Datafest: Hacking Campaign Finance Data over a Weekend


This weekend I had the pleasure of helping to run Datafest, a two day campaign finance hackathon at Stanford. The event brought together around 60 journalists, technologists and students, many of whom had no prior experience in campaign finance or data driven journalism. The enthusiasm and dedication of the group–as well as the results–were truly heartening to see.

On Saturday morning the participants split themselves into ten teams, each of which came up with an original project to implement over the following day and a half. Some teams focused on original reporting using political influence datasets. Many produced visualizations. On the more technical end, there were interactive apps and statistical modeling. At the end of the weekend we announced an overall winner and three runners up, all of whom deserve a bit of recognition for what they achieved in such a short time.

Team Frienemies Screenshot

Team Frienemies took home first place with a system for grouping and visualizing political entities by their common donors. The project is compelling not so much for its catchy visualizations, but for the insights they were able to derive. Groups that are non-partisan in name only, such as Emily’s List or Club for Growth, are correctly placed at the center of the Democratic and Republic clusters. The visualization also suggests non-obvious associations, such as various telecommunications industry groups leaning Democratic. The project authors also proposed using the tool as a recommendation system: donors could find other candidates or organizations they may support and political groups could find other funders that may be interested.

Team Gophers chose to dive deep into the influences of one company as a case study in using the many datasets available. Their visualizations touched on campaign finance, the revolving door of lobbyists, federal contracts, and bills before Congress. One key insight they were able to show was that the company gave predominantly based on committee membership of the member, not based on ideology, expected election outcome or geography.

Team KeyStoners took a slightly different approach: they focued on one particular issue–the Keystone Pipeline authorization in Congress–and looked at a variety of influences on the outcome of the vote. In particular, they tried to explain what accounted for 41 Representatives who flipped their position before the final vote on the bill. Interestingly, the found that rather than campaign contributions, lobbying or prior voting record, by far the most explanatory factor was being a Democrat in a district surrounded by Republican districts.

Team MostExcellent started with a simple but interesting question: is the out-of-state money going into the Wisconsin Governor’s race unusual? Through data analysis and interactive visualizations they show that out-of-state money is not an anomaly. This was combined with two short news pieces on the topic, giving a nice example of using data both to investigate a story and to present that story to the reader.

If you’re interested in hacking on your own, check out the data section of Influence Explorer, where many of the teams at Datafest found their data. And many thanks to Knight Fellow Teresa Bouza for organizing the event!