Thursday, December 26, 2013

Forum Ontario Poll - 38% PC to 31% Lib, Wynne Minority?

Forum Research really needs to stop doing seat projections. Their most recent Ontario "horserace" poll gives us a seven-point Ontario PC lead over the incumbent Liberals, yet their "seat projections" give Kathleen Wynne's severely bruised party 51 seats to 43 PC and 13 NDP - currently there the numbers are 49-37-20. So somehow, Hudak's party has done sooo horribly in the GTA that Forum could not give him enough seats to overtake the Liberals.

Yeah, no. Forum has yet to put up their numbers for this December poll, but I have to imagine that their November poll, which showed the PCs leading 38% to 32%, cannot be too different. Either way, their seat count in that poll was 47 PC to 44 Lib and 16 NDP - definitely not what I found.

With the regional swing projection model I put Forum's November numbers into, I got a Hudak majority - 57 PC, 31 Liberal, and 19 NDP. This was in a poll that had the PCs at 33% in Toronto and 43% in the 905, and had the Liberals at steady numbers in Southwestern (28%), Eastern (34%), and Northern (30%) Ontario regions.

Now, I don't know the December numbers but I can tell you right now that to get such a disproportionate result requires quite a lot of tinkering, or an absolute, 100% collapse of the PCs in not Toronto, but the entire GTA, but with numbers rising simultaneously in every other region of the province with a complete and utter Liberal and possibly NDP collapse outside of the GTA. Not impossible, but highly improbable and I can't really get there with my model either, so I have zero idea how Forum is.

I mentioned before that I believe Forum may be taking the numbers from their samples - i.e., if they poll five people in Burlington riding, and three of them support the Liberals, Burlington is therefore marked down as a Liberal riding. If it is the case, then the sample size is way, way, wayyy to small to do anything with, and building a projection off of it is a bad idea - but I stress that I don't know what their methodology is, because they never respond to my queries about it.

It is a tremendously bad idea for a company which has taken so much flak recently to then give the Toronto Star the headline of "another minority government," specifically another Liberal minority government. Part of Forum's issue is that it just polls so much that it has become the main scapegoat for people critical of recent pollster mistakes and misses, of which every pollster has been making but Forum gets the bad rap for. When Lorne Bozinoff then decides to throw out there completely contradictory projections that probably don't have a solid base, he only digs his company a deeper grave.

For the record, though I don't have the regionals, I can do an overall swing based on Forum's topline numbers for Ontario. On topline numbers of 38% PC, 31% Lib, and 24% NDP, I got 58-28-21, which is hardly another minority and definitely not a Liberal-led one.

1 comment:

  1. 301 is the minimum sample size for an even barely reliable result, period.