Our new constitution is now established and has a semblance of permanence; but in this world, nothing can be considered certain except death and taxes.
The human brain has evolved to make predictions.
Born into a world of uncertainty, early humans had to deal with threats, weather, and food sources in order to survive. Evolution has ensured that humans are good at this and enjoy it.
Today, we face similar questions. Will monkeypox develop into a serious pandemic? Will Russia and the North Atlantic Treaty Organization get directly involved? China-Taiwan? Climate change? Global food supply?
It is impossible to answer these questions in advance. The future is uncertain, and the systems these questions address are too complex and interdependent to fully predict. But that doesn't preclude patterns. And so we can make reliable, partial predictions. You can't tell for sure if a parent is angry or disappointed when they call out to you, but hearing their full name is a darn good signal. By studying patterns based on the accumulation of experience and knowledge, humans have developed a system for quantifying uncertainty and making predictions.
Clay Graubard and Andrew Eaddy are the founders of Baserate.io, the publisher of Global Guessing, a geopolitical forecasting website, and Crowd Money, a newsletter and podcast about prediction markets. This article is a preview of a talk they will give next week on the Big Ideas stage at Consensus 2022 in Austin, Texas.
Weather forecasts, once considered useless despite millions of times practiced, are so reliably wrong that they can be trusted. You won't know at 8 a.m. if it will rain at 2 p.m., but thanks to advances in meteorology, the Dark Sky app can predict a 24% chance. Sure, unlikely. But in 2016, Nate Silver gave former U.S. President Donald Trump a 28.2% chance of being elected president, so it's probably wise to bring an umbrella.
In the social sciences, we have found that two democracies are less likely to go to war with each other than other combinations of states, that crime is more likely to occur in neighborhoods with "broken windows," that civil wars are almost twice as likely in countries with a per capita gross domestic product of $250 as in those with $500 (15% vs. 8%).
If you don't believe in prediction, we live in a probability-driven world. Forecasting is something we all do, whether we know it or not. It is a tool, a process, and most importantly, an ability to generate information. Fortunately, it is a skill that we can improve and become good at.
This is essentially the main finding from the research of Philip E. Tetlock, Barbara Mellers, and many others. Their research has shown that if we quantify, record, update, evaluate, and practice, we can make accurate predictions about complex issues. We can see at least partially through the "fog of war."
That's why we should pay attention when forecasters say there's a 28% chance the World Health Organization will declare monkeypox an international health emergency. Or a 10% chance that there will be a direct conflict between Russia and NATO before July 2023, and a 15% chance that there will be a war between China and Taiwan before 2024. And so on. And so on.
This is also why we should wonder and even argue about how little attention and funding forecasts receive.
Forecasting provides good information, forces accountability (you can't just brush aside a track record!), and rejects sophistry, sensationalism, and biased thinking. It is a perfect check on traditional media and a potential antidote to polarization. You can't make predictions and disregard an important perspective. Otherwise, accuracy suffers. This is perhaps why forecasters are less politically polarized than non-forecasters.
Unfortunately, the value of forecasts is no guarantee of immediate or even relatively timely widespread application. History is replete with such examples, including such obvious things as seat belts. Forecasting is inherently disruptive, making it difficult to introduce into organizations.
Forecasting is also difficult, time consuming, and (currently) underpaid (trust us). We believe that prediction markets are one of the most hopeful approaches because they introduce monetary and psychological incentives and market efficiencies into the forecasting process.
What are prediction markets?Prediction markets are not new. Even in ancient times, people made bets on the future outcome of events. And the term "wisdom of crowds" can be traced back to early 20th century England.
Century. In 1906-1907, Sir Francis Galton, a Birmingham scientist and mathematician and a relative of Darwin, witnessed a contest in which some 800 villagers in Plymouth, England, were asked to estimate the weight of an ox. While each individual estimate was too high or too low, Galton found that the mean was within 0.8% of the weight measured by the judges of the contest. The mean of the estimates was exactly right.
Prediction markets are marketplaces where participants trade future outcomes on specific topics. Think of the stock or crypto markets, except this one is about events.
Prediction markets are usually binary, offering two defensible values for a given market (think "yes" or "no"). These assets trade between 0% and 100% (e.g. $0 to $1), with the current market price representing the consensus of the crowd.
When a predicted event occurs, traders who bought shares with the correct outcome receive $1 for each share they owned. Similar to long-established stock markets, the main incentive for participants in prediction markets is profit, while the byproduct of their prediction activity is information.
However, while traditional prediction markets often operate as described above, it is important to place this technology in the context of the larger ecosystem of prediction platforms that use different methods to produce forecasts. There are prediction markets that use real money or play money. (Among real money platforms, some place bets in fiat, others in crypto).
Other platforms, such as Metaculus, use a poll-based method in which participants continuously submit their subjective probability assessments on a question. Many of these platforms create incentives through reputation, community, and other factors to generate more high-value information.
One advantage of prediction markets over other methods is that, all other things being equal, they produce more accurate forecasts than traditional prediction platforms (or surveys) such as Good Judgment Open. The forecasts are less biased, less noisy, and better informed.
Another benefit is that betting makes things more exciting and often increases engagement, not only on hot topics like sports, but also on other important and previously neglected topics like the risks to humanity from artificial intelligence. And as individuals are able to make consistently accurate and well-calibrated forecasts, the opportunity exists for smart traders and elite forecasters to earn reliable returns - especially as returns for accuracy can increase over time.
Prediction MarketsTodayLarge, modern prediction markets continue to build on this insight, creating more efficient and reproducible means of aggregating forecasts to provide accurate information. Within the ecosystem, however, there are different categories and classifications for prediction market platforms.
Some prediction markets, like Kalshi, are centralized and use U.S. dollars to support trading. Others, like Polymarket, are partially decentralized and trade the stablecoin USDC. And still others, like Augur, are fully decentralized, complete with decentralized oracles to facilitate decentralized resolution of issues.
There are also differences between prediction markets in terms of their stated purpose. For example, Google has been using prediction markets internally since 2008 to improve business decisions and understand employee sentiment. The incentive structure and functionality of a prediction market like Google's are understandably different from those of a publicly available prediction market.
Finally, there are differences in the technical functionality of prediction markets. Some prediction market platforms, especially those on blockchains, use automated market makers (AMMs) to provide liquidity. Other platforms use a traditional order book-based trading system.
AMMs provide consistent liquidity in a market and often provide a more positive trading experience for market participants, but when liquidity dries up, slippage can occur, meaning that order sizes can affect order prices, which is not good. Order book trading, on the other hand, is much simpler. A trading screen displays a selection of buy and sell orders for an asset at different prices. As a trader, you can execute any order in that book.
Prediction Markets of TomorrowAside fromimproving the accuracy of forecasts and increasing engagement, prediction markets can help reform the media, drive community in social media, and inform key decision makers in the private and public sectors through systems like Robin Hanson's Futarchy, shown below
.Such a system would reward attentive and engaged citizens. It would also incentivize good decision making, because even if you choose the "right" and implemented policy, it has to work to win money. Over time, bad forecasters would be weeded out. Through futarchy, prediction markets can support fundamental and vital values within a healthy society, namely information (through forecasting), governance (through futarchy), and institutions (prediction markets).
Even without futarchy, prediction markets have the potential to transform today's political process by engaging people in the conversation and providing more reliable information about the future.
We live in an uncertain and complex world. Critical thinking and reliable information are needed to make good decisions. Quantified predictions are an invaluable but still underutilized tool, and prediction markets seem to be an important vehicle for their adoption. At baserate.io, we believe this needs to happen sooner rather than later, because even certainties like death can be uncertain today.