Crowdsourced forecasting methods - Should be applied to finance

 

Crowdsourced forecasting methods are being tested and assessed by the intelligence agencies and the US government to help forecast some thorny issues in geopolitical analysis. This work is still in its infancy and is not accepted by everyone, but the results from testing suggest that this can be an important alternative tool for forecasting under uncertainty.

There has been a growing interest in prediction polls, forecasting competitions, prediction markets, and different forms of  crowdsourcing to help with forecasts, but these tools have often been ad hoc and have not been integrated as part of any normal forecasting process with organizations. The premise for crowdsourced forecasts, however, is simple and based on the concept that there is "wisdom of crowds". Now the potential for groupthink when crowds are used in present; nevertheless, significant evidence exists that a set of diverse independent opinions creates better forecasting outcomes from crowd aggregation. 

Think of the averaging of individuals from a crowd as a form of ensemble  modeling. Along with value from the averaging of diverse opinions, individuals can be tracked and compared to others in order to find individuals that have forecasting skill. Crowdsourced forecast can complement existing methods while also uncovering additional useful information on specific topics. 

Crowdsourced forecasting methods can uncover unique information if it is conducted through precise methods that  ask  for likelihood to specific events. The forecasts have to be falsifiable. For example, will the 10-year yield be above 2% by the end of 2021? Will the YOY CPI inflation be below 2% at the end of the year? The answer can be in a form of a probability that can be updated. This is the preferred method for posing questions over broad questions. 

There is limited value from asking general opinions on a topic. Asking question in the form of a probability of an event occurring can provide precision in thinking on uncertain events. Output from precise crowdsourcing forecasts leads to tracking over time as well as a means of assessing the  skill of predicting specific events. 

Forecasts that are falsifiable have been around for the longest time on Wall Street when analysts are asked about such things as the level of the stock market, interest rates, or GDP. This is the bread and butter of Wall Street forecasters, yet many tasks could use better crowdsourced precision. This precision is especially the case within investment committee structures. It is not just important to assess outside forecasters. Rather it is the inside forecasts where there is more value. For example, what is the likelihood that hedge fund manager X will exceed a 10% total return? Or, what is the likelihood that volatility (VIX) will exceed 25% at the end of the third quarter. Crowdsourcing techniques and analysis increase accountability to those involved in an investment process.  

See the recent report: "Keeping Score: A New Approach to Geopolitical Forecasting" from Perry world House (University of Pennsylvania)