The Paper Napkin Seat Model
Posted May 7, 2009 by Sacha Peter - Link
Category: Analysis
Comments (1)
I remember reading a paper from an SFU professor of political science (who is also trying to predict the outcome of this election, but unfortunately I tried to Google for his name but couldn’t find it – named Kennedy Stewart) stating a simple model for predicting the number of seats the BC Liberal Party would get in an election.
If you look at the past four elections, you have the following data:
| Popular | ||||
| Year | Vote | Seats Won | Total seats | % Seats |
| 1991 | 33.25% | 17 | 75 | 23% |
| 1996 | 41.82% | 33 | 75 | 44% |
| 2001 | 57.62% | 77 | 79 | 97% |
| 2005 | 45.80% | 46 | 79 | 58% |
It turns out that the popular vote of the BC Liberal Party is linearly related to the percentage of seats the party wins in an election. The linear coefficient in this case is 3.11 – i.e. for every 1% increase in popular vote the BC Liberals obtain in the election, they will win 3.11% more seats in the legislature – starting from a popular vote of 26%. The R-Squared value on this is over 99%, which is a very tight fit. The model obviously has constraints – for example, the percentage of seats in the legislature are bound by 0 to 100%, but within the relevant range of 33-58%, it has been eerily predictive. I think the base assumption is that if people don’t vote for the “main right-of-centre party”, that the vote would get split to some other right-of-centre party instead of the NDP.
Whether this is mathematical coincidence or whether this simple model has any predictive value is an academic argument, but I thought it would be amusing to present the “Paper Napkin Model”.
His name is Kennedy Stewart. His current prediction based on the most recent five polls is BC Lib 51, NDP 34.