Prediction markets derive their predictive value from the crowd (a.k.a. wisdom of the crowd). An example given by Suroweicki (2004): “If one asks a large enough number of people to guess the number of jelly beans in a jar, the averaged answer is likely to be very close to the correct number.” But how’s this theory used in practice?
Prediction markets are very applicable to political elections, and an important election is coming up. On the 8th of November a new president will be chosen in the USA. Hillary Clinton running as the democratic candidate, and Donald Trump for the republicans. In the images below the prediction markets for the presidential candidates are shown. What can we infer from this?
Source: http://predictwise.com/politics/2016-president-winner
The upper graph shows the probability of Clinton and Trump winning the presidential elections. The market predicts that at this moment, Clinton has a probability of 87% of becoming the (first female) president. Her probability of becoming president increased sharply after the first televised debate, as well as today’s news about Trump’s opinion towards women. The changes in the probabilities for Clinton winning can be seen more clearly in the graph below. The question underlying these probabilities is as follows: how much are you willing to pay to receive one dollar if Hillary Clinton is elected president? And vice versa: how much are you willing to pay to receive one dollar if Donald Trump is elected president? The dollar amounts (willingness to pay) the market reflects is the probability the crowd assigns to an event to occur.
Source: http://predictwise.com/politics/2016-president-winner
According to the market, Trump only has a chance of 13% of becoming the next president of the USA. What would your advice be based on the information above? Should Trump suspend his campaign or is the crowd not getting it right this time?
References:
Surowiecki, R. (2004) Available at: http://www.diplomacy.edu/resources/books/reviews/wisdom-crowds-why-many-are-smarter-few
Dear David, thank you for your blog! While I truly hope that prediction markets are correct (which they have proven to be), I still wonder what will happen in the elections. Also, can only US citizins take part in the prediction markets you used as examples? Or can anyone around the world join?
Based on the article we had to read for IS, I believe prediction markets are a good indicator of what will happen during the elections. My advice based on the information above would be for Hillary to continue her campaign and convince more people to vote for her, as to limit the chance of Trump winning. While I hope the crowd is getting it right, I doubt Trump will stop his campaign solely based on prediction markets. Since prediction markets have been proven to be a good indicator, tho, it would very much come as a surprise to me if Trump was elected. However, Brexit for example also came as a surprise, so I guess we will just have to wait and see on the 8th of November…
Dear David thank you for your interesting blog post. Although you mentioned in your blog that according to the market Trump only has a 13% chance of being elected, I do think that the poll average is closer. Therefore I don’t think that Trump will be willing to stop his campaign because of the market predictions. Nonetheless, history has proven that the market is a good indicator and I truly hope that this is also the case for the upcoming election. I wonder how close the results will be.
Predictions move over time, and the crowd has it wrong from time to time. A very good recent example of this is the referendum whether Britain will remain a member of the European Union or not. The prediction markets expected the chance for Britain to remain in the EU to be well over 90% (http://andrewgelman.com/wp-content/uploads/2016/06/brexit-bremain.png) up untill the actual day of the referendum. Yet, as we all know, the leave camp won in the end. I would therefore not suggest Trump to suspend his campaign. Crowds can get it wrong from time to time.
Hi David, prediction algorithms are indeed really cool. We should however be careful to focus too much on those predictions. The most recent example where algorithms were completely off the track was during the Brexit referendum. All polls indicated that the UK wouldn’t withdraw from the European Union, and everyone knows what happened in the end.
Political polls are always difficult. Some subgroups in the population are not willing/able to participate in the polls, which is also happening in the Netherlands now. Some say that the DENK party will receive a considerable amount of votes, but the polls cannot confirm this due to the lack of Islamic/Turkish respondent. Therefore prediction models should be interpreted carefully in some cases!