The history of economic thought is not a food fight: Phillips Curve edition

One thing I hate and tried to teach my students to avoid in the History of Economic Thought class is the tendency to view intellectual history as a food fight - some kind of extended battle between the good guys and the bad guys. A great example of this is the Keynesv. Hayek rap and the whole idea that this is a fight stretching over the centuries. Of course there are disagreements - and certainly there was a disagreement between Keynes and Hayek - but nothing like these epic battles which dumb down the actual scientific discussion. This weekend I got into an argument with Phil Magness, a program director at the Institute for Humane Studies, about the Phillips Curve. He sees the history of the Phillips Curve very much in these terms: a Keynesian vs. non-Keynesian fight where Keynesians opportunistically used relationships in the data to push a policy preference. Friedman, Phelps, and Lucas came in to save the day and destroy the Phillips Curve, and nobody makes use of the Phillips Curve now except for "peripheral Keynesians" (his words).

Two notes: (1.) This will be long, but I’m going to divide it into sections to help focus on the main points. So if you’re not going to bother reading it all, please at least skim the section headings. (2.) None of this is hidden knowledge or original digging on my part – in fact I think it’s fairly widely known among people that care about this stuff. A lot of this is pulled from  Leeson and Young's (2008) "Mythical Expectations", Robert Gordon's (2011) "The History of the Phillips Curve: Consensus and Bifurcation", several articles by James Forder, and some from the primary sources. I'm not going to cite them formally below because this is not a formal write-up.

This is written sort of on the fly. There are a lot of subtle differences between these perspectives (prices vs. wages, direction of causality, reasons for LR/SR differences, etc.) so if some of it is a bit off don’t get too upset with me and let me know so I can adjust.

+++++++++++++++++++++++++++++++++++++++++++++++

"No one supposes that a good induction can be arrived at merely by counting cases. The business of strengthening the argument chiefly consists in determining whether the alleged association is stable, when accompanying conditions are varied"

- John Maynard Keynes, 1921

+++++++++++++++++++++++++++++++++++++++++++++++

1. Expectations augmentation did not start with Friedman and Phelps.

 Expectations have been important to economists for a long time. Malthus, Ricardo, and Bastiat all extensively discussed how taking expectations about the future into account often changes results in static models or models that don't include expectations. When it comes to expectations in macroeconomics, this was not new in the 1960s either.

The two really obvious cases are Keynes (1936) and Hayek (1937). Expectations play an enormous role in the General Theory in determining policy effectiveness. Keynes's interest in expectations and belief formation generally go back to his Treatise on Probability, some common interests with Frank Ramsey, etc. This is all very well known, and it was at the time. Hicks said in his review of the General Theory that 'From the standpoint of pure theory, the use of the method of expectations is perhaps the most revolutionary thing about this book' (see Forder, "The Historical Place of the Friedman-Phelps Expectations Critique"). For Keynes expectations fed primarily into entrepreneur's investment decisions and the liquidity preference function. They played less of a role in his analysis of consumption, at least in the General Theory (he seems to get at some of these ideas as they relate to consumption in How to Pay for the War, but I have yet to look closely at that). In any case, as far as the science is concerned this would come in later with Modigliani and Friedman, etc.. Hayek initially did not make expectations as central as Keynes as far as I can tell, but he made up for that in his 1937 paper "Economics and Knowledge" which laid a lot of the groundwork for how to think about defining a dynamic equilibrium in terms of expectations. 

Much of this was not formalized until later - a point I'll come back to. In the early formalizations of all of these ideas in the 1930s and before of course you start out simple, perhaps not formalizing the more complicated ideas, and then build up. But that is very different from saying that nobody understood or thought about the role of expectations.

This is getting far afield of the Phillips Curve (of course there was no Phillips Curve per se at this point). However, even at the very beginning of work on the Phillips Curve people understood the importance of expectations and appreciated each other’s insights. Most notably, Leeson and Young (2008) report that Friedman actually got the equation for adaptive expectations from Phillips in 1952. He was fascinated by Phillips’s work (at this point, principally the machine although the Phillips Curve grew out of that) but understood that the implicit assumption of the model was an expectation of stability. Friedman asked Phillips how he would model expectations that were potentially unstable, and he produced the adaptive expectations equation that Cagan would make so famous in his analysis of hyperinflation (which, if you think about it, is just Friedman/Phelps with half a Phillips Curve!). The Cagan analysis would of course be adapted by the rational expectations revolution as well. So Friedman certainly was not under the impression that the early thinking on the Phillips Curve was done in ignorance of expectation augmentation. Interestingly enough, Friedman twice offered Phillips a job at the University of Chicago (which he twice turned down).

So even in these early years expectations were an important part of economics, the germ of later developments came from Phillips himself, and there was no sense that the scientific questions these men were probing were “peripheral”. What about Samuelson and Solow? 

2. The difference between Friedman/Phelps and Samuelson/Solow was a difference of (a.) formalism, and (b.) the natural rate hypothesis. It was not a disagreement about whether the Phillips Curve existed or whether it was stable. Everyone agreed that it did exist and it was not necessarily stable.

I really don’t need to overcomplicate this: If you think Samuelson and Solow (1960) said that the Phillips Curve offers a stable menu of trade-offs between inflation and unemployment there is an extraordinarily high probability that you simply have not read Samuelson and Solow (1960). Not reading them is OK in my opinion. I am not one of those people that think every single person should take history of economic thought. But by the same token if you’re going to make a claim about the article you should probably... I dunno… read the article?

Samuelson and Solow (1960) are actually principally concerned not with a menu of policy options (though that comes up) so much as with the fight going on at the time between cost-push, demand-pull and other lesser varieties of explanations of inflation. But they do come back to how to think about the Phillips Curve on page 193 where they (very famously) write:

“Aside from the usual warning that these are simply our best guesses we must give another caution. All of our discussion has been phrased in short-run terms, dealing with what might happen in the next few years. It would be wrong, though, to think that our Figure 2 menu that relates obtainable price and unemployment behavior will maintain its same shape in the longer run. What we do in a policy way during the next few years might cause it to shift in a definite way.

Thus, it is conceivable that after they had produced a low-pressure economy, the believers in demand-pull might be disappointed in the short run; i.e., prices might continue to rise even though unemployment was considerable. Nevertheless, it might be that the low-pressure demand would so act upon wage and other expectations as to shift the curve downward in the longer run-so that over a decade, the economy might enjoy higher employment with price stability than our present-day estimate would indicate.

But also the opposite is conceivable. A low-pressure economy might build up within itself over the years larger and larger amounts of structural unemployment (the reverse of what happened from 1941 to 1953 as a result of strong war and postwar demands). The result would be an upward shift of our menu of choice, with more and more unemployment being needed just to keep prices stable.

Since we have no conclusive or suggestive evidence on these conflicting issues, we shall not attempt to give judgment on them. Instead we venture the reminder that, in the years just ahead, the level of attained growth will be highly correlated with the degree of full employment and high-capacity output.”

Let’s take a tally of what is here and what isn’t here to better understand what was so important about Friedman and Phelps. First, Samuelson and Solow definitely don’t think the Phillip’s Curve offers a stable menu of policy options. They definitely recognize that it is a short-run trade-off. They also definitely recognize that policy choices impact the long run state of the Phillips Curve and they definitely recognize that expectations determine the long run state of the Phillips Curve.

What don’t Samuelson and Solow offer us? Well they don’t have a clear vertical long run Phillips Curve (i.e., they don’t have a natural rate). They have shifting curves in mind, which actually is what the data do end up looking like. If you read the rest of the article you’ll know that they also don’t have a model or any formal presentation of expectations. Part of the reason for this, of course, is that the whole point of their article is that the jury is still out on the theoretical underpinning of the Phillips Curve. We get both of these things from Phelps and Friedman, who set the ball rolling for modern macroeconomics: rational expectations, the Lucas critique, New Classical macro, New Keynesian macro, and what is sometimes called the “New Consensus” model.

The nature of adjustment depends of course on how expectations are formed, but the result of a NAIRU pops out as long as you assume that (1.) in the long run expectations have to conform to reality and (2.) full employment is determined by technical/exogenous factors. Phelps offers a more formalized picture than Friedman does and therefore has a more direct impact on later New Classical and New Keynesian Phillips Curves.

Some people, when discussing Samuelson and Solow on the stability of the Phillips Curve, like to point out that they thought that the curve shifted somewhat in the 1940s and 1950s or that differences in labor market institutions (principally unions) can explain some of the differences between the UK and the US. That’s all well and good but I think the passage above is what really drives the point home.

3. The contributions of Lucas were (a.) the introduction of much stronger assumptions about expectations and (b.) broader insights about the importance of using structural models.

Lucas comes at all this from a completely different angle because he’s interested in making a point about how we do modeling in macroeconomics. In the seminal Lucas island model, agents are assumed to be unaware of how much of the short run variation in their prices are due to general price level changes and how much is due to changes in the relative demand for their product. If they knew, they would not change their behavior in response to general price level changes but because they don’t know there is production (and therefore labor demand) response to price level changes: a Phillips Curve. The agents know the underlying probability distribution of all these components of the price and they have rational expectations, so in the long-run they can’t be fooled. There is, therefore, a slight difference in emphasis (though I wouldn’t say a fundamental difference) between Friedman’s argument and Lucas’s. In the island model the Phillips Curve comes from fooling people and you can’t fool people in the long run. In Friedman the Phillips Curve comes from more standard demand arguments and you can’t escape real factors in the long run. That’s a little stylized, but put in these terms it’s clear how Lucas is making much stronger assumptions about how agents interact with the world around them.

I’m sure Lucas was interested in inflation and unemployment, but the island model is extremely unrealistic (and when you read the paper it’s clear he knows that). So his real point, I think, isn’t to offer a convincing model of what’s going on so much as it is to point out that you can get the same reduced form relationship from a lot of different microfoundations and if you don’t know what microfoundations are true you can make policy decisions that can come back to bite you in the long run. This was a bit under the surface in his 1972 paper on the island model but it is front and center in his 1976 paper “Econometric Policy Evaluation: A Critique”. That paper is one of the most important in economics in the twentieth century. When I taught history of economic thought it was the only selection that I made my students (undergrads) read from for our single lesson on post-war short run macroeconomics (I had another lesson on growth theory). With Lucas I think you really get the whole path of post-war short run macro, from the Phillips Curve discussion through microfoundations, rational expectations, the rise of New Classical macroeconomics, and the structure of New Keynesian macroeconomics when it emerged. All of this pivots on Lucas.

But what didn’t Lucas say? Lucas definitely didn’t say there was no Phillips Curve or that it was “peripheral” (Magness’s words). It was quite real and like basically everyone before him he said that the short run and the long run versions were not the same thing. Friedman and Phelps brought formal expectations to the table and a NAIRU, while Lucas brought broader points about microfoundations and rational expectations assumptions to the table.

4. The Phillips Curve is extremely important. Essentially everyone uses it; it is not peripheral. There are active areas of research and disagreement over the Phillips Curve.

As with my Samuelson-Solow discussion, I’m not going to overcomplicate this: if you think the Phillips Curve is unimportant or peripheral to modern economics you’re simply wrong. Donald Kohn said that “A model in the Phillips Curve tradition remains at the core of how most academic researchers and policymakers – including me – think about fluctuations in inflation”. Although it’s important to remember there are a few steps to get from one to the other, the Phillips Curve is essentially just an aggregate supply curve, and the business cycle doesn’t really make sense without the aggregate supply curve. However, the fact that that’s settled doesn’t mean that there’s nothing interesting happening in the literature. I’m no expert in this area, but I think there are at least two important discussions going on.

First, it’s not entirely clear that the long run Phillips Curve is vertical. There’s good empirical evidence indicating that the long run Phillips Curve might be backward bending or otherwise downward sloping at low inflation levels. There are a variety of reasons offered for why this might be the case that are typically related to wage bargaining and rigidity or cognitive limitations around very low inflation levels (inflation is most costly to estimate and account for when it is high). This literature is generally associated with Palley in the Post-Keynesian world and Akerlof, Dickens, and Perry in the mainstream literature. Of course it has been of great interest lately with better anchored inflation expectations and subdued inflation during the Great Recession. This one is potentially huge because it does actually claim a (limited) long run trade-off.

The other, older discussion concerns “hysteresis” and whether the NAIRU is really stable. If the NAIRU is a function of past unemployment it could move around. Notice this is not just an observation that supply shocks could occur (which arguably is just a change in exogenous/technical factors and thus consistent with the original NAIRU idea). These are demand-side factors with a long-run impact on the NAIRU (albeit potentially through supply-side channels like skill degradation). This also has Post-Keynesian counterparts, and it is also of interest lately given the experience of persistent high unemployment.

None of this is “peripheral” stuff, I should add. Some of the biggest names in the field have wrestled with both of these questions, particularly the hysteresis issue.