The following poll has been added to the model:
EKOS' 9/18-22 national poll (current weight among national polls: 6%)
For a full list of included polls, see previous Projection posts.
This EKOS poll is further bad news for the Liberals in QC, which moves the projection slightly. In ON, it's mixed: good numbers from an absolute standpoint, but EKOS had been producing very Liberal-friendly numbers in ON.
EKOS says that the "blackface" effect appears to be receding after being initially severe. Suppose the effect was bad 9/18-20, but went away on 9/21. This is consistent with the Nanos tracking (9/18-20 being the Liberals' worst period). It also reconciles the DART (9/20) and Campaign Research (9/18-19) polls being very poor for the Liberals with the Abacus (9/18-22), EKOS (9/18-22) and Angus Reid (9/21-23) polls producing milder results. This theory does not explain, however, the trend observed by Mainstreet (still worsening as of 9/20-22) or the good 9/19-21 Liberal numbers from Forum. Nevertheless, given the noisy nature of polling, the hypothesis seems plausible to me - it's normal that a couple of polls wouldn't quite tell the same story.
I should mention that the model simply takes the midpoint date of polls, and does not attempt to infer the numbers from individual nights. So for now, the effect of "blackface" isn't receding in the projection. We shall see what the next few days hold...
Projection as of the latest national poll (midpoint: September 22)
CON - 151.3 (36.5%)
LIB - 145.4 (33.0%)
BQ - 18.9 (5.1%)
NDP - 17.8 (12.5%)