So, which pollster had it right?
I recorded the last polls of the election here.
Here they are compared to the results.
Source | Liberal | PC | NDP | Green | Dates | Details |
RESULTS | 37.6% | 35.4% | 22.7% | 2.9% | Oct 6 | THE ELECTION! |
Angus Reid | 33% | 36% | 26% | 5% | Oct 2-4 | Online; Sample: 2223; 2.1% 19/20 |
Ipsos-Reid | 41% | 31% | 25% | Sep 30-Oct3 | Sample 1020; 3.1% 19/20 | |
Abacus | 37% | 34% | 24% | 4% | Oct 3-4 | Sample 1001; 3.1% 19/20 |
Abacus has the best call, perhaps because their sample was concentrated closest to election day.
The huge Forum Research Poll, also under that link, didn’t report percentages, but did make a seat call of 45/45/17, which was damn close, and nailed the NDP count. Given that that poll Forum conducted the poll over the weekend, days away from the election, the reason for the discrepancy is clear.
Ipsos-Reid was way off, but also was working with the oldest data, and a smaller data set sprawled over many days.
For those who don’t know, the final count was 53/37/17, the largest minority possible.