The problem with Atlantic polling breakdowns

When Atlantic Canadian numbers are broken out of a national poll, they begin as suspect because their sample is so small it likely has a margin of error that is greater than 10%. But that is just the beginning of the problem.

Atlantic Canada is not a homogeneous region. In all elections from 2000 forward (and to a lesser extent 1997), NDP poll numbers are misleadingly low. This is because while the NDP polls well in Nova Scotia and ok in New Brunswick (thanks to Yvon Godin scoring over 50% in his riding), they have been in the toilet in PEI and Newfoundland. Thus, any substantial numbers from Nova Scotia are washed away and muted.

This election will be much worse. I wouldn't be surprised if the Tories were polling below 15% in Newfoundland & Labrador thanks to Danny's popularity and visceral opposition to them. Similarly, the Tory numbers are likely at 2006 levels or less in Nova Scotia due to lingering anxiety over the Atlantic Accord. What non-Maritimers fail to realize is that the Atlantic Accord would have done nothing for New Brunswick nor for PEI and in fact, some felt it advantaged NS and NL to our disadvantage.

Therefore if the Tories are polling at 35% in Atlantic Canada, they could well be at 55% in NB, 45% in PEI, 30% in NS and 15% in NL. Even with a proper sample, Atlantic-wide numbers will be largely meaningless, especially this time around.

Every federal election, CRA does a poll with a sample of about 800 per province so we can get a real picture. It will be especially important this year. I can't wait.