All I can say is that working from the data I had is why ended up where I did. ARS seemed to get the mood right and that this mood was in place from before the election started.
The low voter turn out really caught me off guard, it is where most of the error seems to come into it. As I pointed out several times, I had some major issues with the turn out numbers the pollsters were suggesting. In averaging them all and using the weighting I did, I ended up too high for the Liberals, way too low for the NDP and too high for the Greens. I was close on the Others.
I did make a prediction based on the ARS numbers and came up with 46 Liberal, 38 NDP and 1 Ind. So when I have the right data, I can come close, not that I was trusting their numbers.
In the last two days before the election I was doing some modeling that boosted incumbency and it gave me ranges of results that seemed 'odd' given the data from the polling companies. I ultimately did not give this as much weight as I should have when I look at the election results. Incumbency was a very strong factor in this election.
All I can do now is fondly think about my glory days from predicting the 2001 election in April 1999 to within two seats. Ok on a riding by riding basis I was wrong on four in total, but not bad looking two years forward. Think of this para as a me dealing with my wounded ego.
1 comment:
Bernard, a lot of people were saying it will be close. My prediction was 45 to 40.
There was a sense of complacency, especially among the electorate. Many Liberals did not vote because Mustel and others were saying the spread is 10 to 17%.
It just goes to show the limits of polls. Even ARS was wrong because either it was forecasting (which it should not), or it was saying the spread is 2% when it was 8% or more at the time of the ARS poll.
ARS manipulated the election and got its way to fire up the NDP zombies. Mustel manipulated but got it wrong and caused complacency.
With friends like Mustel, who needs enemies?
Post a Comment