If it ain't broke . . .
By Scott Sumner
Paul Krugman has a new Substack, and his first post revisits the famous cycle of rising inflation and disinflation during the 1970s and 1980s, which led to a revolution in macroeconomic theory. The subtitle is:
Did we get the whole macro story wrong?
I’m going to argue that the answer is “no”.
The model that won out in the 1970s and 1980s was mostly developed by monetarists like Milton Friedman, who argued that the Phillips Curve was an unreliable policy tool and that expansionary demand-side policies would have only a temporary effect on unemployment. Once expectations of inflation caught up to reality, unemployment would return to its natural rate.
Krugman suggests that he initially accepted much of this thesis, but now has some doubts. In my view he’s focusing too much on the Keynesian interpretation of Friedman’s argument, which makes the data look more puzzling than it actually is.
In Friedman’s Natural Rate model, unexpectedly high inflation causes low unemployment, and vice versa. The Keynesian version of the same model reversed the causation. Now it’s low unemployment (i.e. “economic overheating”) that causes high inflation.
To give a sense of how this distinction matters, consider this comment by Krugman:
The truth is that I’ve always been a bit uneasy about the standard story of inflation in the 1970s, even though it seemed to fit the prediction of clockwise spirals. My uneasiness came from the sense that the economy never seemed to run hot enough to explain such a big rise in inflation. I actually remember the 70s! And while there were years of good job markets, they never felt as good as the 60s, the late 90s, or 2019.
But that’s not at all a problem for monetarist theory, as the monetarists always insisted that inflation was not caused by a hot economy, it was caused by rapid money supply growth. And monetarists also suggested that the effect would be quite transitory, with unemployment returning to its natural rate after a few years. And finally, the natural rate was itself volatile, and could not be used as a guide to stabilization policy. (To be fair, the preferred monetarist indicator, money supply growth, also eventually turned out to be faulty.)
Rapid growth money pushed unemployment down below 4% in the late 1960s, as inflation soared. This fits the monetarist model. Then it rose back to its natural rate during the 1970s, rising above that rate during occasional periods of disinflation or adverse supply shocks. This also fits the monetarist model.
So Krugman’s right that the 1970s don’t fit the Keynesian interpretation of Friedman’s Natural Rate model (low unemployment cause high inflation), but the data does fit Friedman’s actual Natural Rate Hypothesis, if you assume rapid money growth and sprinkle in a few supply shocks. The Keynesian model says high inflation should have delivered a booming 1970s, the monetarist version does not. To the monetarists, the employment gains from stimulus were mostly exhausted by 1969.
Much of the discussion is framed in terms of the slope of the “Phillips Curve”, which I view as an unhelpful construct. People debate whether the Phillips Curve is flat or steep, which is like debating whether the price and quantity combination in the apple market is flat or steep. It entirely depends on the nature of the shocks hitting the economy.
If inflation fluctuates wildly between 0% and 12%, as it did from 1945 to 1982, and unemployment also fluctuates, then the Phillips curve will likely not be flat. If a central bank successfully targets NGDP growth at a stable rate, then the Phillips Curve will likely slope the wrong way. And if the central bank keeps inflation close to 2% (as they’ve mostly done since 1990), and unemployment moves around for reasons unrelated to inflation, then the Phillips Curve will be fairly flat. But the slope of the Phillips Curve is not a deep parameter of the economy; it’s an outcome that is contingent of the sorts of real and nominal (or supply and demand) shocks hitting the economy.
Perhaps the most interesting aspect of Krugman’s post is his discussion of some research by Jonathon Hazell, Juan Herreño, Emi Nakamura & Jón Steinsson (HHNS), which re-examines the Volcker disinflation:
Now, the Friedman/Phelps story behind clockwise spirals did involve changing expectations: high unemployment was supposed to lead to lower actual inflation, which would over time be reflected in lower inflation expectations, which would feed through to further inflation declines. But the 80s decline is too fast to be explained that way, and seems to have started a bit before actual inflation came down.
They [HHNS] suggest that there was a regime shift: When people realized that Volcker was really willing to put the economy through the wringer, they marked down their expectations of future inflation in a way that went beyond the mechanical link via unemployment.
I think HHNS are correct, but not because Friedman was wrong. As noted above, Krugman seems to assume that Friedman advocated the Keynesian interpretation of the Phillips Curve—unemployment causes low inflation. Instead, Friedman argued that high unemployment is caused by lower than expected inflation. And both are ultimately caused by tight money.
Nonetheless, the HHNS research does present one problem for Friedman’s monetarism—why did inflation expectations fall quickly with a decline in the actual rate of inflation? Friedman used an adaptive expectations model, where a decline in actual inflation should lead to a decline in expected inflation with a substantial lag.
Krugman then discusses other examples of where “regime changes” quickly broke the back of inflation, such as Spain after joining the euro.
Fortunately, there is another model that can explain the HHNS findings—rational expectations. When the public saw that Volcker was serious about reducing inflation, the expected rate of inflation fell faster than what one would predict using an adaptive expectation model. And this also explains why inflation fell faster than predicted by Keynesian models that focus on causation going from unemployment to disinflation. Unfortunately, Krugman has already dismissed the usefulness of rational expectations models:
Second, since the Friedman/Phelps prediction was based on trying to assess what rational price-setters would do, their apparent success gave a big boost to the notion that all economics should be based on maximizing behavior. Friedman always had too strong a reality sense to personally go down the rational-expectations rabbit hole that swallowed much of macroeconomics, but given the law of diminishing disciples it was bound to happen.
Not surprisingly, Krugman fails to draw the conclusion that HHNS’s interpretation of the Volcker disinflation is a big win for rational expectations models.
In fairness, the most extreme forms of Ratex models don’t fit the Volcker disinflation. Some economists argued that inflation could be brought down costlessly if the policy of disinflation were fully credible. I always doubted that view, due to sticky nominal wages. Furthermore, in practice any disinflation will almost never be fully credible. After all, Volcker first tried to bring inflation down in early 1980. Then, after unemployment soared in the spring, Volcker reversed course and cut rates sharply (before the November election), and then made a renewed attempt to bring down inflation in the spring of 1981. This time he persisted, but can you blame the public for being somewhat skeptical?
Krugman is right that macro took a wrong turn in the 1980s, and is also correct that the conservative wing of the profession was especially prone to going down “rabbit holes”. But the actual problem was not too much reliance on the rational expectations; it was too much reliance on “classical” models where labor and goods markets are assumed to be in equilibrium. (Actual classical economists believed in sticky wage models.)
To summarize, unemployment rose sharply during the Volcker disinflation, but if one uses a Keynesian model then the rise was not large enough to fully explain the Volcker disinflation. Friedman’s adaptive expectations model of inflation also fell short. Instead, the best model seems to be rational expectations combined with an assumption that the Volcker disinflation was partly anticipated and partly unanticipated—as if the public thought he had perhaps a 50% chance of successfully bringing inflation down before political pressures forced him to relent.
To me, that seems like a triumph of Chicago school economics (before it went off course), not an unexplained phenomenon that cries out for a new explanation.
PS. I don’t like either the (old) monetarist or the Keynesian view of causation (P –> U or U –> P). Instead, monetary policy causes NGDP growth, which causes trend inflation. Variations in inflation are caused by either supply shocks (when NGDP growth is stable) or demand shocks (variations in NGDP growth.) Unemployment fluctuations are mostly caused by unanticipated moves in NGDP growth, i.e. “monetary policy”, properly defined.
HT: Tyler Cowen