The Federal Reserve’s costly gamble on inflation

The Federal Reserve could raise interest rates in September, with the decision likely to remain uncertain until the meeting, whilst one question deserves greater attention in the public debate. Why do its forecasts continue to predict a return of inflation to 2 per cent, which is then consistently postponed? My suspicion centres on an assumption adopted by the Board’s staff at the end of 2023, which might help explain such confidence in a process of disinflation that is regularly postponed.

In March 2023, the median forecast by Fed members indicated that core PCE inflation – excluding food and energy – would stand at 2.1 per cent in 2025; in June 2024, a return to 2 per cent was forecast for 2026; in March 2025, the target shifted again, this time to 2027. According to my calculations based on the projections published since 2023, forecasts for the following year have generally underestimated inflation by around half a percentage point or more, with errors reaching around one percentage point and no substantial improvement between March and December. For the two-year forecasts, the average underestimation was close to one percentage point, exceeding it in some cases.

For a central bank with a 2 per cent target, such large and persistent errors require an explanation, which may lie partly in bad luck. Wars, tariffs and supply disruptions have complicated forecasting, whilst high federal deficits have made it more difficult to curb demand. The Fed was in good company in being caught off guard, but as early as 2022 there were serious reasons to believe that the US economy was exposed to negative supply shocks that were more frequent, more severe and more persistent than in the past. To expect it to have foreseen the next war would have been absurd, whilst attributing a higher probability to stubborn inflation – not least in light of the size and persistence of the deficit – would have been reasonable.

A second explanation concerns the assumptions embedded in the models, starting with the Phillips curve used by the Fed, in which inflation depends on an underlying trend, its own past dynamics, pressures on the labour market and shocks such as rises in energy or import prices. When these shocks subside and the labour market returns to equilibrium, projected inflation converges towards the assumed trend, at a rate that depends on the model’s dynamics and the persistence of the shocks. Setting that trend at 2 per cent does not automatically guarantee that it will be reached within two years, but it determines the target level when it is assumed that other pressures will also ease.

A report by Ekaterina Peneva, Jeremy Rudd and Daniel Villar, published in August 2025, reconstructs the editorial team’s decisions. Before the pandemic, the trend was set at 1.7 per cent; in December 2021, a rule was introduced allowing the trend to rise in response to high inflation, adding a maximum of around half a percentage point to the forecasts; in December 2023, however, the trend reverted to a constant level, assuming it had returned to 2 per cent by the end of 2022. In other words, since 2023, the staff has assumed that the inflation trend was equal to the target – a highly significant decision that would have warranted a far more in-depth public debate.

The projections by Fed members – each contingent on the monetary policy deemed appropriate by the individual participant – must be distinguished from staff forecasts, and their similarity does not prove that this assumption caused the errors. It does, however, merit serious examination as a possible source of recurring optimism, because the inflation trend was still, until recently, at least around 2.5 per cent – plausibly between 2.5 and 3 per cent – with significant persistence above the target. It is an unobservable variable; therefore, this assessment too must be tested against the data, bearing in mind that errors may stem from an underestimated trend, from shocks that are more persistent than expected, or from both.

We do not know whether the staff subsequently revised this forecast, nor can we infer this from public projections alone, whose convergence towards 2 per cent could also reflect the expected effects of monetary policy. However, after more than five years of inflation above target, assuming that the trend is already at 2 per cent remains a risky empirical gamble, because an estimate that is too low may repeatedly suggest disinflation that requires greater monetary tightening or more time than anticipated. Chairman Kevin Warsh should be asked whether this assumption is under review, what evidence supports it and what evidence would lead to it being abandoned, because a central bank must be able to revise its views when the data continues to contradict them.