Bad forecasters can be good policymakers

Thomas J. Sargent, Martin Ellison, 24 November 2009



The value of the Federal Reserve’s Open Market Committee (FOMC)1 has recently been questioned in a highly provocative paper by two professors at the University of California, Berkeley. The two professors are husband-and-wife team Christina and David Romer, who are amongst the most influential economists in the world today. Christina Romer is Chair of the Council of Economic Advisers in the Obama administration and a co-author of Obama’s plan for recovery, and David Romer is the author of a very popular macroeconomic graduate textbook. Their paper was published in The American Economic Review, arguably the most influential journal in economics.

The Romers criticise the FOMC because of its poor performance in forecasting economic developments. Specifically, the Romers show that the FOMC is even worse at forecasting than its underlings, the staff of the Federal Reserve System. This is surprising because the FOMC should have all the advantages when forecasting. The FOMC has the staff forecast available when preparing its own forecast and the FOMC presumably knows its own policy objectives and preferences better than anyone else. Despite this, the Romers find that:

  1. It is best to ignore the FOMC forecast when predicting inflation or unemployment.
  2. The FOMC makes larger forecast errors than the staff.
  3. Monetary policy reacts when the FOMC forecast differs from the staff forecast

The Romers use these findings to paint a bleak picture of the FOMC as "not using the information in the staff forecasts effectively" and accuses that the FOMC "may indeed act on information that is of little or negative value". In their opinion, the evidence is sufficiently damning to warrant a radical restructuring of the role of the FOMC in policymaking:

"a more effective division of labour within the Federal Reserve System might be for the staff to present policymakers with policy options and related forecast outcomes, and for policymakers to take those forecasts as given. With this division, the role of the FOMC would be to choose among the suggested alternatives, not to debate the likely outcome of a given policy."

These criticisms are understandable in a world where consumers, workers, policymakers, and researchers perfectly understand the workings of the economy. In such a context, it is difficult to justify the apparently poor forecasting performance of the FOMC. Our defence of the FOMC therefore rests on asking what happens if the FOMC doubts how much the staff understands about how the economy works (Ellison and Sargent 2009). In our view of policymaking, the staff uses state-of-the-art but imperfect economic models to produce the best possible forecasts, but these forecasts are not taken at face value by the members of the FOMC. Instead, the FOMC suspects that the staff's model is imperfect and wants policies that will work well even if the staff model is misspecified.

The technicalities of how the to make decisions when the policy maker does not completely understand the economy are laid down in detail in the engineering literature on robust control. The basic idea is that the FOMC should pay special attention to events that give particularly bad outcomes. To achieve this, the FOMC needs to "exponentially twist" the forecasts of the staff by putting greater probability on bad outcomes that involve inflation and unemployment being a long way away from target levels. Twisting the staff's forecasts in this way requires the FOMC to construct worst-case scenarios, differing in their severity according to exactly how much the FOMC distrusts the model used by the staff. These worst-case scenarios are a key input to the decision-making process, because by responding to them the FOMC can design policy that works well even with imperfect understanding of how the economy operates. Our defence of the FOMC argues that the forecasts the FOMC publishes are exactly these worst-case scenarios, and they should not be interpreted as forecasts of what the FOMC thinks is going to happen. Instead, they are worst-case scenarios used to make decisions when the staff does not perfectly understand how the economy works.

Our equating of FOMC forecasts with worst-case scenarios immediately causes us to question the forecasting horse race run by the Romers. In our interpretation, the forecasts of the staff and the FOMC are incomparable, like apples and pears, because only the staff forecast can be fairly compared to actual outcomes. The FOMC forecast is a worst-case scenario that by construction is likely to be a poor predictor of future events if, as the Fed hopes, the staff's model is actually correct. In this light, it is not surprising that the Romers found that staff forecasts outperform those of the FOMC. It is what we would expect if the division of labour within the Federal Reserve System is as we have described.

Furthermore, if the FOMC does behave as we suggest, then the worst-case scenarios it publishes will definitely influence the policy actions actually taken; it is precisely when the worst-case scenario differs from the staff forecast that the FOMC needs to take pre-emptive policy steps. The finding of the Romers that the difference between forecasts predicts monetary policy actions can, therefore, be completely rationalised as due to the concerns of the FOMC that the staff’s model may be misspecified.

Some policymakers have gone on the record with arguments that support our view. For example, on 4 January 2008, Forbes reported on a discussion of the Romers’ paper given by former Federal Reserve Monetary Affairs Director Vincent Reinhart at the American Economic Association meetings in New Orleans:

"However, former Fed staffer Vincent Reinhart said while it may look as if ‘the FOMC's contribution to the monetary policy process is to reduce forecast accuracy’, they are not there primarily to be forecasters. Instead, they exist in a political system and have to be held accountable for the outcomes of their decisions. ‘They can be bad forecasters and good policymakers’, Reinhart said, ‘if the diversity of views about the outlook informs their policy choice.’"

The arguments we make amount to a spirited defence of the FOMC. Once we identify FOMC forecasts as worst-case scenarios, there is no need to reorganise the division of labour within the Federal Reserve System. In our story, policymakers do "use the information in the staff forecasts effectively" and do not "act on information that is of little or negative value". The model where the FOMC doubts the staff model is consistent with all of the Romers’ findings and explains the differences between FOMC and staff forecasts as a rational response of the FOMC to doubts about the specification of the staff model.


1 The FOMC sets monetary policy in the US. It is currently chaired by Ben S. Bernanke, and is comprised of members of the Federal Reserve Board and presidents of the regional Federal Reserve Banks.


Ellison, M. and T.J. Sargent (2009), “A Defence of the FOMC”, CEPR Discussion Paper 7510, October.

Forbes (2008), “Kohn says Fed operating with diverse views, not just strong chairman”, 4 January.

Romer, C.D. and D.H. Romer (2008), “The FOMC versus the Staff: Where Can Monetary Policymakers Add Value”, American Economic Review 98, 230-235, May.

Topics: Macroeconomic policy
Tags: Fed, forecast, monetary policy

Professor of Economics at the University of Oxford and Michael Cohen Fellow of Exeter College

W.R. Berkley Professor at New York University