The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER A S UGGESTED D IVISION OF L ABOR D ISCUSSION OF E LLISON AND S ARGENT: W HAT Q UESTIONS ARE S TAFF AND FOMC F ORECASTS S UPPOSED TO A NSWER ? James Bullard* President and CEO Federal Reserve Bank of St. Louis 10th EABCN Workshop on Uncertainty over the Business Cycle European Central Bank, Frankfurt March 30, 2009 Any opinions expressed here are mine and do not necessarily reflect those of other Federal Open Market Committeee members. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER S ARGENT ’ S ADDRESS IN A S UGGESTED D IVISION OF L ABOR S EOUL Motivated by a speech by Chairman Bernanke in October 2007. Bernanke’s speech was about how policymakers might think about model uncertainty: Bayesian versus robust control methods. Policymakers that can place a prior over competing models are Bayesian. Policymakers that cannot instead use multiple priors and robust control, using a worst-case analysis to put bounds on performance. Relevant, or techno-babble? Paraphrasing Greenspan: Uncertainty is the defining feature of the economic landscape. Re-analyze a comment David and Christina Romer on FOMC forecasting. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER A S UGGESTED D IVISION OF L ABOR M ODEL UNCERTAINTY The level of model uncertainty faced by the FOMC is profound. Many of the “big ideas” from academia are simply not part of the analysis. This leaves one wondering how important these omissions are. Examples: Output gap measures, credit market frictions, multiplicity of equilibria. Is it right to be so far off of the research frontier? I do not think we would allow it in the space program or in the design of weapons systems. One important defense: forecasting. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER F ORECASTING IS A S UGGESTED D IVISION OF L ABOR TRACKING Most of the focus in policy discussions concerns today’s state vector. Further out is normally slow mean reversion. Through experience, forecasters learned that the near random walk model works best. Tracking requires careful attention to incoming data and revisions to recent data. What we would really like is to accurately predict the consequences of an unparalleled policy action. This is the big ticket item. For this we would need much better models than we have. Result: FOMC must make its own judgements from models which are plainly deficient in many respects. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER W HAT THE FOMC A S UGGESTED D IVISION OF L ABOR DOES The FOMC members make quarterly forecasts which are summarized and published. The staff “Greenbook" forecast is released to the public only after five years. One purpose of the FOMC forecasts is to communicate to the public, especially concerning medium run expectations. This alone is valuable. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER M ORE ON WHAT THE FOMC A S UGGESTED D IVISION OF L ABOR DOES The FOMC members forecasts are made "under appropriate monetary policy." The Greenbook forecast necessarily contains a policy assumption. For many years this was a constant or nearly constant policy rate. Greenbook forecast is heavily judgementally-adjusted. If members feel that actual policy would differ materially from the assumption, they should turn in a different forecast. Also, should a single member try to predict what the Committee will do? If so, one might forecast poor macro performance because of a lack of faith in the Committee. If not, one might forecast what may achievable if the Committee followed the fully optimal policy. T RUTH AND P ROBABILITY D OES S OME P ROBLEMS WITH R OMER AND R OMER THE FOMC A S UGGESTED D IVISION OF L ABOR EXERCISE MAKE SENSE ? Not as a pure forecasting exercise. For that, one would have to assume, like Prescott, that the actual path of monetary policy has a very limited impact on macroeconomic outcomes. This is probably the opposite view from Romer and Romer. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER W HAT R OMER AND R OMER A S UGGESTED D IVISION OF L ABOR DO Romer and Romer (AER Papers and Proceedings, 2008) compare forecasts of the Committee and the staff. They find that the Committee’s forecasts do not improve upon the Greenbook. They find that the Committee sometimes acts on differences in the forecast. This suggests that the staff forecast should be worse, since the staff is "surprised" by the renegade actions of the Committee. This does not seem to have happened. T RUTH AND P ROBABILITY D OES S OME P ROBLEMS WITH R OMER AND R OMER THE R OMERS ’ A S UGGESTED D IVISION OF L ABOR EXERCISE MAKE SENSE ? No. The "appropriate policy clause" arguably makes the question: What can be achieved in the current circumstances? Since forecasting is tracking and contains little information, perhaps not so interesting. The equilibrium is that the staff and the Committee think exactly alike. But the models in use are plainly missing many elements that may be very important at certain junctures. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER A S UGGESTED D IVISION OF L ABOR W HAT E LLISON AND S ARGENT DO The Committee may be better interpreted as responding to specification doubts. Doubts are abundant on the Committee. Then members do a worst-case analysis over multiple priors. This biases the forecasts toward worst-case outcomes. Question: Why do this? To signal other members about a potential pitfall? T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER A S UGGESTED D IVISION OF L ABOR M ORE ON WHAT E LLISON AND S ARGENT DO The model: Primiceri (QJE 2006), unobserved NAIRU. Comment: the level of model uncertainty is far more profound than the details of this model would suggest. Staff has the approximating model. FOMC has preferences for robustness. Question: The staff works for the FOMC, should they also have preferences for robustness? Question: Learning and robust control—after a long enough time period, policymakers would learn that pessimism does not pay off. T RUTH AND P ROBABILITY D OES S OME P ROBLEMS WITH R OMER AND R OMER THE A S UGGESTED D IVISION OF L ABOR E LLISON AND S ARGENT EXERCISE MAKE SENSE ? It makes more sense than Romer and Romer. It allows policymakers to have a concern for robustness which differs from the staff concerns. This concern could drive a systematic difference between the forecasts of the Committee and the staff. Still, the staff could be asked to calculate forecasts with a concern for robustness. And, this interpretation blurs the communication role of timely publication of FOMC forecasts but not staff forecasts. Level of model uncertainty far greater than suggested here. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER P OLICYMAKING A S UGGESTED D IVISION OF L ABOR WITH A CONCERN FOR ROBUSTNESS Tracking will remain important. Tracking incorporates average past policy behavior. Forecasting the effects of important policy interventions is critical. For that we need better models. Current forecasting/tracking does not tell us much and is unlikely to get better. T RUTH AND P ROBABILITY S OME P ROBLEMS WITH R OMER AND R OMER A S UGGESTED D IVISION OF L ABOR S OME QUESTIONS I WOULD LIKE TO BE ABLE TO ANSWER What are the chances that the global economy could become entrenched in the deflationary trap equilibrium described by Benhabib, Schmitt-Grohe and Uribe (2001)? What is the optimal regulatory response if we interpret private information to be at the heart of the current worldwide financial turmoil? To what extent can global coordination improve on worldwide equilibrium outcomes in the current environment? If I put some weight on a fiscal theory of the price level, should I be concerned about inflationary prospects going forward? These are the big ticket items. To get good answers, we need good models.