Friday, February 17, 2012

Four BC Polls in 20 days

In three weeks we have had four different polls conducted in BC.   This after a 2011 that saw very little polling in BC.  Here are the headline results:

Pollster      date       size   NDP  Libs Cons Green
Forum        Jan 23     (1000)   39    26   22   9
Angus Reid   Jan 27-29   (800)   42    28   19  10
Ipsos Reid   Feb 1-5    (1000)   44    32   16   7
NRG Research Feb 7-12     600    42    36   13  10

  • Forum conducted an automated phone poll
  • Ipsos Reid and Angus Reid used online panels
  • I do not know how NRG Research polled the public - I have now been told that Forum used telephone interviews
I am not really confident in what these numbers are telling us for a host of reasons:
  • The polling is not being done in the traditional manner of a real person calling the public.   IRV and online panels function in different ways and there has not been enough done to understand the systemic problems with these methods.   What needs to be done is one company conducting all three different methods at once and seeing how close the results are to each other.
  • I do not have access to the raw data to know what the questions were and what results were before weighting. I am also very interested to know how many people answered as undecided or were not intending to vote.
  • No one is polling often enough to allow us to see a pattern within any of the pollsters.  Ideally I would want to see at least one poll per company per month and hold off analyzing the polls till four polls are in.  This would allow us to see what the tendencies of the pollster is.   The only company from the above that has even conducted four polls since the last election is Angus Reid.   Their last four polls were conducted over 16 months.
  • The poll results are not that close to each other, if the polling methodology is done well, the polls should all be very close to each other.
I would like to see a poll from Mustel.  They have the longest running polling sequence in BC but their last poll came out in May last year.   Mustel seems to have dropped off the face of the world since last spring.

  • What I can take from the numbers is that the NDP has clear lead, more than strong enough to win a good majority.   50 to 60 seats seems realistic
  • The Liberals are in second and should retain a decent group of seats.   25-35 seats is the range I can see for them
  • The BC Conservatives are borderline in having enough support to be certain of winning seats.   In the mid teens the BC Conservatives could win as many as 10 ridings or none at all.   Everything depends on concentration of vote and we have no good numbers on this.
  • The Greens are unlikely to do any better than in 2009
I will await more polls from each of the companies and ideally some better quality polls as well.

Update at 12:40 - I was told the NRG Research poll was done via telephone interviews.   This gives a more confidence in the polling methodology, though they had the smallest poll

Update at 3:00 - I found the poll on the NRG Research website.


Anonymous said...

BC Liberals on the rise?


Take these latest four polls and lay them out in chronological order (Thanks BvS) for us to examine.

The NDP and Green party are fairly constant remaining steady over the three weeks (Jan. 23 - Feb. 12) the polls cover.

The trend is what is stunning.

The BC Liberals easily break out of the confines of the margin of error and climb 10 points (26% - 36%).

During the same time the BC Conservatives tank 9 points (22% - 13%); again outside the margin of error.

It should be noted that the BCLibs & BCCons combined totals are steady too (48%, 47%, 48%, 49%). More evidence the two right wingers are cannibalizing each other.

Bernard said...

That is not a statistically valid way to look at the polls. Trends can only be seen within the data from one pollster, not multiple pollsters.

I am not convinced of any trends for certain.