Using Prediction Markets to Follow Election News

Keith HopperThere are scores of news sources covering the elections, but which provide us with a clear picture of how things are progressing, especially when so many sources these days rely on sensationalism or political bias to build audience? One potentially under-valued way to understand what’s happening in our world is to take a look at online prediction markets.

Online prediction markets are sites where individuals can place bets - usually non-monetary – on particular outcomes of world events. These "bets" take the form of either/or outcomes, such as whether a particular movie will be number one in the box office or whether a political candidate will win or lose an election. Bets are then traded much like on a stock market. If you bet wisely on a particular outcome, then you stand to "profit" on your bet. As individuals buy and sell on potential outcomes, the collective prediction fluctuates in real time. As bystanders, we might gain some insight into predicted outcomes and their underlying drivers by watching the fluctuations in predictions as they unfold.

But are these predictions accurate? Markets such as Iowa Electronic Market have been recognized as being a fairly accurate and effective predictor of past elections. While past performance is no indicator of future outcomes, it can be enlightening to watch the outcome percentages fluctuate over time. Behind shifts in predicted outcomes lies the aggregated knowledge of all the market traders. In other words, when a percentage shifts there might be new information driving that shift. In the case of an election, this shift might represent the punch of a campaign ad or the release of a new piece of potentially damaging candidate history. Because traders stand to benefit if they move quickly on a piece of new information, outcomes in exchange markets often represent the very latest information and can be used as a bellwether of sorts around particular events as they occur. For example, perceived performance during the course of a live debate might drive real-time market predictions. Additionally, as bystanders we can gain a better understanding of important events without first having to track down and analyze the underlying facts. The underlying information is still critical of course, but instead of finding it and determining what it all means we might do the opposite – look for impacts and then find the information behind it that might prove meaningful.

Monitoring outcome percentages in prediction markets might be useful for the casual news consumer. I have pulled prediction numbers from three online prediction markets on the web around the upcoming presidential election in order to explore whether this theory might be correct. Specifically, I include predictions in real-time from Intrade, Foresight Exchange, and the Iowa Electronic Market around whether the incumbent president will secure a second term1.

Once per day, I will automatically pull the prediction numbers from the three respective markets and publish them to my Twitter feed. My hope is that these percentages and their fluctuations over time might help the casual news consumer get a slightly better understanding of not only the election outcome, but of meaningful news information that might be happening surrounding the election without having to wade through a sea of potentially conflicting reports. I’m curious whether this experiment might be useful or the predictions accurate. I have no idea – perhaps you can help me understand. I’ll be watching responses on twitter and reporting back if interesting stuff arises.

Latest Results:

1 To be clear, I could have tracked whether Obama will either win or lose the election (% chance of losing = 100 - % chance of winning for those keeping score at home). The decision to track the chance of a reelection and the decision to track the presidential election more generally does not represent my views on the outcome (e.g. whether or not I prefer either of these outcomes), but instead represents my desire to do the least amount of math possible, since all three prediction markets provide prediction data in the affirmative (win) rather than the negative (lose). It’s also worthwhile to note that these three prediction markets represent slightly different potential outcomes, for example, it is technically possible for Obama to not secure a second term without losing the election. I feel these outcomes are similar enough for the type of monitoring I propose to treat as equally representative of a predicted reelection outcome.

Intrade tracks Obama's re-election, Foresight Exchange and Iowa Electronic Market track any Democratic candidate winning the presidency.