May 15, 2005

Polls Translated into Seats

In the final days of the 2005 election campaign, BC has seen at least three published polls. On May 10th the Mustel Group released its poll, taken between May 5th and 9th, which showed the Campbell lead over the New Democrats had shrunk to 5%. With a 4% margin of error, that means that the momentum was in favour of the NDP and it was close to a statistical tie. On May 13th Ipsos-Reid released its poll which revealed an eight point lead for the Liberals, the lead that they enjoyed in Ipsos-Reid polls since the beginning of the campaign. On May 15th the Globe and Mail published a poll from Strategic Counsel which had the Liberals 13 points ahead of the NDP, 49% to 36%.

After the votes are counted on May 17th the three polling firms will do doubt claim that they were all accurate. They will argue that polls are a "snapshot" at a particular point in time; the election is at a different time so any discrepancies are due to changes between the time of the poll and the time of the election rather than being due to badly conducted polls. So why did the three polling firms produce such different results?

Whichever poll one prefers it is very difficult to take a measure of province-wide percentage vote and translate it into the number of seats each party might win. In 1972 the NDP formed government with 39.6% of the vote and 38 of 55 seats. In 1979 the NDP became the official opposition with 46.0% of the vote and 26 of 56 seats. The province-wide vote doesn't transfer into seats because there are as many separate elections as there are constituencies, 55 in 1972, 56 in 1979 and 79 in 2005.

Campaign managers for both major parties have more detailed information in their unpublished polls and as a result of canvassing riding by riding, but not even the candidates are allowed to see that information. In the absence of hard data, pundits try to determine the seat count based on past performance and their assessment of local campaigns. One of the best pundits and most detailed seat counts is from Will McMartin writing for TheTyee.ca. On May 14th McMartin said there are 31 solid Liberal seats, 15 likely Liberal seats, 8 seats up for grabs, 10 likely NDP seats, and 15 solid NDP seats. On Saturday evening Global TV reported that McMartin would change his forecast to 51 seats for the Liberals and 28 for the New Democrats. Although there are a couple of seats where I might quarrel with McMartin, I agree with his analysis on the whole.

McMartin's forecast, combined with the Ipsos-Reid poll, which many would argue is the most reliable, leads me to my prediction. My objective assessment has the count at 24 seats for the NDP and 55 for the Liberals. The Greens are unlikely to win a seat. If the NDP finishes with over 20 seats on May 17th, Carole James should be praised for building on the work done by Jenny Kwan and Joy MacPhail and bringing the party back from a near death experience. Unlike the Socreds who took two terms to metamorphose into the Liberals, leaving the skeleton of a once proud party behind, the NDP will demonstrate that it is as much a movement as a party. Of course elections are full of surprises. Some voters may make up their mind in the election booth; some Greens may decide they want to vote strategically and not help a Liberal victory; many Liberals may stay at home, confident of another Campbell sweep. "It ain't over 'til it's over"; Tuesday night's returns will be very exciting as the voters reveal what they want while pollsters and pundits rationalize errors in their forecasts. A final seat count of 38 to 41 might make for the best government, regardless of which party won!