Opening Data on the TV Spend in the US Presidential Election Campaign

Rossum, the London-based disruptive AI company, and, the first technology-enabled global venture capital firm, combined public FCC documents and cutting edge AI technology to publicly release the first highly granular dataset about media advertising of the presidential campaigns.

“We show that Artificial Intelligence and Big Data can inform the 2020 election both from the side of the candidates and the side of the public. Thanks to recent AI breakthroughs, we are the first to automatically extract key data like the advertiser, the network name, flight times and total amount from all 23,000 political insertion orders submitted by media networks to the FCC and make that entire dataset freely available to the public in collaboration with,” says Petr Baudis, Rossum’s CTO.

Why it matters

Data about presidential campaign spending can paint an in-depth picture about how US votes are swayed and how the competing campaigns are valuing potential marginal electoral votes in each state.

What are the biggest hotspots of the presidential election? While Pennsylvania and North Carolina might be the media favorite swing states, the campaign analysts have a different opinion. Florida unequivocally leads the spending chart, totalling almost twice the campaign spend of Pennsylvania. (Fig. 1) Meanwhile, a single vote commanded the highest spend in Arizona, which leads on per-capita spending. (Fig. 2)

Did Bill Stepien take 1.5 months to get up to speed as Trump’s campaign manager? “While Biden’s spending for TV spots has been steadily increasing week by week until late September, Trump’s campaign does not follow a similar growth curve for long. We can see a sudden flat line starting July 16 when the previous campaign manager was replaced, and the campaign only picks back up at the beginning of September,” explains Sebastian Goebl,’ data engineer who analyzed the data. (Fig. 3)

And ultimately, does money really buy votes after all? The data can give a true picture on the perceived value per electoral vote in each state to the campaign over time. (Fig. 4)

The public access difference

The FCC has done an incredible job of collecting political ad information from the networks and making that data available in raw format to the public in near real time.  While research on presidential campaign spending (like Advertising Analytics or Kantar/CMEG estimates) already exists, these are either just approximations and educated estimates, or are expensive to access because they are based on manually sifting through almost 25,000 documents submitted by networks to the FCC.

Rossum and open up access to this data by highlighting the critical trends in analytical dashboards and releasing the full dataset at the individual order level. This data can now be opened to the public without any charges because Rossum’s AI engine captures the data at a fraction of the cost of a team of human analysts.

This data is now freely accessible to academic researchers, the media and the public to explore how campaigns have focused their advertising budgets leading up to the election on a city and state level over time.  Moreover, it opens a new playing field for analysts and investigative journalists to identify patterns or unique discrepancies… or simply take a closer look at the 500 orders covering literally free advertising.

This public data release foreshadows the impact artificial intelligence can soon make on public policy as well as analytical journalism. Many open data initiatives in the past were stumped by the fact that the data was not available in machine readable form. But the Rossum/ project signals that AI is now ready to take on the challenge to process data even from unstructured PDFs and scanned documents.

“We believe that this level of public interest can ultimately help campaign leaders as well as journalists, not just to maintain a level of fair play, but also to think better about how to run a campaign efficiently. We realized that raising and spending money in a campaign is similar to how modern Direct To Consumer startups are scaling, and many of the same metrics do apply,” closes Partner Thomas Gieselmann, who coordinated the project.

Ready to get started?

Make a quantum leap in your document processing approach. Boost accuracy and effectiveness with an AI-powered data capture solution for all documents.