As Parliament Hill transitions from 10 years of a Conservative government to the newly elected Liberal one, one may wonder if the Liberals won their surprising majority thanks to strategic voting.
We need to be very clear and precise when talking about such concepts. Looking at the polls, it seems fairly obvious that some NDP voters switched to the Liberals in the final days of the campaign, especially in Quebec. What we are really interested in, however, is looking at the ridings that were targeted by organizations such as Lead Now or Ali Kashani.
These groups made recommendations in some ridings — which brought some controversy in the case of Vancouver–Granville — and even went as far as polling some of them.
But did it actually make a difference?
Many parties, including these organizations, were quick to declare that the answer was a definitive yes. The problem we have is that they are looking at the results and concluding, because "their" candidates won, that the results were proof that their actions worked or had an effect. To really conclude this, however, we would need to observe what would have happened had Lead Now or Kashani not made any recommendations or done anything. In other words, it's well possible a lot of these candidates would have won anyway, especially the Liberal ones.
To answer the question, we'll compare the results in these ridings with our projections, using the actual percentages of vote of the election (thus eliminating the error we made because of the polling errors in Quebec, for instance). While we acknowledge that the model made mistakes (and a thorough investigation will be necessary), let's also remember that with the correct vote percentages, the model would have correctly predicted a Liberal majority (albeit a smaller one than actually happened) as well as the correct winner in more than 80 per cent of the ridings. If these organizations really had an effect, we should see a systematic effect in favour of the candidates they recommended.
So we took the 29 ridings of Lead Now and the 16 of Kashani (two ridings were chosen by both groups, thus leaving us with a data-set of decent size with 43 ridings) and we compared the official results with our projections (which, in this case, should be seen as what we'd have expected to happen had there been no strategic voting in the ridings).
Out of the 29 ridings where Lead Now made a recommendation, only two — Calgary Centre and Kootenay–Columbia — had a different outcome than we would have predicted.
In, Calgary Centre, riding polls were indeed showing the Liberals much higher than our model, so this is one example where we should have trusted these polls more (although in general, they didn't do much better). Their results were really representative more of the general gains of the Liberals in Calgary than anything else.
The Liberal candidate got 13 points more than projected. The gain was made almost evenly at the expense of their three main competitors. The Tories were three points too high in the projections (although a campaign like Lead Now shouldn't affect this number), the NDP were five points too high and the Greens also five points too high.
So this is one riding where it looks like Lead Now's recommendations might have had an impact and the anti-Tory vote gathered around one candidate.
Kootenay-Columbia is the other one. The NDP candidate narrowly defeated the Conservative one and we'd have projected otherwise.
Wrong recommendation in three ridings
Lead Now actually made the wrong recommendation in three ridings, included the now infamous Vancouver–Granville one. We also need to realize that many of the ridings for which Lead Now made suggestions were not at risk of vote splitting. Out of 29, only five ended up really close. In many cases, the winner (in particular the Liberal candidate) won easily and were projected to do so. For instance, we have to wonder how the New Brunswick riding of Fredericton could ever have been chosen as vulnerable to vote splitting and where strategic voting could be needed.
The Conservatives did end up winning only four of Lead Now's 29 ridings, which is why some are calling strategic voting a success. We would argue, however, that the Tories would have won only six of them anyway.
Ali Kashani's recommendations fared better. His data-driven choice of ridings seems to have identified potential ridings better. With that said, his campaign for strategic voting drew a lot less media attention than Lead Now's did. He also recommended the wrong candidate in two ridings in British Columbia.
Out of the 16 ridings he identified, the Conservatives won only five, compared with 10 they would have got had the projections been correct. It should be noted that one — Haldimand–Norfolk in southern Ontario — was actually projected to go to Liberal but ended up turning Conservative. Out of the six ridings where the Tories were projected to win but didn't, they fared worse than expected in every one.
It's debatable whether a campaign to promote strategic voting should actually lead to a lower share of the vote for the Conservatives. After all, these organizations are trying to prevent vote splitting more than anything else. We could argue that such movements also helped in getting more voters out, which would have benefited the Liberals or NDP.
In these six ridings, the candidate recommended by Kashani ultimately won in four. That's a success rate of 66.67 per cent. And in two of the four successes, the race was projected tight and ended up tight, just leaning for the party he suggested. In other words, the differences observed would fall completely within the normal margins of error of riding level projections.
Testing systematically with statistical methods, we find that the effects go in the expected direction. For instance, the recommended Liberal candidates got 1.4 points more than expected and the NDP candidates 2.2 points less (in ridings where a Liberal candidate was recommended of course). That total difference of almost four points is, statistically speaking, not significant.
In other words, in ridings where either Lead Now or Kashani recommended voting Liberal, the party's candidates did not outperform the projections significantly. But there is a measurable effect. The lack of significance could be attributed to the small sample size.
The recommended NDP candidates saw similar results, although the effect is significant. In that case, the NDP candidates seemed to have enjoyed a boost of about three points while the Liberal candidates were mostly not affected.
In any case, significant or not, it means the effect of such strategic campaigns would be less than five points. As we said during the campaign, such strategic voting really requires too much information and co-ordination to be really successful. It can, at best, make a difference in very close races. But as we saw with projections, identifying these races isn't a simple task, especially if there is a last minute momentum for one party.
All in all, this analysis shows that the strategic voting organizations didn't really have an impact on the electoral outcome. That doesn't mean strategic voting didn't play any role at all. We believe that some NDP voters ultimately switched to the Liberals in order to defeat the incumbent Conservatives. Given that the majority of the population was expecting a Liberal victory anyway, the switching might have had more to do with a desire to have a majority and no election (and instability) for a while.
We have shown skepticism for organizations trying to target ridings to motivate people to vote strategically, and these results just reinforce our opinion. Lead Now in particular should choose a better method to select the ridings.
All of this might be moot if the Liberals do indeed get rid of the first past the post electoral system by the next election.
That change would likely make strategic voting unnecessary.
ALSO ON HUFFPOST:
Rookie MPs To Keep Your Eye On
Bryan Breguet has a B.Sc in economics of politics and a M.sc in economics from the University of Montreal. He founded TooCloseToCall.ca in 2010 where he provides electoral analysis and projections. He has collaborated with the National Post, Journal de Montréal and l’Actualité.
He provided analysis and updates for The Huffington Post Canada throughout the federal election campaign. For more, visit his website.