Well, my model did not do as well in the 2016 Manitoba election as it did the others.
Expected: 20-25 Progressive Conservative, 6-8 New Democratic, 1-5 Liberal, 1-2 Green
Reality: 21 Progressive Conservative, 12 New Democratic, 2 Liberal, 0 Green
Expected: 22 Progressive Conservative, 0 New Democratic, 0 Liberal, 0 Green
Reality: 19 Progressive Conservative, 2 New Democratic, 1 Liberal, 0 Green
Expected: 42-47 Progressive Conservative, 6-8 New Democratic, 1-5 Liberal, 1-2 Green
Reality: 40 Progressive Conservative, 14 New Democratic, 3 Liberal, 0 Green
The results were in the projected ranges for the Progressive Conservatives and Liberals in Winnipeg, and for the Greens outside Winnipeg. However, the PC’s underperformed the projections in the rest of Manitoba, NDP outperformed throughout Manitoba, and the Greens underperformed in Winnipeg.
This indicates two potential problems my model might need to solve going forward:
1) It underestimates the incumbency advantage, which happened in the Newfoundland 2015 projection as well.
2) It overestimates the pull of small parties (though the Greens did come close to winning a seat in Winnipeg).
My model will adjust for this in the future.