To make better decisions under uncertainty, you must first accept one limitation of human psychology: On your own, you are an unreliable forecaster. Ask one person to predict the outcome of an election or a football game, and their intuition often fails. Bookmakers build empires on this exact vulnerability.
We process reality through a customized lens of cognitive biases and psychological defense mechanisms. These are adaptation strategies that build up over a lifetime for functional reasons: They protect our sense of self, manage anxiety, and help to maintain social cohesion.
But heterogeneous crowds can make predictions differently. Correcting for individual error requires a large number of predictions, people with access to different information, and some skin (or cash) in the game.
The crowd and the market
When you pool the predictions of thousands of people who assign a monetary value to their confidence, idiosyncratic errors cancel each other out.
This is how prediction markets work. If you believe a political candidate will win an election, you can purchase a contract that pays out if you are correct. You’re not locked into that contract. You can sell or trade it at whatever the market price is, as the odds shift.
The market is an ongoing financial conversation about the likelihood of something happening, and every new transaction is additional information.
The emerging empirical evidence for this collective accuracy is impressive. A study that has just been published in Proceedings of the National Academy of Sciences shows that prediction markets pretty much get the probabilities right.
Prediction markets are good at predicting the odds
The researchers evaluated over 1 million contracts drawn from major prediction market platforms, with data spanning 12 distinct topics and 10 years of market activity. To capture genuine foresight rather than obvious hindsight, the researchers extracted probability estimates at standardized fractions of each contract’s lifespan.
Think about an election that is 100 days away. If the prediction market opens 100 days before the election, the researchers examined the market’s probabilities when the election was 75, 50, and 25 days away. This revealed whether the crowd held genuine early insight long before the outcome became obvious.
The results reveal an excellent ability to predict probabilities. When a market assigns an event a 25 percent probability, you can expect that event to occur between 20 and 35 times out of 100. Even when information is scarce, the crowd tends to get their predictions right.
Just to be clear, this doesn’t mean if a prediction market says that an event is likely, the result will always occur. If a prediction market estimates there is an 80 percent chance of a sports team winning their next match, they tend to win it 4 out of 5 times, not every time. A 95 percent chance of an election going one way still delivers a surprise upset about 1 time out of every 20.
The crowd also has biases
The market and its participants still have some biases. Whether estimating the outcome of a football match, a national election, or the outbreak of a war, prediction markets remain largely accurate in the likelihoods they assign to events.
However, the crowd consistently underestimates political outcomes while consistently overestimating the likelihood of military and geopolitical conflicts. When a prediction market prices the probability of a war or security event at 75 percent, the historical data proves the true likelihood is closer to 63 percent.
We also predict the breakout of war better than we can predict peace. By evaluating the sentiment of geopolitical and security questions, the researchers found that forecasts concerning pessimistic or negative outcomes yield consistently higher accuracy than optimistic or neutral predictions. The collective unconscious may be more accurate as Cassandra than Pollyanna.
How to make better decisions about the future
To navigate uncertainty, we have to abandon the myth of the perfectly rational and fully informed individual navigating a marketplace (sorry, economists). On their own, our personal forecasts are unreliable, especially about major world events.
Prediction markets also reveal a vital insight about human reasoning: Groups with identical assumptions repeat the same mistakes. Overconfidence amongst like-minded individuals makes groups worse at predicting the future. Combining information across groups that make different mistakes brings us remarkably closer to the truth.
Improving our own decision-making requires a deliberate shift in what we think we know about the future. Instead of consulting your gut or polling a room of like-minded peers, actively seek out objective ways to pool information you have with people who disagree with you.
To see the future more clearly, adapt alongside those who possess entirely different blind spots.
