Tyler Cowen directed me to a Greg Mankiw essay on inflation:
And as a textbook author, I sometimes get the suggestion that the whole discussion of the Phillips curve be removed from my books on the grounds that the idea is hopelessly out-of-date. By contrast, my own thinking is closer to that of George Akerlof (2002), who in his Nobel prize lecture called the Phillips curve “probably the single most important macroeconomic relationship.”
I have a hypothesis about the source of this disconnect. In the original paper by Phillips (1958) and the famous follow-up by Samuelson and Solow (1960), the Phillips curve was presented as an unconditional relationship—a negative correlation between inflation and unemployment. Some periods, most notably the 1960s, do exhibit a simple downward-sloping scatterplot of points. But that unconditional relationship is long gone. Using data for the past several decades, the scatterplot of inflation and unemployment is now a cloud of points.
My perspective, however, is that the Phillips curve is best considered a conditional relationship. If you think (as most economists do) that monetary shocks—or aggregate-demand shocks more generally—push inflation and unemployment in opposite directions in the short run, then there is a short-run Phillips curve. But this downward-sloping curve is conditional on what shock is hitting the economy.
Mankiw understands the problems with the traditional Phillips Curve, but would like to keep the basic concept, always keeping in mind the distinction between supply and demand shocks. My preference would be to mostly remove inflation from macroeconomics, and instead directly focus on the relationship between nominal GDP shocks and unemployment.
Mankiw identifies one problem with the traditional Keynesian approach:
Whenever inflation moves away from the Fed target, as it dramatically did in 2022, observers are tempted to attribute the change to a shock to aggregate supply or aggregate demand. That might provide some clue as to how transitory the change is likely to be and how much corrective action the central bank needs to take. The problem is that because we don’t know the natural rate of unemployment with much precision, it is hard to disentangle supply and demand. That is true even with the benefit of hindsight, but the task is even more formidable in real time when data are preliminary and incomplete. And it is in real time that policymakers need to respond. . . .
The 2022 inflation surge is a case in point. Even now, I don’t think we can say for sure what happened. The surge could have resulted from pandemic-related interruptions in supply chains. It could have resulted from excess demand, as the 3.6 percent rate unemployment rate was plausibly below the NAIRU, which may well have been altered by the pandemic experience as workers rethought their relationship with the labor market.
From an NGDP perspective, we know exactly what happened. Although there were supply problems in 2022, those have resolved over time. The cumulative excess 11% NGDP growth over the past 4 1/4 years fully explains the excess 9% in PCE inflation. It was essentially all demand side for the period as a whole, despite a portion of the 2021-22 inflation being supply side.
In contrast, if you look at unemployment as a way of ascertaining whether the economy is experiencing excess inflation, you run into exactly the problem identified by Mankiw, we don’t know what the natural rate of unemployment is at any given moment in time—even retrospectively.
To be clear, I am not saying the Phillips Curve model is wrong, at least in the sophisticated version preferred by Mankiw. Rather it is not useful to policymakers.
Mankiw is correct that inflation caused by demand shocks is associated with movements in unemployment that correspond to the basic logic of the Phillips Curve. It is quite plausible that there was a period of time during 2022-23 when unemployment was below the natural rate. The real problem is that it is not a useful tool for policymakers—they are better off trying to stabilize NGDP growth at 4% and avoid trying to guesstimate the natural rate of unemployment.
For similar reasons, I’d prefer the Fed try to stabilize NGDP futures contract prices along a 4% growth path, and not try to guesstimate the natural rate of interest, which is just as unobservable as the natural rate of unemployment.
PS. A few weeks back, Michael Sandifer developed a Chatbot trained on my money illusion blog posts from 2009-24. Eventually, I plan to do something with this, but I have not gotten around to it yet. If you scroll way down in the right column of this blog, you’ll find the link. If I’m hit by a truck, this Chatbot will take over my commentary on current events and do future movie reviews. (Let’s hope it doesn’t endorse Trump!) I guess that means I’m roughly 5% or 10% immortal.