Primer: Understanding The Post-Keynesian Rejection Of Mainstream Inflation Theory

From the perspective of conventional economic analysis, the post-Keynesian approach to inflation is mystifying. If we focus on the Modern Monetary Theory (MMT) school of thought in particular, it is very easy to either find claims that "MMT has no theory of inflation," or non-MMTers "explain" the MMT inflation theory is some random trivial relationship that they just made up. The key to understanding post-Keynesian approaches is that it takes a completely different approach to understanding inflation, and outcomes are seen as very difficult to forecast.

This article is based on Section 8.1.1 ("The Rejection of the Acceleratoinist Thesis") of Professor Marc Lavoie's excellent Post-Keynesian Economics: New Foundations (link to my review). From the perspective of a non-academic, a significant portion of the book would likely be found as arcane, and could easily be confusing to a non-specialist. However, Section 8.1.1 is extremely straightforward, and the most difficult part of my writing task here is staying within "fair use" copyright limitations when describing it...

(Since I raised the issue of MMT earlier, I cannot say whether there are any major disagreements between MMTers and Lavoie's description of post-Keynesian thinking on inflation. My feeling is that there is nothing that a non-academic would get too excited about, other than the importance that MMT ascribes to the Job Guarantee wage in stabilising inflation. For the analysis of a country without a Job Guarantee -- currently, all of them -- this distinction has no practical import.)

Setting up a Mainstream Straw Man

One of the problems with post-Keynesian thinking is that it largely seems to be defined in a negative fashion: it appears to be just a random set of complaints about "mainstream" economics. Although I would prefer to avoid developing the discussion this way, it seems to be the best way forward in this circumstance. I have been writing and reading economic commentary produced in the financial markets as well as the internet for a very long time, and I would argue that there is a certain "conventional" world view that is widely shared by writers and readers, even if they do not qualify as "mainstream." I think it is best to confront these conventional views directly by setting up an overly-simplified version of the thinking -- a straw man -- and then explain what parts of the straw man view are viewed as completely out-of-paradigm. (Please note that my discussion in this section is not based on Lavoie's text; I return to his logic later. Another note is that this discussion applies to inflation theories, as a practical matter, even "mainstream" economists are closer to where I am on the matter.)

When I get around to writing a text on competing inflation theories, I will then run through the various variants, and give them an honest critique. For now, I just want to re-orient the discussion in a certain direction.

A fairly typical mainstream approach is to build models around transactions between agents with an endowment of goods/resources (including labour power) at a point in time; "dynamics" are allegedly introduced by introducing forward markets. Prices are (somehow) determined for this vector of goods/resources simultaneously, and then agents act (work/consume) based on maximising utility based on that vector of prices.

The next step is to lump all non-financial goods/resources into two buckets: labour, and an aggregate good. Therefore, there are two real economy prices that matter: the wage rate, and the aggregate price level (there are financial prices as well, such as the price of Treasury bills, as well as money, which is defined to have a price of 1).

We then engage in various arguments to arrive at the conclusion that wage rates are related to the marginal productivity of labour (in terms of the aggregate good), and so we only have one real economy price that matters: the money cost of the aggregate good.

If we step away from theory, economists decide one particular price index is "the" price level of the economy, and so it is equivalent to the money price of the aggregate good in economic theory. Inflation is the rate of change of this aggregate price.

(The exception to the conventional wisdom would appear to be Austrian economists, who have an aversion to aggregating the price level. Instead, they seem to define "inflation" as the change in the money stock, which may or may not be related to changes in the measured CPI. Since Austrian economics is now largely a financial markets and internet phenomenon, it's hard to give a clean theoretical summary.)

We can now define what inflation theory is: the theory that allows us to predict the changes in some price index, which I will refer to as the consumer price index (CPI) here for simplicity. Everyone scurries off in various directions at this point, but the usual result is that there is an attempt to slap together a reduced form model (that is, computationally tractable) that allows us to forecast the change in the price level, based on a few economic time series.

It should be underlined that practically everyone does this (including me, in my earlier life). For example, it is not hard to find physicists who opine on economics on the internet that build models that assume that there is a reduced form relationship between other economic variables and changes in the CPI; finding that reduced form model is just doing empirical work!

The post-Keynesian rejoinder is: not so fast.

Rejecting Reduced Form Models

It would be an exaggeration to say that post-Keynesians reject all reduced form models, rather they reject pretty well all the ones that the mainstream has proposed.

The first set of objections is anything based on the money supply. Marc Lavoie, on page 541:

In post-Keynesian theory, the level of the money supply does not determine the level of prices; nor does the rate of growth of the money supply determine the rate of inflation. Excess money or monetized government deficits are not a proximate cause of inflation, not even in the case of the German hyperinflation of the 1920s, as argued by Burkedin and Burkett (1992) and Wray (2012, pp. 246-57), nor in the case of the 'Great Inflation' in England in the 1500s, usually attributed to gold discoveries (Arestis and Howells, 2001-02).

We can then go after NAIRU (page 542):

... post-Keynesians reject the concept of the natural rate of interest, and they have very little faith in natural rate of unemployment or the NAIRU.

He quotes Wynne Godley (from the compilation Keynes and the Modern World, published in 1993, page 170):

Indeed if [emphasis in original] it is true that there is a unique NAIRU, that really is the end of discussion of macroeconomic policy. At present I happen not to believe it and there is no evidence of it. And I am prepared to express the value judgment* that moderately higher inflation rates are an acceptable price to pay for lower unemployment. But I do not accept that is a foregone conclusion that inflation will be higher if unemployment is lower. 

Conventional economists have an extremely wide range of variants of the notion of excess demand beyond the notion of the NAIRU, including concepts like the output gap (which can be operationally defined in many ways). From their perspective, all of these concepts are extremely different, although I am not very convinced that they are distinguishable. I am not going to put words into Lavoie's mouth and say that there is a blanket rejection of every such possibility, but I personally would argue that they are all fairly hopeless.

Within mainstream economics, there is an unorthodox school of thought known as the Fiscal Theory of the Price Level (FTPL - a primer). (The Lavoie text does not discuss the FTPL in Section 8.1.1.) As Professor John Cochrane argues, if mainstream economists actually believed the mathematics behind their Dynamic Stochastic General Equilibrium (DSGE) models, they would end up at the FTPL. The logic is sound. Unfortunately, the FTPL is an intellectual horror show: the price level is determined by the discounted value of all primary fiscal surpluses going out on an infinite horizon. The catch is that the only way to measure the value of discounted surpluses is to assume the FTPL is true, and back it out based on the observed price level. The theory is non-falsifiable, and so one could argue that it is correct (in some sense); it only has the minor drawback that we cannot make any predictions with it.

We are then left with the grab-bag of random reduced form models that analysts with access to a time series database and econometric software have churned out. Confronted that mess of internally inconsistent models is well beyond the scope of any individual. However, I would argue that it is safe to reject them based on a loose adaptation of the efficient market theory: if the models actually were reliable, we would hear about them.

What Works?

The message from Lavoie's text is rather scary from the perspective of anyone trying to build an inflation model (page 542):

What then is the basis of the post-Keynesian theory of inflation? Post-Keynesians argue that the 'economy is primarily a money-wage system' (Weintraub, 1978, p. 66). The money wage is the exogenous factor [emphasis mine] explaining the price level that the orthodox authors have sought. As Robinson (1962, p. 70) said. 'in our model, as in reality, the level of the money-wage rate obtaining at any particular moment is an historical accident.'

To anyone familiar with economist jargon, that is like a live hand grenade. I will work through the implications of that short passage for those who are less familiar with academic economist-speak.

The argument is that nominal wages are most important (a "money-wage system"). The catch is that it is an exogenous variable -- a variable that is determined outside the model. You cannot model the wage rate within a mathematical model, since it is an input to the model. 

So imagine that your boss tells you to come up with "an inflation model" for some country (which is a pretty common demand for employees of central banks or investment firms). According to post-Keynesian theory, the "correct" answer is to respond that inflation is an historical accident**. However, I must point out that the theoretically correct answer is also an extremely career-limiting one, so any employee stuck in that particular situation needs to figure out what their superiors want to see, and give them exactly that (even if the model stinks).

Future articles will delve further into the post-Keynesian description of how wages relate to the measured aggregate price level. However, the key takeaways are:

1. The relationship between wages and the CPI is complicated.
2. Wage inflation is dependent upon economic institutions and convention.

In other words, do not expect a cute little regression model (or whatever) between three or four economic time series and CPI inflation to work.

Aggregation Aggravation

I will conclude with another attack on reduced form inflation models, which is based on my own personal experience as an analyst than what I have seen in the post-Keynesian literature. I have no idea how what I am writing fits in with the existing academic theory. (To what extent it might fit in with existing theory, it might be more Austrian theory, which might cause some of my regular readers to faint.)

I see very little value in assuming that aggregate price levels have useful theoretical content. To be clear, aggregate price levels obviously exist, and can be measured. (If one rejects that they can be measured, one will have a very short career as an index-linked bond analyst.) Rather, my point is that since we are lumping together unrelated things, we should not expect the aggregate to follow any reduced form model in the first place. If we are going to apply reduced form models to inflation, we would need to apply them to the components of the CPI index separately.

Although such an assertion would give some academic economists the vapours, it would not be a surprise to inflation-linked market practitioners: that is exactly what they are already doing. Moreover, central bank analysts will do such decompositions when they are preparing detailed inflation forecasts, and such central bankers are mainstream almost by definition. The issue is that this approach coexists uneasily with mainstream theory.

We need to forecast CPI inflation component by component.

• What will cause the price of oil to move? If a cartel of oil producers decide to jack up the global price of oil, the domestic price of oil will move with it (translated to local currency).
• Food commodities are also flex-price markets, driven by varying factors. If those prices were easy to forecast, grain trading would be an easy way to millionaire-hood.
• Imported goods are sensitive to the exchange rate, which is similarly hard to forecast.
• Some prices are administered, such as the ever-rising cost of tuition.
• Other goods and services have varying input costs associated with them -- labour, rent, energy, etc.

For each of these components, the reduced form model will vary greatly. For example, even if you think inflation expectations are really important for the determination of inflation, the expected CPI inflation rate in Australia is going to mean diddly squat for the global price of oil.

Concluding Remarks

Beware reduced form inflation models.

Footnotes:

* "Judgment" is (or at least was) the preferred British spelling, and is used in Canada in legal contexts. So it should be fine to write the sentence "In my judgement, the judgment in the case was incorrect," in a Canadian spelling test.

** North Americans would probably prefer "a historical accident."

(c) Brian Romanchuk 2018