Wednesday, April 29, 2009

Polling in this Election

I am trying to make sense of the Angus Reid poll and it does not fit the other data out there. Something is not right. I am also amazed that for Angus Reid the first two weeks of the campaign had no significant impact on the numbers.

I have still not see the full numbers and details so it is still hard to comment more fully.

I am going to work with all the polls and weight them based on how reasonable they are as representations of public opinion. Polls will be adjusted for size and time. This is how Nate does it over at Fivethirtyeight does it.

My weighting of the companies
  • Ipsos - 2.0
  • Mustel - 1.5
  • Angus Reid - 1.0
  • De Dutch 'Burger Poll' - 0.4
  • Robbins Research - 0.2
I am also using the Ipsos poll from last week that measured some aspects of the campaign, I am giving that one a weight of 0.6 before time discounting.

I am weighting Angus Reid low because they obviously have a huge response error problem because the amounts they are getting for other parties other than the three main parties is 100% higher than an optimistic result would have them get.

I am including De Dutch for the moment because they have reasonable spread of restaurants, they attract a reasonable cross section of the general public, and it is the only thing we have that is close to tracking poll. By weighting it at 0.4, it has a very limited impact on the calculations. I would raise this number if I knew more details about sales of burgers each day and how the sales are tracking each day.

So what numbers do I come up with at the moment?

  • Liberals - 48%
  • NDP - 36%
  • Greens - 13%
  • Others - 3%

I know many of you will disagree, there is not much I can do about that, I am trying to work on a model here that reflects what is going on.

A new Ipsos poll will have the biggest impact on these numbers.
Post a Comment