As we all obsess about how many more minutes could possibly be left until this U.S. presidential election finally (mercifully?) comes to an end, Dish threw some fun data into the maelstrom that could prove interesting to compare to the final results. The company collected anonymized, aggregated viewership data from millions of set-top boxes across its national footprint, and shared the results of its “Viewers to Voters” predictive model on Monday.
It analyzes the potential outcomes of the 2016 presidential election, and based on the anonymized customer viewership data generated from internet-connected boxes – and it’s important to note the italics I’ve added there − the modeling predicts that Hillary Clinton will win. The model also foresees a Republican majority victory in the U.S. House of Representatives, with an estimated 245 out of 435 seats. This data would represent a Dems two-seat gain. Sufficient historical viewership data was not available to meaningfully predict the state of the Senate, Dish reports in a press release.
“With so much focus around national polling, we thought it’d be interesting to see if we could find a correlation between how our customers interact with Dish and how they may vote,” Warren Schlichting, Dish EVP of media sales, marketing, and programming, says. “We recognize that our call on the distribution of seats in the House may be an outlier. Yet when we tested the model against 2014 House elections, we found that we were able to predict the outcome at a 98 percent reliability point.”
So, if you watch a lot of football or tend toward music-based programming, which way are you more likely to vote? Viewers to Voters analyzed the relationship between programming watched and political affiliation, and the model determined that customers who watched more sports, religious, or family-oriented TV were more likely to vote Republican. Subscribers who watched more series/specials, education, or music-oriented programming were more likely to vote for Democrats.
Working with terabytes of viewership data, Dish’s Data Science team reportedly took a two-step approach to developing a predictive model for the 2016 presidential election. The first stage of the model identified the relationship between the shows customers watched in 2014 with the state-by-state outcomes of the 2014 House elections. A variety of variables were analyzed including the types and amounts of programming customers viewed across nine different Nielsen-defined genres. The second stage of the model then identified the relationship between House party control and presidential party affiliation dating back to 1932.
The Data Science team did not draw from any outside polls to reach its conclusions, according to Dish.
