Why We’ll Never Have A Canadian Nate Silver
Nate Silver is in the news again, this time for his move to ESPN to start a new blog/site dedicated to statistical analysis of a broad range of subject matter (but largely sports, it would seem). This seems like as good an opportunity as any for me to write something I’ve been meaning to write for a while, which is why you can’t do what Nate Silver (and Sam Wang, etc.) does in the U.S. here in Canada. It’s one of the reasons pundits keep getting election forecasts so wrong and one reason why you should probably ignore Canada’s cheap Nate Silver rip-off, Three Hundred Eight (who, like the pollsters and pundits, has badly missed a number of recent election calls).
To explain this we need to look at what Silver and others like him do, which is basically to aggregate polls to increase the sample size of respondents. Averaging the poll results seems to create pretty good indications about who will win a given Senate seat or who will win a state in the electoral college (though as a number of people have noted, many states can be predicted using even simpler methodology like “who won this state last time”).
The problem is, there’s no way to do this in Canada. In the U.S. there are many polls to refer to at the state level, which is the relevant electoral level. But in Canada results in federal elections aren’t determined by who wins the popular vote in a given province, but rather at the local riding level. And there are few if any polls done at the riding level. This holds true for provincial elections as well. This is important because the popular vote across the country and the actual seat count never line up. The NDP have a vastly higher proportion of seats in Quebec than their popular vote share, and the same is true of the Conservatives in Saskatchewan (and so on). Because we don’t have polling data on the level that elections are decided at (the electoral district), you simply can’t do what Nate Silver does for Canadian elections. And unless someone is willing to spend an awful lot of money doing riding-by-riding polls, you’re never going to be able to.
