Who will win the US 2012 election?

August 16, 2012
By

Predicting the results of the 2012 elections  is both a complex science and an art.  There are so many variables involved and the dynamics of the situation change almost every day in this closely fought, heavily funded battle between two strong contenders.

It is a science, because, too a large extent, predicting election results requires processing largeamounts of data: national polls, electoral votes, popularity ratings, twitter ratings, the state of the economy, policies, ideology and above all else, the likeability of each candidate. There are  twists and turns to the campaign, such as the announcement of a running mate, or, the 24 hour news cycle. All candidates make mistakes, and with the superpacs in action, there are daily negative ad campaigns.

Some of these factors can be quantified, such as opinion polls. Many others cannot. Hence predicting the results is very much an art and the personal bias of the observer is an ever present problem. Can you trust the blogs from Washington Post, or should you depend on Rasmussen reports or Fox news? While personal bias cannot entirely be eliminated, we believe it can be minimized by gathering information from a variety of sources; hopefully, the biases will cancel each other out.

Also, election prediction requires a simple model that is easily understood and which does not try to accomplish too much. In this post, we propose a simple model that has only one purpose:to predict who will win the election. The margin of victory doesn’t really matter in the long run – there can, after all, be only one president at any one time. In the US system, the winner takes all.

At We Canadians, we believe we are well qualified to carry out this type ofanalysis. Canadians living in Canada do not vote in the US elections, so we are neither Republicans nor Democrats. Of course, we observe the election with great interest, because whatever happens in the USA will, ultimately, impact us.

Prediction Model

Our model is, in essence, very simple. We consider a number of factors, or criteria, which we believe, will affect the election results. Selection is limited to factors that can easily be quantified – such as polling results. We have compiled these criteria from a variety of sources; with a wider net, individual biases, will tend to even out.

To each criterion, we have   assigned a weighting factor, based on our best   judgement about  its relative importance. For example, we believe that the electoral vote is more significant than national polls, because, ultimately, the electoral count will determine the winner. There have been instances where a candidate who trailed in the national polls ended up as the winner. Favourable ratings are important factors, as they include candidate likeability, and this requires a higher  weighting factor.

To keep the math simple, all the weighting factors  add up to 1.00. Each numeric indicator (a percentage) is now multiplied by its weighting factor to produce another percentage for each candidate. These percentages are added up for each candidate, to produce an overall rating, as a percentage. The candidate with the highest overall rating is predicted to be the winner, at the time of the analysis. Of course, these numbers change every day and the prediction is only valid for the day it is made.

As the election date approaches, we will refine the model by adding more criteria, and more data sources to improve prediction accuracy. Currently, the model indicates that Obama is leading the race by a substantial margin.

Indicators Data Source

Weighting Factor

Obama Rating

Romney Rating

Obama Composite Index Romney Composite Index
National Polls Average RealClearPolitics

0.1

47.3%

43.8%

4.7%

4.4%

Wikipedia

0.1

47.7%

43.0%

4.8%

4.3%

Electoral Vote CNN Electoral Map

0.2

45.9%

38.3%

9.2%

7.7%

RealClearPolitics

0.2

44.1%

35.5%

8.8%

7.1%

Favourable ratings RealClearPolitics

0.2

49.4%

42.9%

9.9%

8.6%

Twindex twitter polling index

0.1

27.0%

16.0%

2.7%

1.6%

Fund raising CNN

0.1

75.0%

100.0%

7.5%

10.0%

Total

 

1

47.6%

43.6%

 

 

Last updated: August 17, 2012

 

 

Leave a Reply


Hit Counter provided by Los Angeles Windows