Thansk to some of my readers, I've been informed that the riding of Esquimalt-Juan De Fuca, on Vancouver island, might experience deep changes this time around. Specifically, the NDP got his "star" candidate from 2004 and 2006 back, and the Liberal incumbent is not running again. A lot of people then think the Liberals will drop because of that.
So I followed the advice of the reader "Dylan" and estimate a Randall Garrison (the NDP candidate) effect. I found out that Mr. Garrisson is worth around 5-points, taken from the Liberals. The fact he wasn't running in 2008 explains a lot why the NDP dropped so much in this riding (and why the Liberals didn't).
So I tweaked this riding manually. Instead of using the actual results of 2008, I'm using the projected ones. These are the baseline results (i.e: what the riding would looks like if there was no provincial swing between 2008 and now). See that as the results that would have happened if Randall Garrison had run. The Liberals are at 30.59% (instead of 34.18%), and the NPD starts at 26.61% instead of the actual 22.72%. I'm ok with doing these changes as they are estimated.
Then, I need to take into account of the fact the Liberals are losing a star candidate. This part is not easy as the Liberals had this candidate running there in all my sample (i.e: 2004, 2006 and 2008) and my model doesn't include (yet?) a star-candidate effect, except in rare cases. So I have to go arbitrary here. I decided to take 5-points from the Liberals. 5% is usually what people consider as the maximum a star candidate can achieve. Given that the Liberals were pretty stable in this riding, I think it makes sense not to take more than that.
So at the end, this riding is now projected to be a close race between the Conservatives and NDP, with the Liberals not far behind.
Let me know what you think and thanks for the comments.