Wednesday, June 23, 2010

Slight Methodological Change

I have decided to put the estimated popular vote gap between the top two parties on the left sidebar. I'm using the opportunity to refine my methodology for calculating popular vote averages. The specifics are a little complicated and somewhat arbitrary, but the end result is this:

1. Numbers posted after today are not directly comparable to numbers posted previously.

2. To approximate a fair comparison, you would multiply the previous numbers by 1.04. For example, the gap in the latest projection was 5.7% under the old method, and is 5.9% (approximately 1.04*5.7%) under the new one.

I don't routinely post poll averages because I don't feel they bring much value added to this blog: you can get very similar numbers from other sources (unlike seat projections, there isn't much variation other than resulting from the time frame considered), or easily compute them yourselves. But for what it's worth, I currently have: CON 34.5, LIB 28.6, NDP 16.7, BQ 10.3, GRN 8.6.

Update (April 23, 2011): I should have said a little more about what this change is about. Basically, in past elections, the NDP, Bloc and Greens received fewer votes than suggested by polls. Most of these sometimes went to the Liberals (2004, 2006), and sometimes went to the Tories (2008). What I started doing after this post was to slightly reduce the small parties' numbers (by 0.5-1 percentage points in most cases) and to proportionally distribute those votes to the Grits and Tories before projecting seats.

No comments: