Forecasting the Canadian Election
The following is a guest post from �ric B�langer and Jean-Fran�ois Godbout.
Canadians will go to the polls on May 2 for a third time since 2006. Although many Canadian bloggers are offering their votes and seats projections during the election campaign (see, for instance, ThreeHundredEight.com), our model is currently the only forecasting tool to predict the outcome of federal elections in Canada using political and economic factors, much like models of American elections (e.g., here).
In this article, we forecast the level of electoral support for the incumbent party in the 19 Canadian federal elections since 1953. Our model is based on three factors: the unemployment rate, the popularity of the incumbent, and the longevity of the government. (The first two factors are measured three months before election day.) The model provides a vote share prediction, which can then be used to estimate the proportion of seats going to the incumbent party in the Canadian House of Commons.
The model offers a very good fit to the data (r-squared=0.70), although not as good as some of the better forecasting models found in the United States or the United Kingdom. The model's mean absolute error is of 4.4 percentage points in incumbent support. In close to half of the cases (seven), the difference between the predicted and the actual incumbent vote share is less than 3 points - the typical margin of error attributed to commercial polls. On the other hand, five other election outcomes are predicted with an error of over 8 percentage points.
For the upcoming election, the unemployment rate is 7.7% (average of November-December-January), the incumbent party's popularity is 38% (average of eight published polls during the month of February), and the longevity measure is equal to 63 months (logged value = 4.143). Hence, the model's forecast gives 34.4% of the vote and 114 seats out of 308 to the Conservative Party.
Our model thus predicts that a minority government will again be elected by Canadians this coming May. To put things in context, the Conservative government currently holds 143 of the 308 seats, while the Liberals have 77, the NDP 36, and the Bloc Quebecois 47 (there are also two vacant seats and 1 independent MP). If we consider the model's margin of error of 4%, we predict a Conservative minority government with 30-38% of the vote and 33-41 percent of the seats, which would mean a 5 percentage point decline for the Conservatives in seat numbers assuming the upper bound of our prediction.
However, it is still possible that the government will be formed by a Liberal minority, or by a coalition of Liberal and New Democratic Party members. Indeed, our model cannot predict the politics of government formation under a minority scenario, meaning that we cannot predict the actual composition of the government and how the opposition will be fragmented (unless of course the incumbent government is predicted to obtain more than 50% of the seats).
This forecast appears lower than what early campaign polls have been reporting so far (with a few even indicating a possible Conservative majority). We believe that the main reason for this difference is the weight of the economy in our model. As discussed in our article, among these three factors, unemployment most affects the election outcome. Historically in Canada, when unemployment is high, popularity stands around 35-40 and a government has been in power for a few years (say 5 years), the incumbent is expected to lose votes and seats. Unemployment in 2008, a few months before the previous federal election, was around 6.1%. Now it stands at 7.7%, thanks in part to the financial crisis.
These observations may appear surprising to people who are following Canadian politics. Indeed, the Canadian economy has performed relatively well since the 2008 recession, especially when compared to the United States and European countries. This actually forms part of the Conservatives' rhetoric in the current campaign, with their emphasis on the relative success of their "Economic Action Plan" and their claim that a change in government could jeopardize the recovery.
It is important to note, however, that our comparison is strictly historical and based on Canadian indicators. Our model is not comparing the economic performance of several different countries. Had we included a comparative measure, perhaps our forecast would have predicted a stronger level of support for the Conservative incumbent government. Normally, according to the model, incumbent governments are predicted to lose votes when a federal election is held in the years following an economic recession; the examples here of the Liberals in 1984 and the Conservatives in 1993 come to mind. Still, these previous recessions were much more damaging to the Canadian economy than the most recent one. We shall see how important a role the economy plays in the outcome of the current election.
Powered By iWebRSS.com
