Monday, February 9, 2015

Pining for the Fjords

This post by Simon Wren-Lewis brings the dead parrot sketch to mind (with Simon playing the Michael Palin role). I'm interested in the last couple of paragraphs of his post. Simon is analyzing the U.K. policy problem, and asserts:
For whatever reason (resistance to nominal wage cuts being the most obvious), inflation ceases to be a good indicator of underutilised resources when inflation starts off low and we have a major negative demand shock.
First, for inflation to "cease" to be a good indicator of underutilized resources, it must have once been a good one. I have run into economists who think that the unemployment rate is a good measure of underutilized resources. Then, if inflation is a good "indicator" of underutilizaiton, there must be a stable Phillips curve - a negative relationship between the unemployment rate and the rate of inflation.

Clearly, Simon thinks that there has been a "major demand shock" - presumably he means the financial crisis - which caused the Phillips curve to break down. So, suppose we look at the data on CPI inflation and the unemployment rate in the U.K. for 2000-2006. That's arbitrary of course, but if this Phillips curve is so stable, we should see it in the data for that period. Here's the time series (quarterly data):
The inflation rate rises steadily over that period; the unemployment rate goes down, and then up again. Here's the scatter plot, with the observations connected in chronological order:
You can see that's not much of a Phillips curve. Suppose I knew that the inflation rate was 1.8%, and I could not observe the unemployment rate. What would the observed inflation rate tell me about the unemployment rate, given the last chart? Not much. And, in fact, I can observe the unemployment rate. So what is the inflation rate telling me about underutilization?

So after the "major demand shock" - otherwise known as the financial crisis - occurred, what does the Phillips curve in the U.K. look like? We'll first plot the data for 2007-2014 (quarterly) using the headline CPI to measure the inflation rate. Here's the time series:
I think you can tell that this isn't going to produce a nice Phillips curve. Here's the scatter plot:
In particular, note that, from peak unemployment during the recession, the unemployment rate drops about 2 1/2 points, while the inflation rate drops about 3 points. You can see why Simon is worried about the inflation rate as an indicator of underutilizaiton. Presumably utilization has been rising in the U.K., but inflation is dropping like a rock.

In case you're wondering what happens if we use core inflation instead of headline inflation, here's what that yields:
So, that's even worse - the Phillips curve has the wrong slope.

Simon goes on: seems quite possible that GDP continues to be quite a few percentage points below where it could be without inflation exceeding its target...
So, in spite of the fact that Phillips curve logic is inconsistent with the data - and the problem seems more severe post-recession, Simon continues to use that logic. He imagines that it is underutilization that is holding back the inflation rate, as he states: we continue to waste resources on a huge scale. This is money down the drain that we will never get back. It is like taxing households thousands of pounds or dollars or euros a year and burning that money.
Thus, the possibility of an output gap becomes a certainty - it's now a waste of resources on a huge scale.

Simon finishes the post with:
...if, at low inflation rates, inflation becomes a noisy, weak and asymmetric indicator of the output gap, then focusing on inflation is going to perform badly. In these circumstances it could be many years before it becomes clear that we have been continually running the economy under capacity, and needlessly wasting resources. Unfortunately even when that point of realisation arrives, for obvious reasons monetary policymakers are going to be reluctant to acknowledge the mistake.
Simon's chief worry is the latent output gap - the inefficiency he suspects is there, but seems at a loss to measure. It's unclear why we can't see it now, but might realize sometime in the future that it was there all the time - massive, and highly persistent. That certainly doesn't seem like a Keynesian inefficiency - or, to be more accurate, a New Keynesian inefficiency - as that's a temporary phenomenon. If Simon is so certain this beast exists, he should be able to tell us what it is.

I think it might help Simon, as a start, if he declares the parrot dead. The Phillips curve is not resting, sleeping, or pining for the fjords. It is dead, deceased, passed away. It has bought the farm. Rest in peace.

The Bank of England has been close to the zero lower bound for a long time. The Bank Rate has been set at 0.5% since March 2009. Here's the latest inflation projection from the Bank:
And here's the latest inflation data, up to December 2014:
So, like Simon, the Bank seems not to have learned that the parrot is dead. In spite of a long period in which inflation is falling while the economy is recovering, they're projecting that inflation will come back to the 2% target. But 20 years of zero-lower-bound experience in Japan and recent experience around the world tell us that sticking at the zero lower bound does not eventually produce more inflation - it just produces low inflation.


  1. Great post! We need more people questioning and providing an alternative to the "Intermediate Macro Keynesian" narrative out there. Hopefully you'll blog more about your own "positive" theory of inflation, aggregate level of activity, etc.soon?

  2. You forgot the best one -- the Phillips Curve has joined the bleeding choir invisible. It is an ex Pcurve!

  3. Does the phillips curve assume that the only thing affecting inflation is the labor market?

    1. There are at least two theories of the Phillips curve correlation. The first is a very old one, which was in Lucas's "Expectations and the Neutrality of Money" paper from 1972. Lucas's theory was that the Phillips curve correlation was due to unanticipated monetary shocks. The central bank surprises economic agents with a monetary shock, and people are fooled into working harder. The theory can potentially explain why the Phillips curve shifts, or its slope changes. But that theory isn't taken so seriously any more. The Phillips curve also comes out of New Keynesian models, due to price stickiness, and the forward-looking behavior of firms who are setting the prices. I think that also implies that the Phillips curve is unstable, but no one really explores that. In its crudest form, the Phillips curve becomes a theory of inflation - output gaps, however measured (could be the difference between the unemployment rate and some "natural rate") determine the inflation rate. Some central bankers think in the latter terms.

  4. The post "Government debt, financial markets and dead parrots" from Simon reminds me more of the dead parrot sketch.

  5. from Paul Ormerod's blog:
    A powerful blow against the concept of potential output has been published in the latest edition of the American Economic Association’s journal Applied Economics. Igal Hendel and Yossi Speigel document the evolution of productivity over a 12 year period in a steel mini-mill producing an unchanged product, working 24/7. The steel melt shop is almost the Platonic ideal from a national accounts perspective of output measurement. The product – steel billets – is a simple, homogenous, internationally traded product. There was virtually no turnover in the labour force, very little new investment, and the mill worked every hour of the year. Yet despite production conditions which were almost unchanged, output doubled over the 12 year period. As the authors note, rather drily, “the findings suggest that capacity is not well defined, even in batch-oriented manufacturing”. Time to put the concept of potential output into the rubbish bin

    1. At the aggregate level, the way this concept is used is typically not helpful. What we really care about are the welfare effects of alternative policy rules. To do that calculation, we need a model. Inefficiency (relative to what is feasible under alternative policy rules) is not something we can measure directly.

  6. "So after the "major demand shock" - otherwise known as the financial crisis". Would there have been a demand shock if central banks had responded appropriately to the suddenly increased demand to hold money? The experience of Israel and Australia suggests not.

    1. The point is that the financial crisis is not a "demand shock." That's language attached to some device that was developed to teach something to undergraduates. It can't help us think about what was going on in the financial crisis.

    2. Doesn't Angeletos/Collard/Dellas have a model that delivers a "demand shock" (MIT 2014). That at least is a proper model, not an undergraduate toy.

    3. They can label what is going on in their model any way they want. They also call it "confidence." I haven't read the paper, so I don't have much to say about it, but I'm certain this is far from what Wren-Lewis is thinking about, or what your typical Old Keynesian who says "demand shock" is thinking about. Did "higher-order uncertainty" have something to do with the financial crisis? I don't know.

  7. Loved the Monty Python reference. I should add, that at least in case of Australia, different expectations about central bank policy discouraged such a flight to safety in the first place.

    "But 20 years of zero-lower-bound experience in Japan and recent experience around the world tell us that sticking at the zero lower bound does not eventually produce more inflation - it just produces low inflation."

    Persistent deflation under the C19th gold standard was a result of the stock of gold rising slower than the output of goods and services, just as persistent inflation was a result of the stock of gold rising faster than the output of goods and services, in both cases the demand to hold gold mostly marching with output. With central banks either being (1) not big enough players to affect that much or (2) operating to make it so or (3) some combination of the two.

    By 1919, central banks completely dominated gold holding. Persistent deflation in post-return to gold standard Britain was due to the returned-to money price of gold not matching the British price level. Sharp deflation in the Great Depression was the result of a dramatic increase in the demand to hold gold (thank you, Bank of France).

    So, in none of the above cases do we have to tell an interest rate story. Why do we now tell an interest rate story? Interest rates being the price of credit, and all: an across-time price.

  8. "So, that's even worse - the Phillips curve has the wrong slope." - no, you can clearly see that the percent change vs percent curve has one correlation then the opposite, depending on what 'regime' the data are in, such as 'boom times' or 'bust (post-boom) times'. The analogy with physics: the same fluid will behave differently depending on whether the regime is "laminar" or "turbulent", with the cutoff between the two regimes being quite dramatic (analogous to the charts here), with the cutoff determined by the fluid's "Reynold's number". It's surprising to me that you cannot see this. BTW these sorts of correlations are used in IQ tests all the time...the more you can spot, the higher your IQ.

    1. Some people can see the virgin Mary in a grilled cheese sandwich:

      Maybe people like that have high IQs, or have studied fluid dynamics (or cheese dynamics).

    2. Spotting correlations like these is similar to finding conspiracies -- they are there, of course, and anyone who can't see them is part of the conspiracy.