Saturday, January 19, 2013

Phillips curve? What curve?

The Phillips curve is one of those 'regularities' that is more likely to exist in an economist mind than in reality. Figure below (from Ritschl and Straumann, 2010*) shows the 'trade-off' between inflation and unemployment in Europe in the 1920s and 1930.
They say quite candidly:
"the seemingly obvious connection between deflation and unemployment is less than easy to find in the data. ... Yet while there was ample variation in both unemployment and inflation during the interwar period, no systematic pattern seems to emerge in the data, even if the 1920s and the 1930s are looked at separately. The picture emerging from Figure [above] is rather that the natural rate of unemployment moved quite independently of inflation, irrespective of whether or not a country was on gold" (italics added).
Seemingly obvious, but inexistent in the data. In doubt economists choose the obvious and avoid reality. The other possibility that they do not seem to consider is that there is no natural rate at all, and no systematic relation between inflation and unemployment. Consider the data for the United States below (CPI change and unemployment data from Fred).

Note that there is also no clear correlation, unless you start assuming that it is unstable, and the curve is jumping, or that a supposed natural rate is jumping around too. Occam's Razor would suggest that if you have a theory that does not postulate any systematic relation between prices and quantities** it should be simpler and, hence, preferable.

* Ritschl, Albrecht and Straumann, Tobias (2010), "Business cycles and economic policy, 1914-1945." In Broadberry, Stephen and O'Rourke, Kevin H. , (eds.) Cambridge economic history of modern Europe. Cambridge University Press: Cambridge, UK , pp. 156-180.

** Note that classical-Keynesian theory implies that prices are determined by the technical conditions of production and exogenously determined distribution while unemployment derives from effective demand. Also, unemployment might have an indirect effect on inflation, through the effects of unemployment on the bargaining power of workers and, thus, on income distribution.


  1. // oh do i have a lot to say... i hope i can remember the half of it.

    Occam's Razor would suggest that if you have a theory that does not postulate any systematic relation between prices and quantities it should be simpler and, hence, preferable."

    By that logic, the Empty Theory is simplest and most preferable of all!

    Not perfectly relevant I suppose, but on some other topic Schumpeter said The distinction is, in a sense, quite unrealistic. But if we do not make it, we shall never be able to say any more than that everything depends upon everything.
    I need a theory that is less vague and not so preferable as the one you suggest.

    I think it is throwing out the baby with the bathwater to reject the Phillips curve with the "natural rate" idea. Phillips found empirical evidence over *many* years. I don't think his evidence should be just dismissed.
    My preference would be to try to undermine Friedman and Phelps by offering a different explanation for the inflation of the 1970s.

    Noah Smith had a great Phillips curve graph, showing the curve jumping, the second graph here:

    My thoughts on his graph included these:
    The placement of the curve is related to economic performance. Not "structural" shifts, but performance shifts. A healthy economy puts the tradeoff low and left on the graph; a sickly economy puts the tradeoff high and to the right.

    So the question, really, is not "Why does the Phillips curve shift?" The question is: What makes growth so good in some periods, and not so good in others?

    If you have a powerpoint viewer, David Andolfatto links to a good put-down of the curve here:

    Finally: the CBO relies on the Phillips curve when it figures Potential GDP:

    1. Good Lord! What is this stuff?

      Okay, let's go through a few things. First, why does the curve exist? Well, it is supposed to be a theoretical derivative of an empirical regularity. So far, so scientific, right? But what happens when that regularity clearly and substantially breaks down? Well, the scientific response is to accept that this is no longer a statistical regularity. The economic response is somewhat different: the economist clings dogmatically to the theory as if it made up some existential component of his being.

      Instead they say that the curve "shifts" or base their discussion on this. But if the curve "shifts" then it is no longer a curve. Just because economists put this in geometrical terms should not disguise the fact that they are engaged in sophistical argument. Here's how the argument is structured:

      1. Posit a curve that shows two variables which have a firm relationship to one another -- in our case unemployment and inflation.

      2. Discover that this is not a true representation of the data.

      3. Claim that the curve remain intact but then "shifts". This is done purely to hold the flawed theory together but at the same time admitting that its basically wrong.

      4. Feel comfortable that you have done your job and defended your theory.

      And I know the response I'll get. The interlocutor will have a mild existential crisis and claim that if we don't have the Phillips curve then Everything Falls Apart and life becomes meaningless. Well, no. Some of us just make good learning the different types of inflation and examining historically when they arise so that we are better able to detect them in future. If such a position requires one to exist on the edge of existential limbo, floating precariously between Truth and Nihilism well then, I don't know what to say to that.

    2. TheIllusionist: "Good Lord! What is this stuff?"

      I know :) ... Matias, if I get overbearing or just plain stupid, please feel free to delete my comments. You'll be doing me a favor.

      Illusionist, I explore the internet seeking the interesting and challenging. Thanks for providing both.

      First, why does the curve exist? Well, it is supposed to be a theoretical derivative of an empirical regularity. So far, so scientific, right? But what happens when that regularity clearly and substantially breaks down? Well, the scientific response is to accept that this is no longer a statistical regularity.

      When the science is settled, yes. But science never settles.

      For me, the shifting of the Phillips curve is evidence of something that awaits discovery. What I think is that the Phillips curve shifts in response to changes in monetary balance -- specifically, to changes in the "debt per dollar" ratio, the ratio of total debt per dollar of circulating money -- but I don't know how to make calculations to show it.

      The empirical regularity broke down because of bad policy. You seem to assume that policy could not have caused the breakdown. I think it is the most likely cause. (Occam!)


    3. I don't think you get my point here. There either is a curve or there is not a curve. There is a curve if we can be absolutely sure that this statistical regularity is a constant no matter what.

      In economics the only things we can talk about in such terms with any degree of certainty are what we call identities. So, we can say -- apriori and with absolute confidence -- some of the following things:

      => MV = PT
      => GDP = G + I + C + (X - M)
      => I = S
      => (G – T) = (S – I) – (X – M)

      And so on, and so on. These are apriori truths by construction. If the empirical reality doesn't fit, someone has mucked up the accounting.

      The Phillips curve is an altogether different entity. It was derived from a statistical regularity. Thus it was, despite the fetish-like geometrical form it was given by Solow and Samuelson, just a rule of thumb. Not a law. Not an identity. Just a rule of thumb. It works until it doesn't. And when it doesn't work we drop it. Otherwise we're just squeezing the facts and torturing the argument so that it will fit with what we read in a textbook.

      Personally, I don't think that the Phillips curve should have ever even entered the debate in macro beyond the form that Phillips gave it: namely, an empirical correlation between wages and unemployment that could be interpreted to POSSIBLY explain SOME TYPES of inflation (i.e. wage inflation). The second that Samuelson and Solow turned it into an abstraction it was obvious that it would become a fetish. That it was the very fetish that, in the 1970s, turned back on its creators like Frankenstein's monster and destroyed them is, I think, in the last instance rather amusing.

    4. Don't call it a "curve". Call it a "relation" between inflation and something. The relation may still exist even after bad policy shifts the "curve" away from the origin. Probably still exists, in fact.

    5. But look at what you're actually saying here. You're saying that the relation "may still exist" even though it doesn't. And then you justify this by the fact that "bad policy" has led to the relation not existing (while it still, in another undefined but possibly metaphysical sense, does).

      So, now this becomes a policy issue. You think that the relation SHOULD exist -- this is a normative judgment -- and that it only doesn't exist because policy is not run correctly. Again, let's break down the argument so that people can clearly see the point at which it becomes either nonsensical or normative.

      1. The Phillips curve exists because some data in the past showed a correlation between inflation and unemployment.

      2. Fresh data shows that this relationship doesn't exist anymore.

      3. This does not mean that the curve no longer exists. Sure, we can't actually SEE it anymore. But it continues to exist on another plain of existence.

      4. The only reason we cannot see it is because governments have enacted "bad" policies which are bad because they have made the Phillips curve disappear (well, sort of, as it still exists -- even though it doesn't exist).

      5. So, either the Phillips curve is now a normative rule which we should run policy based on (i.e. policy should be run in such a way that the curve "appears" in the data) or we're just asserting that the curve "probably" exists on some other plain of reality even though it doesn't show up in the data. So, we're either dealing with a strongly normative argument or a metaphysical argument.

    6. Sheesh, I barely have time for coffee!

      "So, now this becomes a policy issue. You think that the relation SHOULD exist -- this is a normative judgment -- and that it only doesn't exist because policy is not run correctly."

      To be clear, I am not saying the policy is bad because it ruined the Phillips curve. I have unrelated reasons for saying the policy is bad. I think the changes in the Phillips curve are evidence that can help us understand what happened.

      And you say "this relationship doesn't exist anymore" but I think it DOES still exist. Only it exists at higher levels of inflation and unemployment that it did when the economy was good.

      I need to go make some graphs to be sure. Thanks.

    7. Good Lord, making graphs won't make you sure of anything that you don't build into your graphs. What is up with the economics profession these days? It's become a complete pack of metaphysicians!

      Again, to be clear, because (and I find this quite shocking) you don't seem to understand the difference between something existing EMPIRICALLY and something existing as a thought experiment UNDER CERTAIN ASSUMPTIONS.

      The Phillips curve, as posited by Phillips himself, was a statistical regularity. It ceases to exist the moment it... ceases to exist. It really is that simple.

      When you draw a bunch of graphs showing that, under certain assumptions, inflation will rise if unemployment falls you're just telling an imaginary story. We could alter the assumptions somewhat and tell an entirely different story if we wanted to.

      You seem to equate what Phillips was doing -- that is, looking at an empirical relationship in a given period -- with what you're doing when you tell imaginary stories with graphs. They're not the same thing. Why don't economists understand this very simple point? Why are they unable to distinguish between what's in their heads and what exists in reality? Well, I have a few theories as to why that is, but I won't go into them here...

    8. Regarding the curve shifting around over time -

      I think this is OK. There can still be value in noticing the relationship between unemployment and inflation. It just means that there is an incomplete understanding of what is going on. It just means that more variables go into determining unemployment than just inflation.

      Obviously, there are more variables that go into determining unemployment rate. Population, interest rates, tariffs, technology, ...

      It doesn't mean that there isn't a relationship between inflation and unemployment. Just that there is more going on than that.

      Imagine that you're in a car, and I'm driving. You can't really see what is going on, but you have an accelerometer and you have some kind of device for measuring the position of my right foot. You're trying to figure out the relationship between the position of my right foot and acceleration, if there is one.

      You find that it seems to be the case that when i drop my foot down, there is more acceleration. You can make some plots of foot-height vs acceleration.

      But instead of seeing one curve, you see a few. At any given time, there is a clear relationship between the position of the petal/foot and acceleration ... but the curve moves around over time. It doesn't mean that there is no relationship between foot-position and acceleration ... it just means that there are other things going on too. (I'm changing gears and velocity.)

      This is what science is. You put together pieces of the puzzle. You don't start out with whole picture.

    9. I completely disagree. This is absolutely not what science is. Scientific laws -- which the Phillips curve and other neoclassical constructions implicitly pretend to be, right up to the sophistical geometrical representation -- scientific laws fall if they are falsified in a consistent manner. The Phillips curve, which is generally thought to be THE explanation of inflation by pseudo-Keynesians, has been completely and utterly falsified. And yet it is still posited.

      Okay, now someone is going to say "Oh, but we never claimed it was a scientific law...". Fine, then what is it? I said that Phillips' original presentation was of a simple, empirical statistical regularity. But since the more recent data has shown this regularity to have broken down, then it can no longer be said to be this. So, what is it?

      I've made clear what I think it is: a one-time statistical regularity that no longer holds, but one which we can theorise and which may or may not occur again. But the neoclassicals et al are NOT clear about what it is. In formal terms it is presented as would be a scientific law. Indeed, they seem to teach and preach it as if it were this. They seem to use it in their models as if it were this. So, what is it? What is it that gives it this special privilege among theories of inflation?

      I'll tell you my opinion: its a fetish; an idol. It's like a fertility statue in a tribe that people attribute mystical powers to. It's a simplistic device that people use to structure their (false) understanding of the world. It is, in its presentation and its form, a superstition.

    10. It's called curve fitting and data mining in investment.

      Generally what you do is come up with some statistical correlation between some values, write a piece on it in some journal, get a following and then flog a book on the subject to some mug punters.

      Then you retire on the loot from the book while the mug punters lose their shirts in the market. You can always explain it away by suggesting that market action filled in the anomaly.

      But fundamentally its the same brain pattern function that sees teddy bears in clouds.

    11. @Neil Wilson

      Haha! Someone needs to do a graph plotting the likelihood that neoclassicals fall for 419 scams!

  2. Oh poor William Phillips. What you lay out above is not a Phillips curve at all, but a Samuelson/Solow curve.A Phillips curve ties inflation to wages:

    It is thus a representative of Keynesian "wage inflation". While by no means perfect this is far superior to the silliness of the neoclassicals who, in their theological quest for easy and bulletproof apriori truths, find only tedious nonsense and garbage.

  3. Many comments, not enough time to respond to all. A few brief comments. Yes empty theory indeed. If there is no systematic relations between excess demand (measured by unemployment below a natural rate, or an output gap with a potential GDP at full employment) and inflation you will indeed be better with no theory. A non event requires no theory, by definition. What's your theory for things falling upwards? Yes you can drwa lots of graphs with jumping curves and other epicycles, the questions is that there is no evidence or theoretical reason to believe that there is a relation and it is highly unstable. The US was close to full employment (say below 4%, an arbitrary, but fairly small number) only four times (WWII, Korean and Vietnam Wars and boom). Inflation acceleration (like the 1970s) is more often than not associated to supply shocks (oil) and wage resistance, and not excess demand.

    On the use of wage inflation rather than prices, as in the original Phillips curve, I don't think it would make much of a difference, or that Phillips himself would have been against. The person to consult on Phillips is always Robert Leeson, the editor of his collected papers.


Lula's election and what lies ahead

It's been a while since I wrote about the Brazilian crisis (a summary of the previous catastrophic election here ). In part that electio...