Don't build your castle on shifting sands: the Phillips Curve(s) edition

The Phillips curve, which tells us the tradeoff between unemployment and inflation is a mainstay of monetary policy. But it moves around, making it risky to use.

Today’s post on the Phillips Curve is primarily for paid subscribers. I discuss a speech by Fed Governor Chris Waller, “The Unstable Phillips Curve,” below the paywall; above it, I sketch out the fundamental issue of its instability.

Source: Getty Images.

The Fed will hike again next week in its ongoing efforts to ‘cool down’ the economy. At this point, it sees the high demand from consumer spending as the primary cause of high inflation. And where there’s high demand, there’s low unemployment. And so, for several months, we have been told that lower inflation will require the tough medicine of higher unemployment. But exactly how much? Many macroeconomists and central bankers turn to the Phillips Curve for the answer.

Here’s one example in June of 2022, when inflation was 9%. Larry Summers argued:

We need five years of unemployment above 5% to contain inflation -- in other words, we need two years of 7.5% unemployment or five years of 6% unemployment, or one year of 10% unemployment,” …

The US may need as severe monetary tightening as Paul Volcker pushed through in the late 1970s early 1980s.

How did Summers come up with these dire numbers? The Phillips Curve. How did using it turn out? So far, not so well. Last June, CPI inflation peaked at 9% in the United States, and it has since dropped 4 percentage points to its current level of 5%.¹ And during that period of substantial disinflation, unemployment held steady at around 3.5%—considerably less inflation without more unemployment.²

Some refer to that as “immaculate disinflation,” but really, it’s just disinflation that their Phillips Curve failed to predict. And that should be a wake-up call for those people using the Phillips Curve in real-time to calibrate policy. Let’s be very careful and modest.

Phillips Curve is unstable when we need it most.

The data show that the Phillips Curve is unstable.³ In the scatterplot of monthly inflation and unemployment—with different time periods grouped together by color—the relationship changes. We went from two decades of basically no relationship between unemployment and inflation, that is, a flat Phillips Curve (shown by the gray dots) to a steep Phillips Curve in 2021 (orange), to a vertical one since then (blue).

Not surprisingly, flat, steep, or vertical makes a huge difference in predictions about how much unemployment we need to bring down inflation. Specifically, a flat one will tell you that it will take a lot of unemployment to bring inflation down. With a vertical one, it does not take much unemployment. Do you need a recession or not is a big difference. Using a tool as unstable as the Phillips Curve for policymaking is dangerous. See my other critical posts on the Phillips Curve, too: here, here, and here.