Inflation has behaved strangely over the past decade.
The recession that followed the financial crisis of 2007-08 sent American unemployment soaring to 10%.
But underlying inflation fell below 1% only briefly—nothing like the fall that models predicted.
Because the only way economists can estimate the natural rate is by watching how inflation and unemployment move in reality, they assumed that the natural rate had risen (an estimate in 2013 by Robert Gordon, of Northwestern University, put it at 6.5%).
Yet as labour markets have tightened—unemployment was 4.3% in July—inflation has remained quiescent.
Estimates of the natural rate have been revised back down.
Such volatility in estimates of the natural rate limits its usefulness to policymakers.
Some argue that the wrong data are being used, because the unemployment rate excludes those who have stopped looking for work.
Others say that the short-term Phillips curve has flattened as inflation expectations have become ever more firmly anchored.
The question is: how long will they remain so?
So long as low unemployment fails to generate enough inflation, central banks will face pressure to keep applying stimulus.
Their officials worry that if inflation suddenly surges, they might lose their hard-won credibility and end up back in 1980, having to create a recession to get inflation back down again.
This recent experience has led some to doubt the very existence of the natural rate of unemployment.
But to reject the natural rate entirely, you would need to believe one of two things.
Either central banks cannot influence the rate of unemployment even in the short term, or they can peg unemployment as low as they like—zero, even—without sparking inflation.
Neither claim is credible.
The natural rate of unemployment surely exists.
Whether it is knowable is another matter.