A shifting benchmark for monetary policy
The study by Farkas, Jakab and Lindé examines a puzzle left by the 2022-2024 tightening cycle: economic activity and labor markets held up even as standard estimates suggested that monetary policy had become highly restrictive.
Existing explanations emphasize strong demand backed by expansionary fiscal policy. They also point to central banks' initial delay in raising rates because they expected supply-chain disruptions and energy-price pressures to fade quickly, limiting the damage to activity from those shocks. The paper adds another explanation: the benchmark against which monetary restraint is measured may itself have risen.
That benchmark is the natural rate of interest, or r*: the short-term real interest rate consistent with output under flexible prices over the long run. Policy restraint depends on how far the actual real policy rate sits above r*, not simply on how high nominal rates have climbed. A simultaneous increase in r* would therefore offset part of the tightening.
Using information about expected policy rates, the authors estimate that post-pandemic US r* increased from roughly -0.5% to 1.5%. Their euro-area estimate moved from about -2% to almost zero. Those changes amount to around two percentage points in the US and 1.5 to 2 percentage points in the euro area.
Why policy expectations change the estimates
Because r* is unobservable, researchers must estimate it with a model. Farkas, Jakab and Lindé incorporate expectations of future policy rates into a medium-scale dynamic stochastic general equilibrium, or DSGE, model for each economy. The framework includes nominal rigidities, financial frictions, anticipated monetary-policy shocks and the value investors place on holding safe, liquid debt securities.
This last component, known as the convenience yield, reflects benefits beyond an asset's financial return. Investors willing to pay more for safety and liquidity accept a lower yield. A larger convenience yield thus pushes down safe real interest rates and r*.
Alongside conventional macroeconomic data, the researchers use overnight index swap rates, long-term government bond yields, corporate bond spreads and long-term inflation expectations. These help the model separate expected policy actions from convenience-yield changes and other forces affecting the natural rate.
Expected policy shocks cannot directly move r* within the model. Instead, information about those expectations changes which underlying shocks the model identifies as responsible for observed economic movements. Including overnight index swap policy rates one to six quarters ahead, together with anticipated policy shocks, lowers the filtered r* estimates by more than one percentage point in the aftermath of the global financial crisis.
The effect on the post-pandemic picture is a larger estimated rise in r* than models without expected policy rates produce. The widely used Holston-Laubach-Williams model, by contrast, shows little change. For that comparison, the researchers use its real-time estimates from the second quarter of 2025.
Safe-asset premiums account for much of the rise
The benchmark model puts the US increase in r* since 2020 at 1.9 percentage points, with convenience-yield innovations accounting for 1.5 points. For the euro area, it estimates a 1.7-point increase, including a 1.8-point contribution from convenience-yield innovations. The authors interpret the rise mainly as an unwinding of forces that had depressed safe real rates after the global financial crisis, rather than as a conventional improvement in trend growth.
Their US estimate is also larger than the roughly one-percentage-point post-pandemic increase reported by Del Negro, Elbarmi and Pham in a 2026 vector autoregression model with convenience yields. Farkas, Jakab and Lindé attribute the additional increase in their estimate entirely to convenience-yield innovations, which they link to a reversal of anticipated policy shocks.
A related argument appeared in ECB Executive Board member Isabel Schnabel's February 25, 2025, keynote. She said an increase in the net supply of government bonds had turned the global "savings glut" into a "bond glut," reducing the premium for safety and liquidity and lifting r*. Citing IMF analysis, she noted that a decline in convenience yields to their pre-2000 average could raise natural rates by about 70 basis points.
The working paper also explores how these shocks differ across the Atlantic. In a separate simulation, it substitutes euro-area convenience-yield innovations for US innovations from the euro's inception in the first quarter of 1999. Under that scenario, US r* and economic activity would have been notably lower, with inflation close to zero after the 2008 global financial crisis. The unconstrained, or shadow, federal funds rate would have fallen further below zero, making the zero lower bound a more binding constraint.
What an unchanged natural rate would have meant
To assess the consequences for the recent tightening cycle, the authors run a counterfactual in which r* follows the flatter Holston-Laubach-Williams path rather than their estimated rising path. The resulting real-rate gap eventually becomes about two percentage points larger in the US and slightly more than 1.5 percentage points larger in the euro area, implying tighter policy.
In that model exercise, the tighter gap lowers output relative to trend by nearly 4% in the US and 2% in the euro area. US growth would have been approximately zero from 2022 onward, while euro-area growth would have remained around -2% for an extended period. These are simulated outcomes, not observed losses.
The inflation response in the exercise is muted because the model's estimated price and wage Phillips curves are relatively flat. The authors caution that inflation effects could be stronger if those relationships vary with economic conditions or involve nonlinear effects. They present the counterfactual as a complement to explanations involving strong demand, expansionary fiscal policy, adverse supply shocks and nonlinear pricing behavior, not as a complete account of post-pandemic inflation.
Other measures show increases, with substantial uncertainty
The broader research does not yield a single estimate of r*. A March 2024 BIS Quarterly Review feature by Benigno, Hofmann, Nuño Barrau and Sandri found significant post-pandemic increases in most established US and euro-area measures, sometimes reaching pre-global-financial-crisis levels. Holston-Laubach-Williams was the notable exception. With the highest and lowest estimates separated by about 1 to 2 percentage points, the BIS authors described r* as a "blurry guidepost" for policy.
Farkas, Jakab and Lindé also compare their results with the Lubik and Matthes measure. It tracked the Holston-Laubach-Williams series closely during 2009-2022, then rose by about one percentage point in 2023-2024. At the end of the sample, it stood slightly above their own DSGE estimate.
A May 12, 2025, San Francisco Fed Economic Letter by Christensen and Mouabbi offers another euro-area estimate, using a yield-curve model fitted to French inflation-indexed OAT€i bonds. It finds that r* declined from above 1.5% in 2002 to below -1.5% at the end of 2021, before recovering about half of that decline. The authors project a further modest increase over the next decade and conclude that euro-area interest rates are unlikely to revisit their pre-pandemic lows.
The Farkas, Jakab and Lindé paper's conclusions are the authors' views and should not be attributed to the IMF, its Executive Board or its management.
