Ask any Federal Reserve chair how far the policy rate sits from neutral and you’d normally get a number, or at least a range. Asked exactly that on 16 September, current Chair Kevin Warsh provided neither. He said the comparison is “useful academically … a discussion to help us think about policy,” but that it carried no “operational effect” on what the Federal Open Market Committee (FOMC) decided that day.
Warsh’s messaging is a notable departure from recent norms in how the Fed describes its policy stance. The real neutral interest rate has been a core component of FOMC communication since Janet Yellen was chair from 2014–2018 (read her 2015 speech). Before Yellen, Chair Ben Bernanke leaned on the same idea: implicitly attributing the drop in the neutral rate after the 2008–2009 global financial crisis (GFC) to economic scarring and the balance sheet repair that followed, packaged publicly as “headwinds” (see his 2012 speech).
However, since being appointed chair this year, Warsh has framed the discussion differently. At Jackson Hole and again in September, he said he’d “be hard-pressed to describe broad financial conditions as restrictive” – a view he said is “widely shared by the committee.” This framing relates the stance of monetary policy and, more specifically, changes in the Fed’s policy rate to how those changes affect broader asset market prices and spreads – credit spreads, equity valuations, and borrowing costs across the economy – rather than the gap between the policy rate and a model-implied neutral rate.
The distinction matters and has near-term implications for Fed policy. Several (but not all) FOMC participants have described policy this year as somewhere between neutral and mildly restrictive – a stance that is at least theoretically positioned to mitigate temporary inflationary pressures.
At the same time, broad financial conditions in the U.S. have been remarkably stable despite energy market disruptions. Higher real rates – which by themselves should tend to restrain economic activity – have coincided with robust equity returns. The S&P 500 sits roughly 13% above where it started the year, supporting consumption through the wealth channel.
Viewed through this lens, September’s Fed rate hike may have been aimed at helping prevent financial conditions from easing further; such easing would potentially add to demand-side inflation pressure. Ahead of the September meeting, markets were pricing an elevated chance of a 25-basis-point (bp) hike. As a result, holding the policy rate steady would have been a notable surprise.
Looking ahead, markets are priced for additional hikes. And our base case is that the FOMC will likely deliver one or two more 25-bp rate hikes through this year and into early next. However, looking further out, anticipating appropriate Fed policy through a financial-conditions-targeting framework has its own limitations. Hence, a neutral rate anchor is still useful.
Ultimately, our outlook is for the Fed to recalibrate its policy stance to position for the risk that inflation doesn’t dissipate. However, as the temporary factors that have been supporting inflation fade – tariffs, energy, and AI-related computing equipment price adjustments – and the FOMC gains more confidence that inflation isn’t persistent, further adjustments beyond that may not be necessary.
A further complicating factor is that r* estimates are time-varying as structural forces in the economy change. After falling for the better part of three decades, model estimates of r* have risen about 100 bps post-pandemic, from roughly 0.5% to 1.5%. Although the factors driving r* are widely believed to be related to the evolution of potential growth and savings and investment preferences, the main driver recently appears to be higher potential growth, which the models that publish it have marked up from about 1.8% to 2.5% (see Figure 2). According to U.S. Commerce Department data, U.S. labor productivity growth – a key component of potential growth – shifted higher post-pandemic even before any widespread AI implementation and adoption could reasonably be expected to show up in the statistics.
The U.S. interest rate curve has repriced accordingly: Term-structure estimates of the expected average short rate embedded in the 5-year, 5-year-forward rate have also risen roughly 100 bps over the same period (see Figure 3).
Whether r* rises further still is another key question. The structural forces behind the original decline aren’t easily reversed, but AI is likely to be a powerful force that fundamentally reshapes the economy. A widely cited analysis is by Rachel and Smith (2015), who documented a roughly 450-bp fall in global long-term real rates over the preceding three decades. The trend was common to developed and emerging economies alike, pointing to a shift in the global neutral rate rather than country-specific factors.
Notably, trend growth – the channel doing most of the work in today’s U.S. growth estimates – explained little of the pre-GFC decline in Rachel and Smith’s analysis. Global growth was fairly steady through those decades, and it was the crisis itself that triggered a broader reassessment of growth prospects. Most of the move, therefore, came from saving and investment preferences: Longer lifespans and rising inequality pushed desired saving up, while desired investment fell with the declining relative price of capital and lower public investment.
Fast-forward to today: If AI delivers sustained productivity gains, it should push r* up, as should higher desired investment. But the societal uncertainty AI creates could also lift precautionary saving. In prior analysis we’ve found no evidence yet in U.S. Treasury yield behavior that AI model releases are moving market-based r* estimates, although that doesn’t necessarily mean AI won’t affect r*. (See a related discussion in the 1 July edition of Macro Signposts.)