Friday, July 27, 2012

Using Basic Economics to Predict the 2012 Presidential Election Results

Updated November 3rd, 2012

Now that America is into the final days of the two party Presidential election cycle, I thought that it was time to take a look at a tool that may well predict the winner of the 2012 election, keeping in mind that the data used has nothing to do with polling and everything to do with the economy with  emphasis on GDP growth levels, particularly when they are above the 3.2 percent level.  This is key after the recent 2.0 percent annual growth rate data release.

The Presidential and Congressional Vote Share Equation, developed by Dr. Ray C. Fair at Yale University, uses three economic variables that are significant to both the end result of the Presidential and Congressional elections, although the impact of the variables on the vote share derived from the House equation is somewhat smaller than it is for the Presidential equation.  For the purposes of this posting, I won't get into detail about the equations themselves since they look something like this:

Interesting, isn't it?

Rather than going through the details, I will take a brief look at the economic variables that impact the outcome of both elections and the results of the latest analysis.

Here are the three economic variables that impact the vote share calculation:

1.) G - The annual economic growth rate of real per capita GDP in the first three quarters of the election year.

2.) P - The absolute annual inflation rate in the first 15 quarters of the administration.

3.) Z - The number of quarters in the first 15 quarters of the administration in which the growth rate of per capita GDP exceeded an annual growth rate of 3.2 percent.

Notice that one of the variables (G) covers a short time horizon while the other two (P and Z) cover the entire period of the administration up to the time of the election.  The last variable, the number of quarters of high growth is the "good news" variable because it measures the number of extreme (in this case, extremely positive economic growth) outcomes that people tend to remember.

Here is a look back at elections from 1916 to 2004 showing the actual vote share received by the Presidential candidate (Vp), the calculated vote share (Vp with accent over top) and the difference between the two (Up):

You can see that, in general, the predictions are historically quite accurate with the difference between the calculated vote share and the actual vote share ranging from +0.3 percentage points to -4.3 percentage points.  Out of all 23 elections, only five (21.7 percent of the total) have an error of 3.0 percentage points or greater and sixteen (69.6 percent of the total) have an error of 2.0 percentage points or less.  I'd say that the accuracy of Dr. Fair's model is rather stunning.

Now, let's go forward and look at Dr. Fair's forecasts for both the Presidential and Congressional over the time period from November 2010 to October 2012.  Here is a chart showing his analysis:

Notice that the value Z, the number of quarters in the first fifteen of the Obama Administration with an annual growth rate greater than 3.2 percent sits at a measly one.  That's not much for voters to remember, is it?

Keeping in mind that Dr. Fair states that the margin of error is plus or minus 2.5 percentage points, notice that as the months have passed since the mid-term elections of 2010, the vote share of the incumbent President (Vp) and Congress (Vc) show dropping values with President Obama currently showing a dropping vote share and, for the second quarter since the mid-term elections, a minority of votes.  The drop over from July to October 2012 is related to the drop in G, the annual economic growth rate of real per capita GDP in the first three quarters of the election year.  

From Dr. Fair's fascinating analysis, we can see that the election will be very, very close.  His analysis suggests that it is quite possible that the final vote count will show the incumbent losing the popular vote but winning the necessary electoral votes to take the prize.


  1. If things bear out as they did in 2000, many will take to the streets. There will be no alternative.

  2. Interesting. However, the model suffers from the same issues as the Colorado model. What if Romney substitutes his dog to run instead?

    The Colorado model has Romney winning, cough, cough, all of the swing states including OH and PA. Not gonna happen.

    I enjoyed the overview. Thanks!

  3. In an election that will have the candidates less than 2% apart in popular vote, not to mention that the electoral college will actually determine the Presidency, this model has as much value as a can of air.