Oops! Looks like I derailed the discussion with my Senate reference. That comment was meant somewhat tongue in cheek — I didn't count on it spawning a whole different discussion. But have at it if you like! I'm going to attempt to steer things back on track here.
South Pender wrote:That's an interesting analysis, sj-roc. However, let me point out a couple of things.
Thanks, SP. There's absolutely nothing wrong with anything you've presented, I think we're just at cross purposes. Let me see if I can address your points below.
1. Your analysis seems to suggest that the NDP lost this election because of low voter turnout--or, more correctly, a lower percentage of NDP-leaning voters actually casting their ballots than Liberal-leaning voters casting theirs. What evidence do we have that this is the case? If this is not the case (and I see no reason that it should be), then the inferences drawn from the analysis are hard to understand. For example, if just 106 more NDP-leaning voters in one of your ridings who didn't vote had voted, then the NDP would have won that riding. However, if the same factors that kept some NDP-leaning voters from actually voting similarly affected the Liberal-leaning voters the same way, we could make the same argument that had that factor not had the effect it did, the Liberals would have had another 106 votes in that riding too. We simply don't know that a lower show-up-to-vote rate among NDP-leaning people had anything to do with the outcome. It's possible that an even higher percentage of Liberal-leaning people didn't vote simply because they thought that the NDP was headed for victory. I think that this argument advanced by many NDP pundits is simply a way of saying that the majority (or a plurality) of the population of BC really wanted the NDP, but not enough turned up to vote. To me, this is more of an attempt at face-saving than a reflection of any empirically-verifiable reality. If we try to use the pre-election poll data to make the argument, it could just as convincingly be argued that many early (at the beginning of the campaign) NDP-leaning people simply changed their minds when they saw Christy Clark and Adrian Dix in action.
To be honest, your points here, while certainly valid, exceed the scope of my aim. I was simply interested in a
post-hoc determination, with all else being equal, of what would have been the smallest amount of extra NDP votes, judiciously casted in the ten closest ridings, that would have won them a majority. That answer is about 6,500 (basically the sum of victory margins in the 10 closest ridings that went Lib > NDP). Whether it was possible that the Libs could have also gotten out more of their vote concurrently with those 6,500 is a very valid point to consider in the broader scheme but one I was explicitly not entertaining as it wasn't germane to my ends. It seems rather astonishing that a mere 6,500 judiciously cast extra NDP votes, added to the nearly 1.63M total (
i.e., only a 0.4% increase) would have fully reversed the result from a Lib majority to an NDP majority. In these two scenarios, the one that actually happened, and the other one I constructed by adding 6,500 extra votes in 10 swing ridings, would have given almost identical results for each party's share of the popular vote, and scientifically conducted pre-election polls of these two "universes" would have been very similar in spite of the two completely different outcomes. So I think this shows that projecting outcomes from their polling data is not a straightforward task: the result can be very sensitive to the details. It's not that tough to get things wrong and we saw that loud and clear. It must be tough to project seat counts from polls when the parties are sufficiently close (within about 5% as was the case in this election) in their support levels. You'd need polling data at the individual riding level and it's far from obvious to me that the pollsters actually collected such data.
Another way of interpreting my analysis is to consider the Liberals' success in how they got out their vote. There were 6,500 Liberal voters in these ten ridings who were absolutely crucial in delivering the Liberal majority. All else being equal, if those 6,500 Liberal voters had stayed home Tuesday we'd be talking about Premier Dix and *his* majority. Keep in mind the Liberals won six individual ridings by more than 6,500 votes
each and absolutely NOTHING would have changed in the riding-by-riding results if as many as 47,000 Liberal voters in these ridings stayed home. Anomalies like this are sometimes presented to argue for reform of our first-past-the-post system. My position, incidentally, is that such reforms are unnecessary because going into an election, all parties understand (or ought to understand) the rules of the game. They've been this way forever. Complaints of "not fair! that other party that I happen to dislike got a majority with only 38/42/whatever% of the popular vote! boo! hiss!" make it sound as if there were initially a rule requiring some higher threshold, perhaps 50, 55, or even 60% and that someone from on high made a unilateral decision after the fact to lower it.
In the 2008 federal election, Harper came up just 12 seats short of a majority; at the time I undertook an analysis similar to the above to discover that, likewise, in the 12 closest ridings that they just missed out on capturing, there was only the tiniest fraction of voters (compared to the total number of voters in the country) who came into play in determining majority or minority. So they didn't need a very large across-the-board bump in their support to get a majority next time out and that's pretty much what happened in 2011.
2. I did a parallel analysis of 8 ridings in which the Liberal candidate fell short of the NDP winner by the same < 1227 votes that you used. In my analysis, the margins ran from 52 (in Saanich North and the Islands) to 1137 (Cowichan Valley). Couldn't we make the same argument for an additional 8 seats for the Liberals had Liberal-leaning turnout been higher in those ridings? If both arguments were true (10 more NDP seats in your analysis, 8 more Liberal seats in mine), we have a net +2 seats for the NDP--i.e., no practical change in the standings: 48-35 vs. 50-33.
All very true, but as I've mentioned, it was outside the scope of my aim in demonstrating that the NDP was actually not that far away from pulling off a very 1996-style majority, if only they could have gotten out a few thousand more voters in those 10 ridings.
I suppose one might question whether those 6,500 extra voters were actually there to be had for the NDP in the first place (
i.e., did they exhaust every last vote from the electorate that could have been had). The fact that in 2009 they had a larger share of the pop vote from a smaller absolute turnout seems to suggest that this time around, they might have left some votes in the dressing room, so to speak. But then again voters can firmly change their loyalties from one election to another so it's quite possible that most of the subset of voters who went NDP in 2009 but not now (whether they voted for someone else or not at all) were a lost cause. Perhaps 2009 <-> 2013 comparisons of each of these ten ridings — which I have not undertaken — could provide some insights. If in each of these ridings the NDP got more votes now than in 2009 (against a background of lower overall support now than in 2009) then that would suggest they saturated their support levels in these swing ridings this time out — although it would be tougher to draw any conclusions under the opposite happenstance.
I'm wondering about that guy who nailed the US election last year (correctly predicted all 50 states as to red or blue), whether he would have fared any better than all the pollsters here who got it wrong.