Although polls are a common performance metric in the presidential race, their results do not always reflect the actual chances of each candidate winning. Months before any votes are cast, candidates can lead the field in polling, but never have a true shot at becoming their party’s nominee. Prediction markets offer an interesting alternative to keep track of how candidates are doing over the course of an election cycle.
PredictIt.org offers an online marketplace for trading contracts, paying out based on the outcome of political events. For the presidential primaries, markets exist for all state and national elections. Shares are traded for each candidate, paying $1 only if they win. The resulting prices infer the chance of each candidate winning.
Betfair.com is a British gambling website that allows people to lay their own odds, and place bets based on the bid-ask spreads of its users. Betfair is primarily involved in betting on sports, but also caters to political bets. Betfair's odds provide predictive data at the national level between the Republican and Democratic candidates.
OddsChecker.com collates betting odds from a range of different bookmakers. Bookmakers act as a market-maker in a traditional stock market environment - they set their odds, and adjust them according to the flow of the market. US Political events are a popular subject of wagers, so markets exist to predict the presidential primaries.
Marco Rubio sailed into January 2016 having never polled above 15%, yet prediction markets labelled him as the clear frontrunner against all other opponents. Ben Carson was practically ignored by all of the prediction sites, despite polling significantly higher than Bush and Christie before dropping out. Donald Trump had been consistently beating the other candidates in polling by a 20-point margin in early-2016, but was only priced as having a 20% chance of winning.
As the race moved on, the markets continually realigned, as speculation around candidates shifted. Some clear events led to spikes for different candidates, as investors hurried to change their minds.
Although Bernie Sanders closed the gap in polling throughout the primary process, Hillary Clinton was viewed as an unassailable candidate throughout the election process. In November, Clinton’s status as the presumptive nominee allowed for only a small chance of around 20% that something could happen to derail her campaign.
Different markets disagreed with the scale of a threat that a surging Sanders campaign could pose to Clinton. Although punters on PredictIt and OddsChecker remained steadfast in their belief that Clinton would win the nomination, Betfair shortened Sanders’ odds significantly for a substantial period of time.
In the Republican race, Donald Trump has performed best in contests on the East coast, while Ted Cruz has had his finest showings in rural and Southern states.
Marco Rubio was never predicted to win any early states, despite being hotly tipped to win nationwide. His win in Minnesota, which was a surprise result for PredictIt users, together with Kasich’s victory in Ohio, were the only states that were not dominated by Trump or Cruz.
On the Democratic side, Hillary Clinton enjoyed a significant advantage in Southern states, and bested Bernie Sanders in more well-populated areas.
The North-South divide is evident from the map to the right, where Sanders can be seen to have little chance at any point in winning in the South, although Northern states generally fell to him.
How do the issues that the candidates care about change over time? Is there a relationship between these issues and their polling results?
Brush the timeline below to explore
The Presidential Debates have always been an integral part of the election. Critics have commented that the debates "too often turn on stumbles, errors and style over substance." How do candidates' debate styles change over time?
Transcripts of every candidate in every debate from December until now were analyzed. The polarity score is a float within the range [-1.0, 1.0]. For example, a sentence like "Bicycles are amazingly simple to use" would get a polarity score of 0.4. The subjectivity score is a float within the range [0.0, 1.0] where 0.0 is very objective and 1.0 is very subjective.
The percentage of short sentences is encoded by the radius of the circles. The more often a candidate uses a short sentence, the bigger the circle is. A short sentence has 5 or less words