Recent Posts

Sunday, November 1, 2015

Output Gaps And MMT

A commenter on my blog drew the article "Thin Air’s money isn’t created out of thin air," (by Michael Pettis) to my attention. It refers to output gaps, which appears related to my recent article on output gaps. In his article, he makes some statements about Modern Monetary Theory (MMT). Although I would put myself in the MMT camp, I would note that I cannot write on behalf of an official MMT line; I disagree on a few points, and it is unclear what the MMT position is on some topics.

Pettis' article is quite long and contains a great deal of content. Although I find myself in agreement with much of what he says, he packs in so many theoretical asides that I cannot hope to summarise them. As a result, I am not attempting to respond to his underlying thesis. Instead, I just want to highlight some hidden assumptions around output gaps.

What Does MMT Say?

The discussion of MMT within Pettis' article is found in the following paragraph.
But it is possible not because banks can fund investment by creating debt “out of thin air”. This statement is either highly confused or it too-easily leads others into confusion. There is a related form of this question that often seems to come out of the MMT framework, although I have no idea if this is a misreading of MMT or if it is fundamental to the theory, but while banks can create debt, they do not automatically create additional demand. According to MMT, as I understand it, there is no limit to fiscal deficits because governments who control the creation of money can repay all obligations regardless of their taxing capacity simply by monetizing the debt (which of course means nothing more than exchanging debt which we call “debt” for debt which we call “money”).
Pettis admits that he is unsure what the official MMT position on these topics is. I would note that I am unsure myself; some MMT writings leave themselves open to a few interpretations. My reading is that the MMT literature is fairly close to pre-existing Keynesian and post-Keynesian views, and is not that radical theoretically. MMT is only radical relative to conventional views about fiscal policy.

However, MMT authors appear to subscribe to a different conception of the output gap. If I paraphrase their views using terminology I used in my previous article, they appear to subscribe to a concept of a generalised output gap, whereas the mainstream (and possibly Pettis) believe in a stricter version of the output gap -- where the output gap is the difference between actual (real) GDP and a well-defined "potential" (real) GDP.

Differing Views On Demand Created Out Of Thin Air

Pettis asks what happens if demand is created "out of thin air," either as the result of bank lending (private sector debt issuance) or government deficit spending. It is somewhat unclear as to what he exactly believes, as he runs through a number of different cases. He outlines a few cases.

Firstly, he outlines a "full capacity" case.
In the first case, assume that we are in an economy in which there is absolutely no slack. Workers are fully employed, inventories are just high enough to allow businesses to operate normally, factories are working at capacity, and infrastructure is fully utilized.
If in this case Thin Air’s expenditures cannot be satisfied without diverting goods and services that are being used elsewhere. Whether the new credit or the deficit spending goes to support investment or to fund consumption, all the goods and services that an economy is capable of creating are already being used by other economic entities.
Secondly, there is a case with "excess capacity."
In the second case, assume the other extreme, in which the economy has a tremendous amount of slack – there are plenty of unemployed workers who have all the skills we might need and can get to work at no cost, factories are operating at well below capacity and they can be mobilized at a flick of the switch, and there is enough unutilized infrastructure to satisfy any increase in economic activity.
In this case when loan creation or deficit spending creates demand “out of thin air”, in other words, it also creates its own supply. When Thin Air spends money to buy certain goods or services, those goods and services are automatically created by switching on the factory equipment and putting unemployed workers to work.
He then notes that neither of these possibilities are completely realistic.
In reality no economy will ever have zero slack, as in the first case, or full slack, as in the second, but instead will exist in some combination of the two. In that case Thin Air’s demand created “out of thin air” will partly be met by suppressing demand or investment elsewhere within the economy, and partly by creating the larger amount of goods and services needed to satisfy the increased demand.
In my view, his description looks very similar to a strict version of the output gap. Even though the economy may not be exactly at "full capacity" or "excess capacity", we still have a good idea about the amount of "slack" in the economy. In my view, we need to have a more generalised approach to the output gap, and we end up with a different perspective on government policy.

A Theoretical Explanation Why There Is No Unique Potential GDP

I will first give a theoretical reason why we cannot believe that there is well defined unique level of potential GDP, as is assumed by mainstream conceptions of the output gap.

The reason why mainstream models end up with a potential GDP concept is because they assume that there is only a single (composite) consumer good. Given the amount of capital available, and the number of hours worked, there is a well defined production function that determines the amount of this composite good produced within an accounting period. (In crazier models, there is an uncountably infinite number of consumer goods, but we integrate them all out to create a composite good.) We can then determine some level of production for this composite good which acts as "potential" GDP, as it is used in strict output gap definitions.

When we turn to the real world, we see that there are many different goods and services produced. Moreover, we see that productive capacity is not well defined, particularly for services. We could imagine that by changing the mix of goods and services purchased within an accounting period, we could increase measured real GDP, without hitting "capacity constraints".

For example,
  • people can download more copies of electronic books (a topic close to my heart), increasing sales volumes, without hitting any server capacity constraints. The prices of those e-books would probably drop in response to increased sales, and so we would have a decrease in the price level in response to increased demand (if all else was held equal).
  • People could go to unpopular restaurants instead of eating at home. The amount of food consumed would be similar, but since the restaurant meal is more costly, measured real GDP would rise.
The key is that this increased real GDP is not corresponding to an increase in the demand for "real resources," which is a concept often referred to by MMT authors. The demand for "real resources" can be viewed as the mechanism behind a generalised output gap. Since we cannot tie any particular level of real GDP to "real resources," we cannot decide what is the unique level of "potential" GDP.

However, we see that something like a "potential GDP" measure appears to work in econometric studies. The reason is that increased monetary incomes probably leads to increased spending, and the consumption basket in an accounting period is likely to resemble that of the previous period. Pretty much by definition, people do not flock to unpopular restaurants, they crowd into popular restaurants. If households in aggregate spend more on restaurant meals, the popular restaurants are likely to hit their capacity to serve customers, and start to raise prices.

This appears to salvage the notion of "potential" GDP as an analytical tool. But it does not tell us about policy. The objective of MMT proponents is to make changes to the structure of the economy, and thus their objective is change the structure of the consumption basket.

No Such Thing As NAIRU

My previous point was a theoretical point which is not of great import for the discussion of policy. However, the MMT authors have a very strong disagreement with the mainstream with regards to the role of unemployment within the economy.

The mainstream conception used to be that there was a well-defined level for the "nonaccelerating inflation rate of unemployment" (NAIRU; described in this article). The idea is simple. If the unemployment rate drops below the NAIRU (which can move slowly over time), inflation will rise. (Vice versa for an unemployment rate above the NAIRU.) The concept failed spectacularly in the 1990s, and so it was dropped by most economists, but the various monetary policy hawks kept the NAIRU alive in the latest cycle (and were hopelessly wrong about inflation as a result).

I would paraphrase the MMT critique of the NAIRU as follows. The NAIRU concept sort-of "works" for historical data because the unemployment rate is correlated with the generalised output gap. As long as we keep the same economic structure, a falling unemployment rate corresponds to a greater demand for real resources.

But the whole point of shifting economic policy is to shift the structure of the economy. In particular, MMT proponents argue that the government can eliminate involuntary unemployment by guaranteeing a job for everyone who wants one at a fixed wage. As a result, the unemployment rate would just consist of people in the process of taking time off, or looking for a new job. Most statistical studies pin that population of workers as being about 2% or less of the total. As a result, the unemployment rate would probably be around 2%, which is way below any econometric estimate of the NAIRU. (In fact, the unemployment rate would presumably cease to be a useful measure of anything, and what would matter is the percentage of the workforce in the guaranteed jobs programme.)

However, there is no reason to expect continued inflation to result. In the short term, there would likely be a jump in the price level, as it is assumed that the government-guaranteed wage would be around the current minimum wage. Since a lot of existing minimum wage jobs are terrible, workers would probably leave those jobs and take the guaranteed job. (For example, many minimum wage jobs are part time and on a random schedule, whereas the government jobs would presumably be full time and on a regular schedule.) A lot of employers that rely on poorly compensated workers would either go out of business, or be forced to raise the wages that they pay. 

This one-time structural shift would probably eliminate or raise the price of many low cost goods and services, and so the price level would rise. But once the creative destruction period is finished, wages will have no reason to accelerate further, unless the government raises the guaranteed wage. Eventually, an extremely low unemployment rate could co-exist with price stability.

As a result, real GDP could jump due to increased labour force utilisation without corresponding inflationary pressures. This would not be possible if the strict version of the output gap (NAIRU) were true. Since this jump accompanies a policy-driven change in the structure of the economy, which is why we do not see this type of behaviour in economies with relatively unchanged labour market frameworks.

Open-Ended Capacity?

Finally, I believe that Warren Mosler has made some comments that question the limitations of potential output. (Unfortunately, I do not have a reference, so I am paraphrasing here.)

He observed the jump in production that happened during World War II in the United States. There was no way the private sector could have produced the goods needed to fight the Axis powers solely based on market signals; the government was forced to intervene deeply in the economy in order create the war machine needed for victory. Meanwhile, the recent boom in China provides a peacetime example of a government-induced surge in production. 

There are drawbacks to heavy government interference with the economy. But at the same time, if there is a national interest in augmenting some form of production, a competent government should be able to achieve that. This growth increase could not be predicted by extrapolating the growth of private sector production. The developed economies have a strong free market bias in policy making, but it is unclear whether that bias would survive in the face of some future emergency.

Concluding Remarks

Most commentary on the output gap embed a lot of assumptions. If the economic structure is unchanged, those assumptions may not matter. However, we cannot extrapolate such analysis to an environment where policies are deliberately taken to alter the structure of the economy.

(c) Brian Romanchuk 2015


  1. Any view on Tom Hickey's comments here Brian:

  2. Interesting post. I dug up my notes on the CBO's A Summary of Alternative Methods for Estimating Potential GDP (PDF):

    1. Start with the Phillips curve.
    2. Observe the current relation between inflation and unemployment.
    3. Use this relation to estimate the NAIRU.
    4. Use the NAIRU as "the natural rate of unemployment".
    5. Compare the natural and actual rates of unemployment.
    6. Call the difference "the unemployment gap".
    7. By Okun's law, the output gap is twice the size of the unemployment gap.
    8. Calculate the output gap.
    9. Add the output gap to actual output, to get potential output.

    There's more to it. But this is where the Potential Output number comes from.

    1. Yes, that's either what I called the "squishy middle output gap" or the "generalized output gap". it's an estimate based on observed outputs and a model. As I noted, I think that the basic concept is sound, but I am unsure about the precise methodologies used. Bill Mitchell has done much more detailed analysis of the estimation biases in traditional techniques; my comments are a paraphrase of his analysis.

  3. Nice post, Brian. Conventional Keynesians (including the current ones) would agree with the statement that the value of output gaps, NAIRUs,etc could be altered if structural changes occur (ie. changes to regulations, labour market policies, so on). Using the old language, such policy changes would shift the Phillips curve, NAIRU, etc lower. Similarly, I'm not really sure what good there is in denying potential GDP numbers. I thought MMT gave up this angle. See slide 25 from this Kelton presentation using the potential GDP concept (properly, I might add):

    1. As I noted, this is more my spin on things. I question the idea of a traditional output gap based on the complete inability of anyone to pin what it is down right now.

      Bill Mitchell had written about output gaps and NAIRU, as traditionally defined, and he questioned the estimation methods. I am unsure how to summarise what he wrote, but my interpretation is that he ends up with something that I interpret as a "generalized output gap". It's how I phrase the concept, coming as an outsider to the Keynesian literature.

    2. I looked at the presentation (for those who have not looked, it shows the deviation of real GDP versus trend as a measure of the output gap). Bill Mitchell has also referred to the same type of chart.

      I would argue that such an analytical tool does work within the current "regime", but things get more complicated if policy changes the regime. Those complications show up in the discussion of the introduction of a Job Guarantee.

  4. Brian,

    Are you buying any of this?

    All this talk of the output gap being closed when long term structural unemployment is still at record levels simply defies logic and makes me wonder whether we should just drop the output gap altogether.

    1. I looked at this quickly. That's probably what their model should say, and it's perhaps not too surprising that the Fed staffers will want to blindly follow their models.

      My feeling is that their models are broken, as they assume that the economy will always return to "normal" as "random shocks" disappear. I see little evidence that economic variables are markedly different in 2015 versus 2013. But the Fed is not going to admit that pretty well all their theoretical work is wrong, so I see a good chance that they do what the models tell them to do.

  5. Pettis should stop writing books until he understands a few fundamentals about how a sovereign fiat currency economy actually works. He would do well to start off by reading Warren Mosler's The 7 Deadly Innocent Frauds of Economic Policy. Then follow it up with a bit of Randy Wray,

  6. My reading of Minsky (Stabilizing an Unstable Economy) is that political forces in the United States since World War II favor a policy regime where the mix of investment (I) to consumption (C) is too large causing excess capacity especially during a recession or depression driven by financial instability. The policy mix must shift to encourage reduced investment relative to greater consumption, thus there is less growth in capacity in the economy over time relative to the population. A federal job program would also reduce unemployment and sustain aggregate demand from the installed capacity which should grow more slowly when consumption is a greater portion of the economy than investment. This makes sense if you think of an economy evolving on a logistic growth curve. The ratio of investment to consumption would be large for an emerging growth economy and the ratio of investment to consumption would become smaller so the economy can evolve on a logistic curve rather than showing a pattern of overshoot and collapse.

  7. Larry Summers: Advanced economies are so sick we need a new way to think about them:

    Summers also tweeted this: "There appear to be more cases where recessions reduce subsequent growth of output than where output returns to trend."

  8. The distinction between a generalized and strict output gap is key, but I don't know that MMT itself requires one view over the other. MMT has policy implications only because supply matters. If supply is held constant, MMT offers no useful insights. It remains true in the way it is true that one can compute a de Broglie wave for a major league fastball, but that fact is not relevant, as no major league pitcher cares.

    MMT permits fiscal expansion because it uses inflation as the constraint. Inflation is too much money chasing too few goods. Thus, MMT brings the supply of goods into play, and the supply of goods depends on the relevant output gap, i.e., the output gap with respect to things scarce enough and ubiquitous enough to generate inflation. We can call the ability of the economy to produce whatever additional stuff is demanded without untoward inflation happening "the output gap."

    Note, though, that the importance of supply does not flow from MMT. On the contrary, MMT flows from the importance of supply. For reductionists like me (and, I would assume for Rothbardians!), a single axiom should supply a lot of intellectual juice. I would offer as an axiom "supply matters." From that axiom, one would look at the rate of change in supply over history to discover periods when printing money would have caused unwanted inflation (Weimar Germany without the Ruhr) and periods when it would not (the world with container ships and robots). One would then infer that in "normal times," when the output gap is small and steady, MMT has nothing to offer, because, in such times, traditional monetary and fiscal policy correctly treat supply as one of those "all else" things that remain equal when such policy is made.

    But when a new productive genie or three escape the bottle, the supply constraint on policy relaxes. MMT explains why that is so. The key is not that the output gap is generalized but that it is dynamic, i.e., changing in size (and perhaps, but not necessarily in make-up) sufficiently to require that the rate of change in it (as opposed to its static size as per Okun) be reckoned with. MMT has the tools to do that.

    I raise this quibble because there is a tendency for anti-MMTers to attribute beliefs to MMTers because they are MMTers. I believe that supply matters, so I lean toward MMT, but I don't "subscribe" to it, as I haven't explored all of its nooks and crannies. And anyway, what would be the point? I believe that supply matters, and I believe all of the inferences that I, myself, can draw from that axiom. To the extent I, or anyone, reasons from the notion that supply matters, the label "MMTer" is meaningless. Whether or not Mitchell and Mosler are right about anything else, supply still matters, and inflation is still the only economic constraint on monetary and fiscal policy (pretending for now that those are separate things). For me, the output gap is not a "thing"; it is a reification of the economy's ability to absorb spending. That leads, I think, to preference for the "generalized" notion, whatever MMT may say on the subject.

    1. Those are reasonable points. My concern here is whether we can measure how much slack there is, and not how to remove that slack. If we do not know where the constraints are, how do we know whether we are hitting them?


Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.