Why people can’t agree on where interest rates are going

Stay informed with free updates

Simply sign up to the US interest rates myFT Digest — delivered directly to your inbox.

Everyone is struggling to see where interest rates are headed. Investors are jittery, as shown by gyrating long-term Treasury yields. America’s central bankers are trying to project calm, but they are in a fog too. On August 25, Federal Reserve chair Jay Powell summarised the mood when he said “we are navigating by the stars under cloudy skies”. Economists do have some tools to illuminate the path ahead. But they aren’t very helpful.

The object everyone is searching for is the neutral rate of interest, or R-star for short. (Economists seem to struggle with nicknames.) It is the (real) rate that neither buoys nor depresses the economy once temporary shocks have faded away. Central bankers believe that they can neither influence it nor observe it. Their task is merely to divine it.

Although most agree that over recent decades R-star has fallen, its recent moves are more mysterious. An estimate published by the Richmond Fed suggests that it fell from around 2.2 per cent in April 2008 to 0.8 per cent two years later, but by April 2023 had recovered. By contrast, an estimate from the New York Fed finds that in April 2023 it was still around two percentage points lower than before the global financial crisis. In April, the IMF used a more complicated model to argue that it is probably still very low.

These estimates diverge because of different trade-offs made by their designers. One approach is to make lots of assumptions about how the economy works to strip out the noise associated with shocks. But it suffers the risk that the assumptions — and therefore the results — are rubbish. An alternative relies more on recent data. But that risks results that reflect temporary shocks, not the future once they fade.

Take the data-heavy method deployed by the Richmond Fed, which uses a very sophisticated moving average to forecast long-run rates. Given the recent resilience of America’s economy, it should probably be no surprise that it suggests a rising R-star. Unfortunately, it suffers from statistical error bands the size of a bus. Although the median estimate is 2.3 per cent, the lower bound is 1.4 per cent and upper bound is 3.6 per cent. That is about as useful as being told that one’s Friday night pizza will arrive anytime between 6pm and 11pm.

The New York Fed’s method uses more theory. It assumes a relationship between inflation, the position of the economy relative to its potential and interest rates, then brings in data to infer the position of R-star. The cost of this approach became apparent over the pandemic, when the model was spitting out such implausible numbers that it was temporarily suspended. Now the tweaked model describing America’s R-star is back.

The approach deployed by the IMF is the most theory-driven of them all. In the long run, factors like demographic change, productivity growth and fiscal policy should influence the balance of savings and investment. And the model delivers fantastically detailed breakdowns of exactly how much. Between 1975-79 and 2015-19, demographic change tugged down America’s R-star by 0.5 percentage points, and weak productivity growth by 1.23 percentage points.

Again, the danger is that these results tell you more about modelling than reality. Particularly if you believe a 2017 study from the Bank for International Settlements which argued that the simultaneous events of falling real interest rates, stalling productivity growth and ageing populations “appear coincidental”. Scouring data between 1860 and 2016, they reckon that changing monetary policy frameworks matter more.

The other way of measuring R-star is to look at the long-term interest rates implicit in investors’ pricing. One interesting study calculates the long-term interest rate implicit in the value of British flats before and after their leases are extended. It concludes that the natural rate of return on capital has not risen much since the pandemic. But that rate includes a risk premium associated with owning British property, and may be affected by other distortions. And of course, says Atif Mian, one of the study’s authors, the collective wisdom of the housing market “could be wrong”.

An anxious central banker could drive themselves mad worrying about the uncertainty ahead. What if R-star is indeed slow-moving, but the pandemic has revealed that we overestimated its fall during the 2010s? What if the expectations of central bankers and investors feed off each other? It’s probably not much comfort to say that these estimates of R-star are the best we have. But it should be more soothing to say that any mistakes will be just as hard to pin down.

soumaya.keynes@ft.com