So the Conservatives win another minority, as expected. But it’s a little stronger than most, including DemocraticSPACE, projected. As we noted in our commentary, there were a wide range of possible outcomes due to the margin of error in the projections, with the best-case scenario for the Conservatives being 146 seats and the worst-case for the Liberals being 74 seats. The vote broke just about as well as it could have for the Conservatives, allowing them to narrowly pick up a lot of seats, and run away with it in others. So they came in at the top end of our projection range, and the Liberals came in at the bottom. The NDP and Bloc came in as expected, near the middle of the projected range.

Assessing our accuracy, of the 308 ridings, we made calls in 266, of which 258 (97%) were correct and eight (3%) were incorrect. We also had 42 ridings as “too close to call”, meaning there was an overlap in the projection range for more than one candidate, meaning either candidate was equally likely to win. All the model tells us is that it is going to be very close. To ensure we had a tally that added up to 308, however, in our average projection, we gave the riding to the candidate with the highest average projection, even though we did not make a call. Of these 42 ridings, the projection average correctly put the winning candidate ahead in 25 cases and put them behind in 17 cases, so a 60/40 split. So, indeed, these races were too close to call (although in some cases, the race was not nearly as close as we projected).

So, to sum, where we made a call, we were 97% correct. Including all too close to call ridings in the tally, we were correct in 283 ridings and incorrect in 25 (17 that were too close to call, and 8 genuine surprises), for an overall success rate of 92%. Interestingly, this is exactly the same result we had in 2006 (283 right, 92%), even though the average seat projection was closer to the results in 2006. So at least we’re consistent if nothing else…