Jan 07, 2006
Reading my email takes a while because it's so distracting. For example, I've spent the last hour or so lingering over my beta invite to Newsvine, a citizen journalism website that looks like it has some legs.
Anyhow, here is the first article I wrote for the service, a throw-away about predictions in the Canadian election. I'm pretty impressed with the service, and if they can manage to display articles fairly (my new item did not appear on the front page for some reason, only on 'world') the advertising revenue portends a second career.
OK, maybe not. But I like dreaming that it does.
Here's the article:
Liberals Still On Track in Canadian Election: Election Prediction Project
Despite recent gains by the opposition Conservative Party, the Canadian Liberal Party is still on track to earn a minority government during elections January 23, according to the Election Prediction Project's current tabulation.
While most polling services interview people and aggregate the results, the Election Prediction Project invites visitors to add comments amd make their own predictions. Based on the comments, a winner in each of Canada's 308 electoral districts (called 'ridings' or 'seats') is projected.
Currently, the Liberals hold the lead in the standings at 85 seats. The Conservatives are expected to win 73, the socialist NDP 14 and the separatist Bloc Quebecois 51 seats. 85 additional seats are deemed "too close to call."
Despite its unorthodox approach, the Election Prediction project has been remarkably successful in its projections. In the 2004 federal election, it was correct in prediction 268 of 308 seats, an accuracy rate of more than 87 percent (on election day, every seat is predicted, no matter how close they are).
In other elections the service has performed even better. predictions for the 2003 provincial election in Ontario were 91 percent accurate. In the British general election of 2001 it correctly predicted an astounding 619 of 659 seats, an accuracy rating of almost 94 percent. In the 2001 provincial election in british Columbia, it peaked at 96 percent accuracy. It has never fallen below a 76 percent success rate.
Interestingly, according to the project, "The evaluation process is entirely subjective. Members of the evaluation panel have very different political background and often disagree. Predictions are undoubtedly influenced by panelists' specific insights, experiences, or biases about particular contests."
Thus, for example, while in a riding like Moncton pundits of all stripes will project a Liberal vistory, in closer ridings, such as Saint John, the projections will be all over the map. As the election approaches, there may be a sawy toward the eventual winner, which will be used as the basis for the projection.
In the current election, the Election prediction Project is projecting the potential for heavy Liberal losses in Quebec. In ,a href="http://www.electionprediction.org/2005_fed/p_35on.html">Ontario, a Liberal stronghold, there is some volitility, especially in the north, but the Conservatives have not solidified any gains, with many voters possibly leaning toward the NDP.
Pollsters are showing a much closer race, with the Conservatives even in the lead over the last few days and possibly winning the election. Any gains by the NDP are not being reflected in the polling numbers.
Pollsters in Canada, and especially Strategic Counsel, have been criticized for reflecting a bias. During the British Columbia election, in which the NDP polled much more strongly than projected, Vancouver pollster Angus McAllister criticized the sagency, saying, "The Globe and Mail ought to be responsible enough not to publish a poll where the ballot question is preceded by 14 questions that influence people's answers." Pundits have also questioned the pattern of pollsters' donations to Canadian political parties.
