A recurring theme in forecasting is that the combination of forecasts or ensembles will improve prediction. Blend information from different sources to make better forecasts. Sentiment, beliefs, and opinions help improve the predictive information from prices. This is especially applicable if prices are distorted from the intervention of governments, come from illiquid markets, or face high bid-ask spreads. When it is hard to inject information into the price, looking at beliefs will be helpful.
Prices, through the action of market participants, provide market information but so do the actual beliefs of the traders. Beliefs are a separate form of information. An interesting paper, Are markets more accurate than polls? The surprising informational value of "just asking" in the journal, Judgment and Decision-Making, finds through direct comparisons between self-reports and prices in prediction markets that beliefs are as accurate as prices. The combination of beliefs and prices will actually add value. Prices do not effectively aggregate all information and beliefs from self-reports can be an important source of new information. Notice the difference in Brier scores, the squared error between forecasts and outcomes, between prices and beliefs.
This study was conducted under a set of special conditions of self-reporting forecasts in a prediction market that tracked the beliefs of traders at the same time as trades were conducted. This was a laboratory study and would be hard to replicate in real life. Still, it provides a framework for thinking about how beliefs and prices can be used to build financial models.
Use price information. There is no doubt that they are an aggregator of information, but if you can get beliefs, opinions, and sentiment, incorporate this added information.
There is a move to nowcasting for predictions, but a significant part of the value of any nowcast is using information that is as timely as possible from surveys. Discount the data that is old but employ recent survey data of market participants as an added tool.
The heavy lifting is determining how to weight this information relative to prices. For example, the latest information from the Atlanta Fed business survey was announced yesterday. See the Survey of Business Uncertainty SBU. The most telling numbers were the measures of expectations and uncertainty. These do not look like numbers consistent with robust increases in financial prices. These beliefs may help forecasting future asset prices.
Prices, through the action of market participants, provide market information but so do the actual beliefs of the traders. Beliefs are a separate form of information. An interesting paper, Are markets more accurate than polls? The surprising informational value of "just asking" in the journal, Judgment and Decision-Making, finds through direct comparisons between self-reports and prices in prediction markets that beliefs are as accurate as prices. The combination of beliefs and prices will actually add value. Prices do not effectively aggregate all information and beliefs from self-reports can be an important source of new information. Notice the difference in Brier scores, the squared error between forecasts and outcomes, between prices and beliefs.
This study was conducted under a set of special conditions of self-reporting forecasts in a prediction market that tracked the beliefs of traders at the same time as trades were conducted. This was a laboratory study and would be hard to replicate in real life. Still, it provides a framework for thinking about how beliefs and prices can be used to build financial models.
Use price information. There is no doubt that they are an aggregator of information, but if you can get beliefs, opinions, and sentiment, incorporate this added information.
There is a move to nowcasting for predictions, but a significant part of the value of any nowcast is using information that is as timely as possible from surveys. Discount the data that is old but employ recent survey data of market participants as an added tool.
The heavy lifting is determining how to weight this information relative to prices. For example, the latest information from the Atlanta Fed business survey was announced yesterday. See the Survey of Business Uncertainty SBU. The most telling numbers were the measures of expectations and uncertainty. These do not look like numbers consistent with robust increases in financial prices. These beliefs may help forecasting future asset prices.
This simple belief measure is not like the lab study done, but it can be suggestive of business executive beliefs that can be used to temper thinking about financial prices that have had the benefit of a surge in Fed liquidity.
No comments:
Post a Comment