An algorithm predicting voter turnout?

The Win Number: it is how many votes a candidate needs to win an election.

This number is simply deduced by taking the total number of expected voters, dividing it by 2 (50%, or half) and then adding one (+1).  If a candidate reaches that target number, then presumably the candidate has earned a majority of the votes and thus wins the election.

Therefore, everything depends on the win number.  If the campaign knows how many votes it needs, it can figure out how much it needs to spend on mail, or commercials, or on staff to contact voters.  The campaign can figure out benchmarks and timing and plan a comprehensive strategy.

In other words, an accurate win number is essential.

I have been working on a singular algorithm to assist in assessing the likely voter turnout; thus, more accurately determining the win number.

I am not saying I am finished yet, but I figure I may as well start to share some of what I have done.

In the graphs below, you can see four general elections.  Two are presidential (2008 and 2012) and two are off-presidential (2010 and 2014).

The same singular algorithm was applied to public data related to election results of 25 Illinois House Districts with identical offset between presidential and off-presidential cycles (the data was copy-and-pasted straight from the Illinois Board of Elections. For example, here ya go.).

The algorithm produces 2 numbers: a high-end prediction and a low-end prediction.

As the graphs show, the algorithm produces a high-end prediction that closely mirrors the actual voter turnout of presidential cycles in many districts.

Similarly, the same algorithm produces a low-end prediction that closely mirrors the actual voter turnout of off-presidential cycles in many districts.

Thanks for reading!

 

Screen Shot 2016-07-21 at 8.51.48 AM