Friday, June 15, 2012

Two BC polls, two very different results

In the last two days there have been two very different polls come out in BC

  • Pollster   Liberals NDP Cons Green
  • Ipsos Reid    29     48  16   6
  • Forum         20     50  19  11

The only number where the two polls come close at all is with the support for the NDP.   The support levels for the Liberals and the Greens are so far apart that statistically it is a vanishingly small area of overlap.

I know there will be many out there that will average these numbers, but that is not a valid way to view the data.   At least one company is fundamentally wrong  - the discrepancy between the two is so large that both pollsters could be fundamentally wrong.

There is clearly some sort of systemic problem going on here.

Based in the results of the by-elections, I would be inclined to think the Ipsos Reid numbers are a better reflection of the public mood.

Based on Forum polling data in various Canadian elections, there seems to be a systemic bias in their polling against the government and for the main opposition.

In the Federal election they seemed to be 3 points high for the NDP and 4 points low for the Conservatives in the last week of the election.    

In recent Federal polling Forum is the only company to get the NDP at 36%, about 3 points higher than most of the other pollsters.

In Alberta earlier this year Forum was the first pollster to indicate the Wildrose party was leading but were then insync with the rest of the pollsters.   Even with their poll the day before the election, the only one that day, they were still high for Wildrose and low for the PCs.

I do not see an issue in the data for the 2011 Ontario election.

In BC there are two pollsters that have the Liberals and Conservatives effectively tied - Forum and Angus Reid.   Meanwhile Ipsos-Reid, Mustel, Justason and NRG Research all have the Liberals at about twice the support levels of the Conservatives.

There clearly is something fundamentally different being done to have two companies consistently get one set of results and four other companies to get a different set of results.  

Post a Comment