David Strömberg

WH 2008 Analysis, Oct. 9

Phone: +46 8 16 43 77
Fax: +46 8 16 14 43
E-mail: info@iies.su.se
Webmaster:
www@iies.su.se

Last updated,

 October 1, 2008

How the probabilities are estimated

From the previous analysis on October 1. Obama's chances of winning have increased from 79 to 85%. Around half of the improvement is due to more favorable polls, the other half is due to reduced uncertainty. There is less time left for for campaign-changing events to take place. (The errors in the historical out-of-sample predictions are getting smaller as I include opinion polls closer to the elections). One standard deviation in the national prediction error is now 2.5 percent compared to 3 percent September 15.)

 Obama has an 85 percent chance of winning the elections, receiving an average vote-share of 53 percent and an average 330 electoral votes. The risk that the president elect will not receive a majority of the popular votes is around 7 percent. It is more likely that Obama will be in this position (4.2 percent) than McCain (2.5 percent). Consequently, Obama benefits slightly from having the Electoral College system (this increases his chances of winning by 2 percent).  

Obama's 85 percent chance is higher than what you might expect looking at betting markets. The probability can be understood simply by looking at the distribution of the historical national-level out-of sample prediction errors. Two elections, in our sample of 15 elections, had out-of-sample pro-Republican national shocks larger than 2.6 percent. These elections were 1960, when Kennedy had a 49-45% lead in the polls at this stage and yet the election was very close, and 2000. (In two elections there were pro-Democrat shocks of more than 2.3 percent, 1968 and 1976). Now, and 2/15 is around 13%, so that would be McCain's odds.

In the model, I assume that the errors are normally distributed. One standard deviation in the those errors is around 2.5 percent in the vote share. Assuming normality, this means that the chance of a 2.5 percent pro-Republican shock in vote shares is around 16 percent. 

Intuitively, to counter the pro-Obama effects of the current banking crises, an offsetting shock must be something comparable in size, but working in McCains favor. Although I can think of such shocks -- a terrorist attack on American soil, or a devastating debating performance by Obama -- they don't seem very likely.

The table below show the forecasted democratic vote shares, win percentages and the share of candidate visits that should go to each state, based on opinion polls available October 9.