Man sitting in front of a computer screen.

Blog Name

The pandemic era underscored how messy economic forecasting is. Here's how to deal with that.


Photo Credit: Ismagilov/Shutterstock

Author's note: I am grateful to William English, David Reifschneider, Olivier Blanchard, Patrick Honohan, and Joseph Gagnon for comments on an earlier version of this piece.


Like many other central banks around the world, including the Federal Reserve, the Bank of England failed to anticipate the inflation surge of 2021 and 2022. Now Ben S. Bernanke, former chair of the Federal Reserve, has been commissioned by the Bank of England to review the Bank’s “forecasting and related processes during times of significant uncertainty.” As the chair who led the Federal Reserve through the global financial crisis of 2008–10, Bernanke is arguably the best person to take on the job.

The Financial Times reports [paywall] that, unsurprisingly, the European Central Bank and the Fed have also undertaken similar, albeit less formal, retrospective looks at what went wrong, and how they might do better. The Bank of England’s move to commission a formal report prompts an interesting thought experiment: If the Federal Reserve were to request a similar outside review of its forecasting operations, what should it say?

First, an acknowledgement: I served for 30 years as a member of the staff at the Federal Reserve Board and make no pretense of being an objective observer. That said, I led the forecasting operation there for seven-and-a-half years, so I had a good understanding of how the staff forecast was developed, including its strengths and weaknesses. Perhaps more importantly, I understood how forecasts were used in the policymaking process. I am not up to date on changes that may have taken place since I left at the end of 2018 other than what has been discussed through public channels.

Forecast errors are a fact of life

Forecast errors are inevitable. They are a fact of life and probably will be forevermore. Yes, Fed policymakers and the staff at the Federal Reserve Board were wrong by a wide margin and for a long time about how serious a problem inflation would prove to be. I would have made the same mistake if I had still been in charge of the staff forecasting operation. A few observers issued warnings that turned out to be prescient about how big a problem inflation would prove to be, but—in a further illustration of how hard it is to forecast accurately—some of those same observers have been seriously wrong thus far about how long it would take to bring inflation back under control and the price that would have to be paid in the form of higher unemployment.

The fundamental cause of the failure by the Fed and most other forecasters to anticipate the extent of the inflation problem during the COVID era was that the shocks to the economy were huge and extremely unusual. These shocks included the massive shift from services consumption to goods, the lockdowns at national and factory levels, and workers’ sudden unwillingness or inability to work in person for many reasons, including fear of contracting the disease and because their children were home from school.

The previous global pandemic was in 1918–19, when the economy bore little resemblance to today’s and long before the modern era of economic statistics. A similar problem of unfamiliarity confronted forecasters when the financial crisis erupted in 2008; nearly 80 years had passed since the last threat of a general systemic collapse of the financial system. It’s hard to imagine economists ever being good at forecasting the consequences of events that occur so rarely and are so unusual in nature.

The good news is that an accurate forecast is not a prerequisite for a good monetary policy. To be sure, it’s better to have an accurate forecast than an inaccurate one, but a central bank can run a successful monetary policy even with an inaccurate forecast. The essential requirement for a successful policy is that policymakers be prepared to respond to surprises in the economic environment by adjusting their policy stance in a sensible, systematic, and sufficiently vigorous manner to suit the new circumstances.

Advances in forecasting have occurred

Of course, policymakers and staff should strive to have as good a forecast as possible. Models, computing power, and economic theory have advanced enormously over the past 40 years—so much so that future breakthrough improvements in forecast accuracy are not likely to stem from further advances along these lines. Instead, the bigger payoff is likely to come from improvements in “nowcasting”—the effort to accurately identify the current condition of the economy and measure how that condition is changing.[1] The Fed has been investing in such capability for years, especially with respect to employment and consumer spending. Less progress has been made in developing alternative measures of inflation. Even during the recent inflation burst, it took many months before the Fed recognized it needed to abandon its view that inflation was “transitory,” so it’s not obvious that a few weeks’ advance notice on inflation developments would have helped, but it’s worth looking again at whether more investment in this area is warranted.

The economic shocks from the pandemic caused the economy to behave abnormally. For example, for the past 18 months, inflation has slowed even while the labor market has remained tight, and the job vacancy rate has declined even while the unemployment rate has remained unchanged. Perhaps the most important lesson to be drawn from this experience is that periods of abnormal behavior are possible, and that central banks and others must constantly be attempting to identify when the next such period has begun. (Because such periods are, by definition, abnormal, they will be almost impossible to predict before the fact.) During periods of abnormal behavior, models useful in normal times may need to be set aside temporarily. The COVID-related period of abnormal behavior probably has not yet ended. The next big task for the Fed and other central banks will be to recognize when “normal” macroeconomic behavior has resumed.

One aspect of the Fed’s approach to monetary policymaking deserves another look: In September 2020, with the economy still reeling from the COVID-related shutdowns and the federal funds rate still pinned at its effective lower bound, the Federal Open Market Committee (FOMC) for the first time set out a multi-part test for when the Committee expected it would be appropriate to raise the funds rate: The labor force would be fully employed, and inflation would be running at 2 percent and on track to “moderately” exceed that pace “for some time.” The inflation part of this test was met much sooner than appeared to be the case with the employment part.[2]

The FOMC seemingly had provided itself with an escape hatch in the statement it first adopted in September 2020, by including these words: “The Committee would be prepared to adjust the stance of monetary policy as appropriate if risks emerge that could impede the attainment of the Committee’s goals.” But in real life, the FOMC seemed to interpret the three-part test as an ironclad commitment, possibly contributing to their tardiness in beginning to tighten policy. The comprehensive review slated for 2025 might provide an occasion for clarifying why they didn’t exercise the escape hatch and whether they might behave differently in the future.

Don’t overstate a forecaster’s knowledge

An important lesson from these complications is that in communicating to the public, policymakers should be careful not to overstate the extent of their knowledge or understanding of the future. They should deliver certainty where they can—with regard to their objectives and their determination to pursue those objectives. But they should be equally clear about where they cannot deliver certainty, including economic outcomes and the actions they will be required to take in pursuit of their objectives.

The FOMC already communicates extensive information about uncertainty. For example, figures 4.A, B, and C in the Summary of Economic Projections (SEP—see here for the most recent edition) show confidence intervals around the median projections for GDP growth, the unemployment rate, and inflation, respectively. As an example, figure 1 below shows the most recent median forecast among FOMC participants of PCE inflation (the red line) and an estimated 70-percent confidence interval, calculated from forecast errors made by non-Fed forecasters over the past 20 years.

Figure 4.D in the SEP shows that, through most of the COVID period, FOMC participants assessed the economic outlook as unusually uncertain; that impression has only recently begun to subside. For example, figure 2 below shows that every FOMC participant saw more uncertainty than normal surrounding their inflation forecasts from the middle of 2021 through the end of 2022. In the most recent set of forecasts, that proportion was down to 79 percent.

In light of the inability of economic forecasters to see the future clearly—especially in times of stress when the economy is more likely to behave in abnormal ways—some forms of “forward guidance” should not be provided.[3] Specifically, it’s almost surely inadvisable to make calendar-based promises. (An example of calendar-based forward guidance would be a statement along the lines of: “we won’t raise the policy rate for another two years.”) With the benefit of hindsight, it’s probably also inadvisable to set complicated conditions that depend on multiple circumstances being met. (An example of this type was the statement they began using in September 2020 to the effect that they wouldn’t raise the funds rate until they judged the workforce to be fully employed, and inflation to be back at 2 percent and on track to move above that level.)

In normal times, when the economy is behaving in textbook fashion, promises of that type probably wouldn’t entail much risk, but such promises don’t get made in normal times. Therefore, in its upcoming framework review, the FOMC might consider committing to confining itself to simpler forms of forward guidance—for example, contingent on only one economic variable rather than the joint behavior of multiple variables—or committing to exercise an escape clause if circumstances warrant.

The United Kingdom is reportedly considering using alternative scenarios to communicate more effectively about the sources and extent of uncertainty in the economic outlook. That approach probably is feasible for the Bank of England, given that their policymaking committee adopts a true consensus outlook. But it’s hard to see how the Fed could do something similar.

The Fed staff based in Washington has long included alternative scenarios in the forecast document prepared for the FOMC. But in preparing those scenarios, the staff works from its own baseline forecast, using a particular model to describe the evolution of the economy under the hypothesized scenario. Staff members also use a specific monetary policy rule to trace one possibility for how the Committee might respond. FOMC participants have never been willing to sign onto a consensus forecast—baseline or otherwise—much less a monetary policy rule that would govern their responses to off-baseline developments. Neither have they ever endorsed a specific model as providing a good enough description of the economy. Unless that changes, the best they can probably do is to continue communicating incessantly about their subjective perceptions of uncertainty.

Monetary policymaking needs to be designed mainly for normal times

In summary, the Fed did not perform perfectly through the COVID era. It would have been handy if they’d had an unclouded crystal ball, but like other central banks, they didn’t. Real damage was done as a consequence of the inflation outbreak—damage that could end up being severe and long-lasting. Therefore, it’s important to review the record of the past three years dispassionately and thoroughly. But it’s also important to learn the right lessons from that period.

The most important lesson of all might be how unusual the events of the past three years have been. Policymakers would be well advised to avoid changing too much in the wake of an extremely unusual episode. Monetary policymaking needs to be designed mainly for normal times, and the past three years have been anything but normal.


1. One excellent example of the value of these efforts is the fact that the staff was able to measure the extent of the labor market collapse just after COVID erupted, weeks before the official government data corroborated it.

2. Recall that the unemployment rate remained above 6 percent through April 2021 and didn’t move below 4 percent until the end of that year. By that time, personal consumption expenditure (PCE) inflation was 6.2 percent and rising.

3. When the interest rate it controls has been driven down to its lowest-possible level, it’s become common for central banks wishing to provide more economic support to resort to describing when or under what circumstances they will move up their policy rate up from the lowest-possible level.

Data Disclosure

This publication does not include a replication package.

More From

More on This Topic

Related Topics