The standard (Keynesian) view of the Phillips Curve is that a strong economy leads to higher inflation. If I’m not mistaken, Milton Friedman reversed the causation, arguing that higher than expected inflation led to a stronger economy:

There is always a temporary trade-off between inflation and unemployment; there is no permanent trade-off. The temporary trade-off comes not from inflation per se, but from unanticipated inflation, which generally means, from a rising rate of inflation. The widespread
belief that there is a permanent trade-off is a sophisticated version of the confusion between ‘high’ and ‘rising’ that we all recognize in simpler forms. A rising rate of inflation may reduce unemployment, a high rate will not.

That’s also my view of causality, although I think inflation is the wrong variable. The model should use the rate of growth in NGDP, not prices.

Here’s Nick Rowe:

Andy Harless’ tweet (about the US economy) got me thinking.

“There’s a frog-boiling aspect to this economy. The consistent lack of *rapid* improvement throughout the recovery is enabling us to reach levels of employment that might not otherwise have been attainable.”

It reminds me of my old post “Short Run ‘Speed Limits’ on recovery“.The basic idea is simple: actual inflation (relative to expected inflation) might depend not just on the level of employment (relative to some unknown level of “full employment”), but also on the speed at which employment increases.

I’ve added an epicycle to the Phillips Curve that I think makes it fit the facts better. But I added that epicycle 10 years before the facts that Andy’s tweet asks us to explain. And it’s based on an idea that make sense, and goes back further still:

It’s difficult and costly to increase employment quickly, and easier and cheaper to increase employment more slowly, even for the same cumulative increase in employment. So if demand for output suddenly increases by (say) 10%, individual firms will raise prices and wages relative to the prices and wages they expect at other firms, or raise them more than they would otherwise have done if demand for output had slowly increased by that same 10%. Even with no underlying trend growth in productivity. Even with the same average level of demand for output.

There’s another way of thinking about this question. Inflation is not determined by economic slack, rather it’s determined by monetary policy. When monetary policy is highly expansionary and prices rise much faster than expected, then output tends to rise rapidly. That’s Friedman’s view. Thus it’s no surprise that a subdued rate of inflation is associated with a slow recovery. Phillips curve models that predicted otherwise, i.e., mainstream Keynesian models, are simply wrong.

I do believe that Andy and Nick are making valuable observations here, albeit not because they provide a useful tweak to the Phillips curve theory. It’s better to simply drop the Phillips curve and focus on other models, such as the relationship between unexpected NGDP growth and changes in employment.

Instead, Andy and Nick are showing that the natural rate of unemployment is a slippery concept. If inflation is stable (or better yet if NGDP growth is stable), then the economy will gradually move toward its natural rate. Because of costs of adjustment, however, the fact that unemployment is currently above (or below) the long-run natural rate does not mean that monetary policy is off course, even if inflation (or NGDP growth) is exactly on target, and even if the Fed has a dual mandate. Monetary policy is off course if expected future inflation/employment outcomes are not consistent with the Fed’s dual mandate, as was the case during 2008-16

During 1933, both prices and NGDP rose rapidly, and yet unemployment was roughly 25%. Now you could certainly argue that even faster nominal growth would have been desirable. But even with appropriate monetary policy, unemployment in 1933 would have been well above any reasonable estimate of the natural rate.

Some people argue that the current 3.6% unemployment rate shows that money was too tight a couple of years ago when unemployment was 4.1%. That’s not the case. Money might have been slightly too tight in 2017, but only because inflation was also running a bit below target. The optimal unemployment rate might well have been 4.1% in October 2017 (due to costs of rapid adjustment) and 3.6% today. But it certainly was not 10% in October 2009.

Even when monetary policy is producing appropriate nominal stability, it would not be unusual to see the unemployment rate gradually falling (or rising) toward its long-run natural rate.