Home  Blog  Election Polls  President 2012  Senate 2012  House 2012  Governors 2012  GOP Primaries 2012  Redistricting 2012  Formulas  Results  About me


    "The Blogging Caesar at Election Projection is one of the web's best prognosticators" - Hugh Hewitt

2012 Projections


Election Projection cannot screen all advertisements appearing here.  Therefore, I do not necessarily endorse the products and/or services shown.


April 2013
March 2013
February 2013
January 2013
November 2012
October 2012
September 2012
August 2012
July 2012
June 2012
May 2012
April 2012
March 2012
February 2012
January 2012
December 2011
November 2011
October 2011
September 2011
August 2011
July 2011
June 2011
April 2011
March 2011
February 2011
January 2011
November 2010
October 2010
September 2010
August 2010
July 2010
June 2010
May 2010
April 2010
March 2010
February 2010
January 2010
December 2009
November 2009
October 2009
September 2009
August 2009
June 2009
May 2009
April 2009
March 2009
November 2008
October 2008
September 2008
August 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
November 2006
October 2006
September 2006
August 2006
July 2006
June 2006
May 2006
April 2006
March 2006
February 2006
January 2006
December 2005
November 2005
August 2005
July 2005
June 2005
May 2005
April 2005
March 2005
February 2005
January 2005
November 2004
October 2004
September 2004
August 2004
July 2004
June 2004
May 2004
April 2004
March 2004
February 2004
January 2004
December 2003

  2012 Election Projection formulas

Presidential Election Projection Formula
Attributes of the Formula

I.  Purpose
This formula's intent is to give a general sense of how the presidential contest would turn out if the elections were held today.  As we move closer to election day, its usefulness as a prediction tool will increase.

II.  Objectivity
In developing the formula, I made every effort to be objective, rather than biased toward the GOP. Yes, I am Republican, but this formula does not, in any way, falsely inflate the GOP's standing.

III.  Polls
I try to be thorough and objective in gathering polling data.  I mostly browse RealClearPolitics and TalkingPointsMemo for polls.  However, there may be times when I overlook a poll.  If this happens, I urge my readers to alert me to such polls so that I can include them in the projections.

The Formula's Components

I.  Presidential Job Approval
Job Approval is the one statistic that most closely correlates to actual vote totals for an incumbent president.  As such, it is an integral part of my formula.  Up to twelve job approval polls are used, and only polls conducted less than two weeks prior to the most recent poll will be used.

II.  Adjusted National Polling Data
Up to twelve national polls are used to compute an average national margin between Barack Obama and Mitt Romney.  Only polls two weeks old or less will be used.  The average national margin is then subtracted from the national 2008 result to arrive at the national polling data adjustment.  This adjustment is then added to the 2008 margin in each state, yielding the national polling factor for the state.

III.  State Polling Data
The aggregate average of polling data is used in the following manner:

  • At most, six polls from a state are used.
  • In the interest of currency, only polls conducted within thirty days of the date of the present projection will be used.  Non-competitive states may be exempted from this time limit.  (Note: Closer to the election, this 30-day limit will be reduced to two weeks.).

IV.  The Formula in action, an example
The state and national head-to-head and job approval polling factors are combined to produce each state's Election Projection.  The weight of each factor is variable depending on the number of recent state polls available.  If no state polls are available for a given state, then the national head-to-head and job approval adjustments will each be weighted 50%.  If state polls are available, the weight of these polls will be defined as follows with the remaining weight split evenly between the national head-to-head and job approval adjustments.

The Formula in action, an example

  State Polling Weight:  20% + (Number of state polls * 10%)
  National Polling Weight:  (100% - State polling weight)/2
  Job Approval Weight:  (100% - State polling weight)/2

  Election Projection
      St. Poll (Avg*Wgt) + Nat. Poll (Adj.*Wgt) + Approval (Adj*Wgt)

Here is an example of the calculations using the formula for Ohio (2008 numbers are from May 28):

  • 2008 National Result:  Obama +7.27%
  • 2012 National Polling:  Obama +1.50%
  • National Polling Adjustment:  Obama -5.77%
  • Job Approval Polling:  Approve +48.50%
  • Job Approval Adjustment:  Obama -3.19%

  •    NOTE:  49% approval rating is considered baseline.  In other words,
          at 49% approval, Obama and Romney would be expected to win
          an equal share of the national vote.  That would represent a
          3.64% decrease in Obama's vote total in Ohio.
  • 2008 Ohio Result:  Obama +4.58%
  • 2012 Ohio Polling (4 polls):  Obama +4.75%
  • Weight:  State polls = 60%, National polls = 20%, Job Approval = 20%
  • Projected 2012 National Result: Obama 49.49%, Romney 48.99%

  •    NOTE:  1.5% is allocated to any third-party candidates.
  • Projected 2012 Ohio Result: Obama wins by 1.86%
What to do about outliers?

Until this year, I have resisted picking and choosing which polls to use in my projections.  My policy until now has been "if it's out there, it's in there."  However, I have come to believe this policy degrades the quality of my projections, not so much at the finish line, but along the way.  So I have decided to do something a bit different this year.  No longer will I take every poll and plug it into the calculations.  I will be identifying and avoiding certain polls that I consider outliers.

Now, this process will be neither hap-hazard nor driven by hunch or emotion.  As is my M.O., I have developed an "outlier test" which will objectively screen all polls for worthiness to be included in my projections. 

The Outlier Test
In order for a poll to be considered an outlier, it must fulfill both the following attributes.

  • If three or more additional polls have been released within the currency periods described on this formula page, the margin of the outlier poll must be more than 7 points outside the average margin of the other polls.
  • And the margin of the outlier poll must be more than 5 points outside the margin of the closest poll used to compute the average of the other polls.

Once discarded, any "outlier" can be reinserted if a poll or polls released subsequently cause the outlier poll to no longer qualify as an outlier when judged by the two points above.  I will use this test for all polls testing Senate, gubernatorial or presidential races, and I will adhere to it rigidly.

Senate and Gubernatorial Race Projection Formula
Attributes of the Formula

I.  Purpose
Again, the formula's intent is to give a general sense of how the Senate and gubernatorial elections would turn out if they were held today.  As we move closer to election day, its usefulness as a prediction tool will increase.

II.  Objectivity
Same as above.

III.  Polls
The aggregate average of polling data is used in the following manner:

  • At most, five polls from any race will be used.
  • In the interest of currency, only polls taken up to 30 day prior to the most recent poll of a race will be used.  For example, if a poll was conducted in Ohio on May 15th, another on May 23 and a third on June 20th, only the last two polls will be included in the projection calculations.  (Note: Closer to the election, this 30-day limit will be reduced to two weeks.)
  • If more than five polls meet the time constraint above and multiple polls were released on earliest date, the following order of "tie-breakers" will be used to determine which polls are used
       1. Non-partisan
       2. Likely voters
       3. Largest survey size - least margin of error
       4. Latest survey start date
Contested House Race Projection Formula
General Description

The formula for contested House races is substantially unlike the formulae I described above.  There are several differences between this formula and the ones used for the presidential, Senate and gubernatorial races.

First, published polls for district races are much more difficult to find than for other races, so I needed another metric which would be reliable and consistent throughout the election season.  The aggregate race ratings of four well-known political pundits is that metric.  The predictions of Stuart Rothenberg, Charlie Cook, Larry Sabato, and the Roll Call will be used.  Since their predictions are qualitative (i.e. leans, toss-up, solid, etc), I will assign a margin of victory to them as follows:

No Clear Favorite
Tilt Lean Likely/
0% 2% 4% 8% 16%

Second, when polling data can be obtained, their weight will be based on whether one poll or multiple polls are available.  If one poll is available, it will get a 25% chunk of the projection.  If two polls have been published, the average will be taken, and that result will be 50% of the projection.  Currency stipulations posted in the Senate and gubernatorial formula explanation will apply here as well.

Third, partisan polls will be used with a caveat.  The results will be adjusted by subtracting 2 points from the party of the pollster and adding them to the other party.  For example, if a Republican polling firm publishes a poll with the GOP candidate winning 54% to 40%, the poll will be entered into the calculations as 52% to 42%.

Fourth, an adjustment will be applied to the projection after the pundit rating and polls are included in the calculations.  This adjustment will be based on an average of all published generic congressional polls released in the previous week which test for likely voters.  Conventional wisdom holds that a slight Democratic lead (2 points) in the general poll average translates to a status-quo electoral result.  However, after the red wave election of 2010 which left the balance of power in the House substantially out of balance, that conventional wisdom may not apply.

In that election, Republicans enjoyed a 7-point advantage over Democrats in all votes cast for House candidates across the nation.  That represents a 9-point shift away from "a level playing field."  To compensate for this shift, I am instituting a new composite generic offset.  This year the offset will split the difference between the normal DEM +2 value and the actual GOP +7 mark from two years ago.  That baseline will be GOP +2.5 points.  At that mark, the adjustment will be zero.  Each point in either direction from the GOP +2.5 baseline will increment the adjustment by one-quarter point.  In other words, a generic average of DEM +2.5 would produce an adjustment of 1.25 points for the Democratic candidate in each tracked House race.  Likewise, a GOP +3.5 average would result in a 0.25-point adjustment for the Republican.

The Formula in action, an example

For races with no published polls:

(Pundit Avg. * 100%) +/- generic poll adjustment
   NOTE:  Because I refuse to post a "toss-up" rating on any race, I will decide one way or the other in the event that pundit predictions and the adjustment result in a tie.  The margin for such races will be +0.5% for my best guess.

For races with one published polls:

((Pundit Avg. * 0.75) + (Poll * 0.25)) +/- generic poll adjustment

For races with two published polls:

((Pundit Avg. * 0.50) + (Poll * 0.50)) +/- generic poll adjustment

Here is an example of the calculations using the formula:

  • Pundit Predictions:  Raw value = 7.0 for the Democratic incumbent
    • Stuart Rothenberg:  Democrat by 8% (likely DEM)
    • Charlie Cook:  Democrat by 4% (lean DEM)
    • Larry Sabato:  Democrat by 4% (lean DEM)
    • Roll Call:  Democrat by 4% (lean DEM)
  • Polls:  Raw value = -2.0
    • XYZ Polling Company (R):  Republican (R) 44%, Democrat (D) 38%
  • Generic Adjustment:  +0.625
    • Congressional generic poll average:  GOP 43%, DEM 43%
  • Projected Result:  (7.0 * 0.75) + (-2.0 * 0.2) + (0.625) = 5.5%
    • Democrat projected to defeat Republican by 5.5%
2010     2008     2006     2004

Special Thanks To:

Charlie Cook
Cook Political

Congressional Quarterly
CQ Politics

D.C. Finegold Sachs
DC's Political Report

Dave Leip
U.S. Election Atlas

David Wissing
The Hedgehog Report

Larry Sabato
Crystal Ball

Richard & Tony
The Green Papers

Rose Institute
Redistricting in America

Tom Bevan and John McIntyre

2010 Projections, State-by-State


New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Rhode Island
South Carolina
South Dakota
West Virginia
Election Projection Resources
EP's RSS Feed

2010 Projection Results

2008 Projection Results

2006 Projection Results

2004 Projection Results

Search this site

provided by Google
Contact Me

e-mail Scott at:
-at- gmail.com

©Copyright 2003-2013 Scott Elliott, All Rights Reserved