Political Perspectives is produced by the students and faculty of Carleton University's School of Journalism and Communication, Canada's oldest journalism school.
There has been a proliferation of seat projections in this election campaign, as people try to get their heads around the NDP surge and what kind of parliament that might produce. There’s is an excellent article on seat projections in Pundits’ Guide this morning, and I want to add just a few thoughts.
For the record, I am generally a believer in the usefulness of seat projections, and continue to be so in this election, which promises to be historic in some ways. However, I also think that there are reasons to be more cautious about them in elections where there is potential structural change, and where there is volatility late in the campaign.
A few projections have caught the attention of media in this campaign more than the others. At the outset, many reporters relied on the 308 website because it was frequently updated and took account of all the polls. But the 308 model is deliberately cautious, which also means that it is less sensitive to sudden changes in party support as we have seen in this election. Because it didn’t dramatize the NDP surge story line, reporters suddenly lost interest in it.
Meanwhile a Postmedia story took a poll conducted by Ipsos-Reid (which has consistently shown the Conservatives stronger than other polls), ran it through a seat projection toy on the Fair Vote Canada website, and came up with a Mulroney-style Conservative sweep.
This was simply ridiculous, and unfair to Ipsos-Reid, which was not responsible for the projection, and possibly also to Fair Vote Canada, whose only involvement was to have put up a gimmicky little calculator on their website.
Why would Postmedia get such a strange result? First of all, like most such seat calculators available to the public, it invites you to plug in national polling numbers. As every watermelon knows Canadian elections are fought regionally. There’s a four party fight in BC, a three party fight in the Atlantic and Manitoba, practically no fight at all in Alberta outside a couple of ridings, and in Quebec a party that doesn’t exist elsewhere.
The Fair Vote projector had one ridiculous flaw, apparently, that it didn’t realize the BQ only ran candidates in Quebec. (Kudos to Glen McGregor who reality-checked his own organization’s story the next day.)
But it is not just such gross errors that matter when it comes to projecting seats regionally. The fact is that seat projections are most accurate when you look “sub-regionally”. That is, Ontario needs to be broken down into five or six sub-regions, and Quebec almost as many. The Prairie or Man-Sask regions reported in many polls need to be broken down by province because Manitoba, Saskatchewan and Alberta have very different patterns of party support.
At the moment, I very much doubt that any of the seat projections you are seeing in the public domain are using sub-regional figures for their projections in this way, though I imagine some projections will do so when it comes to the final days.
Because the New Democrat surge in Quebec is so large and unprecedented, it is mostly guess-work how broadly spread the vote is. If it is evenly spread across the province, the New Democrats might run second to Liberals and the BQ in Montreal and to the BQ and the Conservatives elsewhere. That might produce relatively few seats for the votes they get (the dilemma of the Greens nationally in the last election, and the PCs from 1993 till their merger with the Canadian Alliance). On the other hand, if the NDP votes are concentrated in particular sub-regions of Quebec, it might produce a lot of seats.
The other projection to get a lot of (negative) attention, was one done by EKOS, at the time of their first poll showing a big New Democrat breakthrough. In part this was because the projection, showing the NDP with 100 seats, was simply so startling that it was difficult to absorb. But it may also have been because of concerns about how “efficient” the New Democrat support would be in translating into seats.
The Pundits Guide article I mentioned above showed the seat projections from the last election, which had EKOS at the top. I was working at EKOS at the time (though I am no longer). One huge advantage EKOS had in projecting seats at the end of the campaign was that it had an enormous number of cases, because their IVR (robo-call) methodology allows them to poll many more people each night than conventional methods.
These large “n’s” as they are called in the trade allow very detailed sub-regional analysis. Moreover, it is possible to generate reasonable n’s even for individual Atlantic and Prairie provinces by going back a week or so. I think it was EKOS’ ability to do this at the very end that helped produce the best result last time.
However, to repeat, I doubt that anyone is using that degree of sub-regional analysis in their seat projections at the moment. That will only come this weekend.
The additional challenge this time will be that there seems to be quite a bit of volatility at the end of the election. Even polls taken Friday or Saturday, on which the projections will be based, may be outdated by Monday. (Late movement in the 2004 election confounded many pollsters’ seat predictions, and most called the result — a Liberal minority — wrong).
The Pundits’ Guide article suggests that in the past projections have tended to understate the seats for the parties that are growing, and overstate the seats for those that are shrinking. This might mean that the current projections for the NDP understate their potential wins and overstate those for the Liberals and potentially the Conservatives.
Perhaps. But in this election we don’t really know whether the NDP will still be in surge mode on Monday, or shrinking due to buyer’s remorse under the late-campaign attacks of opponents.
What the seat projection models are giving us is a sense of what kind of outcomes are possible — and some are quite different than what we might otherwise have expected. But at least this year, don’t shut down your RRSPs to bet in the office pool on the basis of any of them.
Paul Adams is an associate professor in the School of Journalism and Communication at Carleton. He is a former Parliament Hill reporter and worked in the polling industry. You can follow him on Twitter @padams29