Why are there biases in economic forecasts? We have been steeped in the belief that full information rational expectations are the basis for economic modeling. Individuals use all available information to generate unbiased for economic forecasts. There can be errors, but, on average, the expected value is zero because smart forecasters will drive out poor forecasters from the market. There will not be any biases and errors because all information is used. If there is a difference between the forecast and the realized value, it will be cause by new unanticipated information.
Unfortunately we find that there are errors and biases in forecasts and the world does not seem to behave in the expected rational manner. There is a correlation between forecast revisions and forecast errors. There has been some path-breaking work on information rigidity that can help explain why there may be errors in forecasts and why they may not go away. The piece, "Information Rigidity and the Expectations Formation Process: A Simple Framework and New Facts" by O. Coibion and Y. Gorodnichenko in the August American Economic Review provide some insights on why errors occur and how we can exploit these problems.
There has been a view that forecast errors occur because of loss aversion by forecasters. It is a behavioral issue. But, the correlation between revisions and errors does not have the correct sign for this story. What seems more likely is that there are forecast errors from information rigidities.
These rigidities occur through two possible channels. One, there are rigidity from infrequent updates because of fixed costs of forming the forecasts. This does not seem likely since there is a low cost of gathering information within macroeconomics, but even with low information costs there is a cost with going through the forecasting process. Sticky forecasts seem more likely for forecasts on the micro-level where costs may be higher. Two, noisy information creates a signal extraction problem. When there is more noise in the information gathered, there will be less willingness to change. There is a rigidity between the prior beliefs and the information that will form new beliefs.
Forecasts can be affected by sticky behavior and noisy prices. Both of these will cause a link between forecast revisions and errors. The behavior of the forecaster can be rational but the information rigidities can still create errors in the formation process.
So what can be done by those who trade? Two simple solutions make sense given information rigidity problems. First, update forecasting regularly with the latest information. Stop from being sticky. This can be in the form of nowcasting. Use the best available information without delay. Forecast should be updated at the worst on a monthly basis. Second, be aware of priors. a strong prior view will be slow adjustment to any new information. The noisy information problem is harder to address because it requires a willingness to change when new information enters the market. This issue will have to be balanced against transaction costs, but there cannot be a fear of changing forecasts.
Investors should always worry about whether they are using all information and forming rational forecasts. They should not worry about information rigidities if they better plan their processes.
There has been a view that forecast errors occur because of loss aversion by forecasters. It is a behavioral issue. But, the correlation between revisions and errors does not have the correct sign for this story. What seems more likely is that there are forecast errors from information rigidities.
These rigidities occur through two possible channels. One, there are rigidity from infrequent updates because of fixed costs of forming the forecasts. This does not seem likely since there is a low cost of gathering information within macroeconomics, but even with low information costs there is a cost with going through the forecasting process. Sticky forecasts seem more likely for forecasts on the micro-level where costs may be higher. Two, noisy information creates a signal extraction problem. When there is more noise in the information gathered, there will be less willingness to change. There is a rigidity between the prior beliefs and the information that will form new beliefs.
Forecasts can be affected by sticky behavior and noisy prices. Both of these will cause a link between forecast revisions and errors. The behavior of the forecaster can be rational but the information rigidities can still create errors in the formation process.
So what can be done by those who trade? Two simple solutions make sense given information rigidity problems. First, update forecasting regularly with the latest information. Stop from being sticky. This can be in the form of nowcasting. Use the best available information without delay. Forecast should be updated at the worst on a monthly basis. Second, be aware of priors. a strong prior view will be slow adjustment to any new information. The noisy information problem is harder to address because it requires a willingness to change when new information enters the market. This issue will have to be balanced against transaction costs, but there cannot be a fear of changing forecasts.
Investors should always worry about whether they are using all information and forming rational forecasts. They should not worry about information rigidities if they better plan their processes.
No comments:
Post a Comment