Tuesday, October 7, 2008

Politics: Predictions? No Clue

TORONTO, ONTARIO - A striking difference between the ongoing elections in the United States and Canada is the quality and availability of the polling data available. While I will be the first to dismiss poll results in favor of the one poll that really matters, the net impact is that US citizens have a much better idea how their election, four weeks away, is going to turn out that Canadian citizens do about their election, just one week away.

The proliferation of web sites providing meaningful polling information in the United States is rather stunning. Individual pollsters have their own sites, and then there are aggregation sites like Electoral-Vote.com and Real Clear Politics; my personal favorite is the left-leaning FiveThirtyEight.com, a reference to the total number of electoral votes in the presidential election. The "Super Tracker" in the right-hand column of FiveThirtyEight.com summarizes quite well the long story of the election and Obama's presently-widening lead.

Of course, the Presidential race, even if viewed as about fifty separate races (Maine and Nebraska's apportionment of electoral votes in a non-winner-take-all fashion make it more than just 51) is not really the correct analogy for Canada's parliamentary election, which is really 308 separate elections in individual ridings. The closer analogy in the United States is the election of the House of Representatives, with 435 separate races. Granted, part of the difference is that political gerrymandering and the power of incumbency mean that most of those seats are quite safe, but even for the House, there are fairly predictive polls. Real Clear Politics considers only 50 of 435 races worth watching, and only 17 of those as toss-ups--indicating a certain majority for the Democrats. Electoral-Vote.com similarly picked only 59 races worth tracking. ElectionProjection.com projects 246 seats for the Democrats when only 218 are required for control. Short of enormous change in the political atmosphere, US citizens already who will likely control Congress.

In contrast, Canadian polls tend to track party preference over the entire country, such as the well-known NoDice.ca poll aggregator. This is a nearly meaningless metric, as even at the provincial level (much less the riding level) the only party that actually has somewhat even support is the Green Party at 10% across the nation. While it would be surprising if the first-place party in this kind of poll did not form the government, it does little to say if that party will form a majority or not, which seems to be the biggest question for pundits in this election.

A few sources do try to look at what really matters--the results, riding by riding. Probably the best of these is at ElectionPrediction.org, which of as today is projecting 118 Conservative seats, 71 Liberal, 37 Bloc Quebecois, 26 NDP, and independents Bill Casey and Jean Paradis hanging on to their seats. But, they show 54 seats as too close to call--enough to give the Conservatives a clear majority, or (much less likely) give the Liberals a plurality and hence minority government.

People who wish to put their money behind predictions can participate in the University of British Columbia's Election Stock Market. While the raw party numbers there look much like the actual polls, they have an interesting market in who will form a majority--and as of today, 80% of investors (and rising) think there will be another minority government, and only about 20% are betting on a Conservative majority.

The bottom line is that those who would like to tell the public what is going to happen on Tuesday don't have an especially clear idea, unlike in the United States. There is ample incentive, particularly for those in the 45 key ridings and others with less-expectedly close races to get out and help your favorite candidate. Certainly, there is ample reason to get out and vote next Tuesday.

No comments: