Predicting Congressional Bill Passage
Looking at how Congress works has always been an area of intense study, and analyses like this can be incredibly effective, especially when they can give citizens a clear picture of how likely it is for a bill that directly affects them to become a law.
Academic research has shown that the text of the bill is incredibly predictive. I was interested in seeing if the summaries of bills would be predictive by themselves.After some data cleaning and discarding documents with inadequate information, I worked with about 12,000 bills, resolutions, and amendments from the 113th and 114th Congress in total, and focused my initial analysis on the summaries written by the Congressional Research Service. Of my dataset, roughly 7.5% of the bills, resolutions, and amendments were passed. I managed to beat the (pretty high) baseline score by accurately predicting whether or not a bill would pass roughly 96% percent of the time. I also used a different technique to pull out the key topics associated with each bill.
I'll be tweaking the model and using the output to try to predict how individual Congressional representatives will vote to help activists be more efficient in the way they contact their reps!
Image source: Clerk of the U.S. House of Representatives