We’re just over a month removed from Election Day 2016 and still the postmortem campaign analysis rages on. Much of the discussion has focused on Donald Trump’s rise from fringe candidate to president-elect in a span of 18 months. The extent to which economic factors, the media, and/or our political parties’ dysfunction fueled his ascent is an ongoing debate. Regardless, it’s clear that pundits and pollsters alike drastically underestimated his appeal and chances of winning.
That’s why working with Good Judgment Open (GJO) provided an awesome opportunity to gain some perspective on the campaign. GJO is a platform birthed out of University of Pennsylvania professor Phil Tetlock’s research on the merits of crowdsourced forecasting. It runs forecasting contests on a variety of global issues, asking users to assign a probability to the chance that, say, Scotland will set a date for another referendum on independence before July 1. Users can update their forecasts as new information surfaces.
During the 2016 election season, GJO partnered with The Washington Post’s Monkey Cage blog to pose 45 questions related to the presidential, congressional, and gubernatorial races. A total of 8,359 unique users inputted forecasts across all the questions. After being approached by GJO about a collaboration, The DataFace team built the following visual tool to sift through the data. You can look at how the average probability on certain questions evolved over time, the sources of information from which GJO users were drawing to make their predictions, and how positively or negatively they were viewing the election as it progressed.
 We excluded stop words like “a”, “the”, “and”, “or”, and others when computing the most frequently used words in GJO comments. (Back to reading)
 Most questions didn’t have a sufficient number of comments submitted by users who were updating their forecasts for this part of the analysis. Refer to the second dropdown on the “Sentiment Analysis” dashboard to see which questions were included.(Back to reading)