper would get his majority for sure. I'm not saying it will, but this post wanted to show you that it won't necessarily be "2008 all over again". One big factor that could have played a role in 2008 and will likely do again this time around is how the undecided vote. In 2008, it is possible a lot of undecided voted Conservatives. This time, who will they choose? Given that the big question seems to be whether Harper will get a majority or not, I guess it will all come down to whether these undecided want a majority and 4-5 years without a new election, or don't mind another minority government (and/or a coalition -give it the name you want, a formal arrangement, etc).

Don't underestimate the current lead from the Tories, especially in Ontario. A lot of polls at the end of the 2008 campaign kept showing Dion and the Liberals first in this province. That was one major source of bias for seats projections. Hence, even though the current lead for the CPC in this province is less than 1 month ago, it is still a lead in almost every poll. I said it before on this blog, but I would still bid on a Tories majority today. I think this party will be able to go and grab the 10-12 seats they need to win. Of course, we still have almost 3 weeks before Election Day.

Share this

Related Posts

Previous
Next Post »