Scott Sumner  

The Fed relies on a model that only works when its policy fails

Robert Murphy on Minimum Wage ... Kling on Phishing for Phools...

I've made this point before, but it's worth repeating, given all the recent talk about the Fed relying on the Phillips Curve. The Phillips Curve model only works when the business cycle is driven by demand shocks. When we are hit by supply shocks the Phillips Curve actually slopes upward; inflation is higher during recessions.

It's the Fed's job to prevent demand shocks, which means it's the Fed's job to make the Phillips Curve model false. So why are they relying on that model to predict an upswing in inflation as unemployment falls below 5%?

Here's an easier way to think about it. Imagine a single mandate, where the Fed keeps inflation right at 2%, all of the time. In that case the Phillips Curve is horizontal; there is zero correlation between inflation and unemployment. Now assume Congress adds a second mandate---employment. The Fed would do a bit more expansionary policy when unemployment was high, and vice versa. They would still keep inflation stable at 2%, on average, but allow some year to year fluctuation in inflation to smooth out unemployment. A bit more than 2% inflation when unemployment is high, and a bit less than 2% inflation when unemployment is low. In that case inflation will be countercyclical, exactly the opposite of the prediction of the Phillips Curve model.

So why is the Fed using the Phillips Curve to forecast inflation?

And why is it that when I mention this simple point, right out of EC101, so many economists give me a "funny look"?

PS. I do understand that the Fed can't entirely prevent demand shocks, and hence occasional periods of procyclical inflation. But surely they should set policy in such a way that they forecast an outcome that is consistent with their policy goals.

Comments and Sharing

COMMENTS (10 to date)
Steve writes:


What's your take on using immigration to help increase wages?

Currently, the US allows one million to immigrate every year. If they set the number of immigrants to accept contingent upon wage inflation and the unemployment rate, that would probably increase wages of workers in the US and also encourage others to enter the workforce. Why should the US accept one million immigrants/year regardless of the macroeconomic situation?

E. Harding writes:


-Sounds like a pretty boneheaded idea. The Fed has a lot more influence over money than U.S. immigration services have over movement of people.

Michael Byrnes writes:

Is the Phillips curve, then, a sort of measure of the effectiveness of monetary policy? As in, if the curve slopes upward, the Fed is doing its job; if downward, then monetary policy is procyclical?

In other news, apparently Mr. B was fully assimilated by the FedBorg.

ThomasH writes:


Immigration will affect real wages of those whose labor is highly complimentary or substitutable with sub groups of domestic workers, but the overall effect on average nominal wages will probably not be great. Immigration is both a supply and a demand shock, so the effect on monetary instrument use ought to be is also dubious.

@ Scott,

You seem to be assuming that the Fed is politically unconstrained in its use of instruments. Clearly policy 2008-2015 cannot be understood as carrying out the dual mandate with some below - target inflation needed to restrain excessive employment. Rather, I'd argue, the Fed faced numerous constraints: an inflation rate ceiling, a no-negative interest rate constraint, and a limit on the amount of QE it could carry out even thought NGDP and the price level were below target.

Joseph Calhoun writes:

Why does the Fed use the Phillips curve? Because Yellen is a labor economist? Because Yellen is a bad labor economist? Scott, just say what you're thinking....

Jon writes:

Scott, something is bothering me with your claim. I'm struggling to write it down in math but even if the fed targets inflation, the residuals should still show the Philips curve relationship.

So if unemployment leads inflation, we can use unemployment to observe the error in inflation vs the IT, and if we go to the next order model and look at the rate of change in unemployment, we can put that into our reaction function too.

Majromax writes:


The Philips Curve will only show up in the relationship if the Fed persistently performs less monetary intervention than it should.

Imagine the Fed controls in a proportional-derivative way, where they look both at the current inflation rate and the rate of change of inflation/unemployment. In that case, the residual you mention based on first differences should also disappear.

You can perhaps recover it by going one-derivative-higher than the Fed does, but by this point you'll be swamped by random noise, to the point where there's not likely a useful signal left in the face of supply-side movements.

Jose Romeu Robazzi writes:

@Prof. Sumner
I was puzzled by this post. Is the Fed capable of "controlling" real variables like unemployment? What if unemployment rises, and the fed eases, but unemployment refuses to budge? One probably will just get more inflation ... What am I missing here?

Scott Sumner writes:

Steve, I'd like to see a dramatic rise in immigration, to at least 3 million/year, regardless of where we are in the business cycle.

One side benefit is that it would boost trend growth, keeping the US away from the zero interest rate bound (see Australia.)

Michael, Yes.

Thomas, I don't agree that the Fed was constrained in prior years. They had all the tools they needed, they simply refused to utilize them in the right way. We could have had much faster NGDP growth, with much less QE.

Joseph, This isn't about Yellen, most economists seem to agree with her.

Jon, I don't follow. If they perfectly target inflation, there is no downward sloping PC. If there are random errors, it depends whether those random errors reflect supply or demand shocks---the slope is ambiguous.

Jose, They can't control it in the long run. But in the short run monetary policy does have an effect on employment.

Jeff writes:

Scott, in what instance does the Fed use the PC to forecast inflation? When I look at the "dot plots" I see them all converge to 2.0% over time, which seems consistent with their policy goal.

Comments for this entry have been closed
Return to top