Nova Scotia 2013 election: September 20 projections (a tale of 2 projections)

We’re now full steam into the Nova Scotia election campaign, and we’ve finally seen the first poll release of the campaign, coming from top Atlantic Canada pollster Corporate Research Associates (CRA). They polled 406 Nova Scotians between Sept 12 and Sept 18 and found the following:

	Sept 12-18 poll	Change from Aug poll
Liberal	48	+7
NDP	28	-3
PC	23	-2
Green	2	-1

Just when we thought the NDP would be making a come back with CRA’s poll conducted in August (which showed the NDP 10 points down, up from a 19 point deficit in the Spring), this poll suggests the Liberals have picked up steam over the early part of the campaign.

I have put these numbers into a spreadsheet to see how this break down would results seat-wise based on the transposed results of the last election, in 2009. The results are no surprise; the Liberals would win a majority- but not the landslide that a 20 point lead might suggest:

	Projected seat total
Liberal	28
NDP	13
PC	10

Breakdown by seat: 

Map based on these numbers

But…

A huge concern in the polling industry in this country is the under-polling of a certain percentage of “silent” voters who for whatever reason (perhaps due to not answering pollsters, lying to pollsters, or legitimately deciding who to vote for at the last minute) don’t show up in polling data. We have seen in recent provincial elections in Alberta and BC where the incumbent party went into election day well behind in polls, but ended up with decisive victories on election day. Will Nova Scotia show the same phenomena? Maybe, but maybe not. The incumbent party is behind, and is unpopular in the province. Many NDP supporters may be lying to pollsters because of the unpopularity of such an opinion. There is also a large percentage of undecided voters (22% according to the CRA poll), which means if most of that 22% go NDP, it could make the election close. Both the Alberta and BC elections also saw a large percentage of undecideds making up their mind at the last moment. However, this may be where the similarities end. Both the incumbent parties in BC and Alberta were parties who have been in power a long time, and are parties on the centre-right- perhaps an ideology closer to those people who were underpolled in those elections. The NDP however is an activist party of the left, and could be less likely to appeal to those types of voters. One glimmer of hope for the NDP on this front is, they have governed the province from the centre, so they may still appeal to silent centrist voters. But, voters may be more likely to want to ditch the NDP because they have only been in power for one term, unlike the Liberals in BC who have been in power since 1996 or the Tories in Alberta who have been in power since cavemen walked the Earth (actually, 1971). Finally, the elections in BC and Alberta saw the rise of new right wing parties that changed the electoral landscape (admittedly, the BC Conservatives faded away by the time of the election there, however the threat of vote splitting may have helped the Liberals win at the last moment). This hasn’t happened in Nova Scotia. There are no new major parties that have arisen.  Additionally, the opposition Liberals have had the same leader since 2007, who has already fought a general election. The opposition parties in BC had new leaders fighting for the first time.

Anyways, let’s pretend for a moment that the polls are indeed wrong, and that come Election Day, a large ‘silent’ portion of the electorate come out and vote NDP. In both the Alberta and BC elections, we saw about a 10% difference between the poll numbers and the actual result for the two main parties. If we apply this to the CRA poll, this gives us a 38-38 tie between the NDP and Liberals. And what does my spreadsheet say for such a result?

	Adjusted projected seat total
NDP	26
Liberal	16
PC	9

Breakdown by seat:

Map based on these adjusted numbers

Despite a popular vote tie, the NDP squeaks out a bare majority of the seats, based on the 2009 transposed results. The story behind this is that the electoral geography in the province really benefits the NDP. While the Liberals rack up massive majorities in their safe seats (like in West Nova and the Halifax suburbs), the NDP ekes out a number of wins in marginal seats across the province. Of course, when you apply the real poll numbers, all of these marginal seats disappear leaving the province very Liberal red and could result in the NDP being nearly wiped out. Of course, this is just what my spreadsheet is telling me. The real story on the ground is very different.

Methodology

An important note is that these numbers are in no way my actual predictions for the election. They are simply just taking the 2009 transposed results and redistributing each party’s proportion in each riding based on poll numbers. (I have only made an adjustment in one riding- Cumberland North due to a strong (former PC) Independent running in 2009, whose vote share I have combined with the Tories). I have done this across the province as a whole, since CRA has not published any regional breakdowns in their polls. If any pollster does do regional breakdowns, I will alter my data accordingly.

 I have in the past done actual predictions, but since my new position with a reputable polling firm, I have decided against it due to possible conflict of interest.