Wednesday, August 20, 2014

The polling plight of the little parties

The inclusion of parties in Canadian polling is quite standard. There are the Liberals, New Democrats, Conservatives (or whatever version of them exists at the provincial level), Greens, and the large Quebec parties. But what about the fringe parties in Canada?

One of the analysis topics that was requested during last year's Kickstarter campaign came from Greg Vezina, who asked me to look at the effects of polling (or not polling) the little parties. It was an interesting question, and that is the focus of this analysis.

For the purposes of defining a small party, I have included all parties that had between a 25% to 67% full slate of candidates, or a party that received less than 3% support with a fuller slate of candidates, or a party that was routinely included in polls with a slate of candidates less than 25% of the total.

I have ranked these small parties below by the average share of the vote they received in a riding in which they ran a candidate in a recent election. This was roughly estimated by multiplying the party's province-wide support by the number of seats divided by the number of candidates. 

I have also included below the average support registered in polls, when available, over the last five days of an election campaign. When there were no polls in the last five days, I used the last polls that were published.

There are a few cases that stand out, and I have highlighted these in yellow. These are small parties that were included in polls that had better performances than parties that were not included, despite having a smaller slate than the average of non-polled parties.

The Liberals in Saskatchewan and the Greens in Nova Scotia and Quebec may have benefited from being part of a national brand. But what was it about the Alberta Party, the Island Party, or the People's Alliance that gave them such a boost over parties who ran much larger slates, like the Ontario Libertarians in 2014 or the Ontario Freedom Party of 2011?

Granted, the People's Alliance in 2010 was included in the leaders' debates, and the Alberta Party got a lot of press in the run-up to the 2012 campaign thanks to a floor-crossing MLA. But nevertheless, their performance stands out from the pack.

Note also that the Evergreen party in Alberta's 2012 was included in some polls earlier on in the campaign, but not in the final stages, and is ranked here as the highest non-polled small party.

Now, the low performance of small parties and their non-inclusion in polls could simply be correlated rather than a case of causation. Perhaps these small parties were not included in polls because they were unlikely to do well or were getting scant media coverage. Or perhaps some of these small parties would have received more media coverage if they had been included in polls. But is it the responsibility of the media to give coverage to small parties that have little support, or the responsibility of pollsters to report how small that support really is, which in turn might raise the profile of those parties? It is a bit of a chicken or the egg question.

But there is certainly a difference. The average result for a party that was included in polls was 1.7% compared to 0.4% for parties that were not included. On a per-candidate basis, small parties included in the polls averaged 3.4% support, compared to 0.9% for parties that were not included. And this despite the average slate of small parties included in the polls being 54.7%, only slightly higher than the 44.9% of parties not included in the polls.

In other words, parties included in polls did 3.8 times better than parties that were not, despite running only 1.2 times the candidates.

It should be noted that when pollsters do try to gauge the support of smaller parties, they generally do quite well. And when they lump them all together in the 'Others' category, they do even better.

Electoral and polling results for 'Other' parties
So the pollsters may indeed be accurately recording the support for smaller parties already, and the effect would be minimal if they were included in all polls. 

But again, it might not be the case that being included in the polls was what boosted these parties, and rather their inclusion in the polls might have had everything to do with the support they were registering in the first place. But it is an interesting thing to consider. Should pollsters be including smaller parties in their polls if the party is running enough candidates?

In the grand scheme of things it may not matter or have a great influence on the outcome of an election. But already in two provincial elections in 2014, 347 people have run under the banner of these small parties and more than 150,000 people have voted for them. It is a question of greater importance to them.