Sunday, March 31, 2019

Examining an MMT model in detail

What is MMT, the heterodox economic theory that has captivated Alexandria Ocasio-Cortez, made its way into the Green New Deal discussion, and inspired dozens of thinkpieces and critiques? What does it say? How can we tell if it's a good theory or a bad one?

These are incredibly important questions. Thanks to Ocasio-Cortez and the Green New Deal, MMT has very quickly gone from an obscure heterodox idea to one of the most potentially influential and important theories in all of economics. 

Formal Models vs. Guru-Based Theories

These days, most economic theories are collections of mathematical models. If you want to know what the theory says, you can parse out the models and see for yourself. You don't have to go ask Mike Woodford what New Keynesian theory says. You don't have to go ask Ed Prescott what RBC theory says. You can go read a New Keynesian model or a Real Business Cycle model and figure it out on your own.

MMT is different. There are many wordy explainers and videos that will explain some of the concepts behind MMT, or tell you some of MMT's policy recommendations. But that's different than having a formal model of the economy. In a criticism of MMT, Thomas Palley writes:
The critical economic policy question is what does the power to money finance deficit spending mean for government’s ability to promote full employment with price stability? This question can only be answered by placing that power within a theoretical model and exploring its implications...Proponents of MMT have a professional obligation to provide [a simple mathematical] model to help understand and assess the logic and originality of their claims. Yet, [MMT proponents Eric Tymoigne and L. Randall Wray] again fail to produce a model...If MMT-ers did produce a model, I am convinced the issues would become transparent, but readers would also see there is “no there there”.
Now, a lot of people like to criticize mathematical models in economics. And they do have their drawbacks. Economists can sometimes become so entranced by the precision of math that they ignore the need to connect that math with the real world. And the difficulty of hacking through math can lead economists to make the models too simple.

Furthermore, Palley is being a bit too strict in demanding math; formal models can be stated in English or in graphs, rather than in equations.

But formal models have important advantages. For one thing, a good formal model can be compared with quantitative data, to see whether it works or whether it fails. Formal models can make testable predictions.

A second advantage of formal models is that you can figure them out for yourself, without having to ask any gurus. If you have to run to the gurus to ask them what the theory says any time you think you've found a flaw, it becomes almost impossible to skeptics or outsiders to evaluate the theory objectively.

This latter issue comes up a lot when dealing with MMT. In a recent post, Brad DeLong expresses his frustration with the theory's apparent slipperiness:
"Functional finance" is a doctrine originated and set out by Abba Lerner...When I said that "functional finance" is at the core of MMT, I got immediately smacked down by one of the gurus...Perhaps the key to the eagerness of [L. Randall] Wray to dismiss me (and James Montier) for saying that MMT is Lerner+ is sociological. Perhaps MMT is not model-based ("IS-LM with a near-vertical IS curve") and not idea-based ("Functional Finance") so that it can be guru-based. (emphasis mine)
It wasn't just DeLong and Montier who conflated MMT with Functional Finance. Aryun Jayadev and J.W. Mason did something similar in their attempted write-up of MMT, leading Josh Barro to do the same thing in his own criticism of MMT. Mason, Jayadev, and Barro criticized MMT on the grounds that raising taxes to control inflation - something you have to do in Functional Finance, and which some MMT advocates agree is necessary - is politically very difficult.

But in response to these critiques, Mason, Jayadev, Barro, Montier, and DeLong were told: No, MMT's approach to fiscal policy is not just Functional Finance. It is very very different. On Twitter, MMT insider Rohan Grey declared that MMT has other tools besides fiscal policy for fighting inflation. On his blog, MMT insider L. Randall Wray said that MMT's main tool for maintaining price stability is the federal Job Guarantee:
Yes, MMT does have another tool to maintain price stability. It is the JG approach to full employment. It has always been a core element of MMT. We have never relied the simplistic version of Functional Finance that was presented by Mason. It would take about five minutes of actual research to demonstrate this.
Now that's a perfectly fine rebuttal. People get theories wrong all the time. It's perfectly possible that Mason, Jayadev, Barro, Montier, and DeLong were all very wrong to conflate MMT with Functional Finance, and that five minutes of actual research would have demonstrated this.

But which five minutes? If you want to know how MMT's price stabilization policy differs from Functional Finance, where do you look? Do you trust Wray's blog? Or Grey's tweets? Or an online explainer? Or a video explainer? Or in one of the many papers written by MMT proponents? Which one?

Because MMT doesn't often include formal models, the question of how MMT thinks inflation works is very difficult to answer for yourself. The same is true of a number of other questions, such as how MMT's Job Guarantee would create full employment with price stability. You have to go ask the MMT People themselves.

But "few formal models" doesn't mean "no formal models". Occasionally, MMT people do write down a formal description of how they think the economy works (or might work). One example is "Monopoly Money: The State as a Price Setter", by Pavlina R. Tcherneva.

Examining an MMT Model (or, "Pavlina Tcherneva and the EMPL of Doom")

"Monopoly Money: The State as a Price Setter" explains the idea of how a Job Guarantee would work. The formal model begins on p.130 (page 7 of the PDF).

Here is the model's "conceptual framework":

The government demands that people pay taxes in dollars, which you can only get by working for the government. So people work for the government because that's the only way they can pay their taxes.

Now let's look at exactly how people work for the government and pay their taxes:

Already I can see one potential problem with this model, which is that everyone in the entire economy dies.

Note the part that I've marked with a red arrow. This economy produces only one service, which is firefighting. But you can't eat firefighting. So if that's literally the only thing anyone does in this economy, everyone will starve to death.

LOL OK, so that's probably too harsh. Maybe the people in this economy are doing other stuff on the side that's not in the model - farming, manufacturing, etc. After all, Tcherneva does mention that T could be "a property tax", and the model doesn't include property. So let's assume there's other stuff outside the model too, so that people don't starve, or freeze, etc.

But that still leaves the question of why the government needs firefighting services in the first place. What if there aren't any fires? If the government hires more firefighters than it needs to actually fight fires, then it's just wasting resources - making people labor in useless toil, taking them away from subsistence farming, or whatever. The more useless firefighters it hires, the poorer each person is. (In the limit, if every person has to spend all their time and effort doing unproductive work for the government, they actually do starve to death!)

Anyway, let's go on. 

So the government can hire 1 firefighter for $10 or 10 firefighters for $1 each, etc. (cents apparently don't exist in this world, so you can't hire 4 firefighters for $2.50 each, which is weird but OK).

Another question arises: Why is anyone willing to work as a firefighter in this model? According to Tcherneva, they need dollars to pay their taxes. But who pays the taxes? Tcherneva specifies that the tax bill "for the entire community" is $10. But how is this tax bill paid? Does the entire community file taxes jointly?

Suppose the government chooses to hire only one firefighter - call her Susan. She gets $10 for firefighting, and everyone else gets $0. Does the government tax Susan $10 and tax everyone else $0? Apparently so. Because Susan has $10 and everyone else has $0, so the only way the government can get its $10 is to tax Susan $10.

So why does Susan go fight fires in the first place? Firefighting takes work. If instead she decides to be one of the 9 people who don't work as firefighters, she can relax and watch Netflix, or farm food, or go do whatever it is that people in this model do when they're not firefighting. Someone else will get the dollars, someone else will pay the taxes. Because if you don't work as a firefighter your tax bill is zero, because you don't have any dollars to tax!

So taxation in this model doesn't make a lot of sense, because the model doesn't explain who actually pays the taxes. Without knowing that, it's hard to know why people would accept the government's prices for their firefighting services.

But OK, there must be some reason they work. Maybe the government makes people work at the price it offers. In this case, people are effectively doing slave labor for the government.

That sounds bad...right? It sounds a bit like colonialism, when European governments would send their armies to Africa or Latin America or India and make the locals labor for their overseas masters. 

At this point you may be saying "Noah, you jerk, why are you slandering the MMT people by equating their ideas with colonialism? That's not fair!" But I have a good reason for drawing this comparison: Tcherneva makes it herself, earlier in the paper! European colonialism in Africa is explicitly held up as a successful example of a "tax-driven currency":

On p.135 (page 12 of the PDF), Tcherneva explicitly states that the case where the government sets prices for firefighters "is the African example." So yes, it's explicit that in the model where the government sets prices, it also sets quantities - i.e., it forces people to work. 

I don't want to editorialize too much here, but "the African example" seems bad. Forced labor, potentially to the point of genocide, does not sound like the kind of "job guarantee" I would want. 

Does Tcherneva offer any alternative? Yes. In her "case 2b" on p.134 (page 11 of the PDF), she relaxes the assumption of forced labor, and allows the price (and therefore also the quantity) of workers to be set by the market. She also allows for the possibility that labor unions might organize to reduce the amount of labor they do in exchange for their tax dollars:

In this model, the government can't determine how much labor it gets or how much each laborer gets paid; it can only determine the total number of dollars it pays for labor. In fact, as Tcherneva suggests, labor unions might even allow workers to get all the government's dollars for a very small amount of labor - good for workers, but bad news if there really is a fire!

Anyway, this new model sounds much less horrifying, but it still doesn't answer the fundamental questions of "why do we need all these firefighters" or "who exactly pays the taxes". But at least it does allow for the possibility that either market forces or organized labor could minimize the amount of potentially-useless labor people were compelled to perform for the government in exchange for the money they need to pay their taxes. Colonial Africa, but with labor unions.

(Tcherneva also develops some further cases, including one where the government purchases park benches and another in which desired net saving is nonzero. But none of these cases answer the basic questions either. But don't take my word for it; check the paper.)

Anyway, this formal model is very useful, because it allows careful, precise, explicit analysis of MMT ideas. It allows us to identify potential problems with the theory, such as the question of whether a job guarantee would represent unproductive toil, and whether that's something we would want as a society. 

Further Questions for MMT

There is much I do not yet understand about MMT, that I would like to understand. 

For example, suppose the government implements a job guarantee and sets deficits, credit policies, etc. at the appropriate level to ensure price stability, but we still have more inequality of income and wealth than we would like? What policies should then be used to reduce inequality, without interfering with the goals of price stability and full employment?

Also, I would like to know how to test MMT empirically. What concrete predictions - about macroeconomic aggregates or other quantities - does MMT make that would allow us to determine whether it describes the economy better than, say, Old Keynesian IS-LM models, or New Keynesian models with financial frictions?

Also, how does MMT model productivity in the economy? It seems like hiring a bunch of firefighters in the absence of fires would negatively impact productivity and reduce living standards. Does MMT assume that productivity is exogenous, does it assume that productivity reductions will always be of secondary concern relative to the importance of full employment, or does it have some other way of dealing with potential hits to productivity?

I'm not confident in my ability to answer these and other important questions by reading L. Randall Wray blog posts, or long online explainers, or wordy MMT papers. I want to be able to read a concrete, formal, well-specified model like the Tcherneva model above, and answer these questions myself. And the rest of the non-MMT econ deserves this as well.


  1. Noah thanks. Good critique. I'm a practical person who thinks things like a job guarantee have lots of stumbling blocks. I also don't care about the theory of where money comes from - which seems so important to MMTer. And so on. But as far as the real world of MMT's descriptions are concerned, MMT has, over the past decaders, explained to me (a) why the economic recovery would be relatively weak (b) QE is not inflationary, (c) debt will not "weigh down" the US economy(d) Europe will lag the US (e)higer deficits will increase US profits and thus bolster the stock market (e)Japan will not have a fiscal crisis (f)the bond vigilantes won't be impacting interest rates, (g) why interest rates would be relatively low and so on. As an investor, I would be both poorer and noticeably less knowledgable on these "big" issues without the exposure to MMT. I can't say for most other economists, pundits, etc.
    That is, you asked about "concrete predictions" from MMT. Maybe these explanations aren't "models" but what more could one ask for? Thanks.

    1. Those are all important points, but they are points that regular modern Keynesians like Paul Krugman and Simon Wren Lewis have made repeatedly, without resort to MMT. So the question is what, if anything, does MMT give us that e.g. SWL's view of the world doesn't give us? (Noting that SWL's view is widely understood and not difficult to explain using simple models.)

    2. Yes, good point. Although not true for Krugman IMO. Wasn't long ago where he felt a Japanese debt crisis was around the corner, for example. He's still apparently in a IS/LM world and believes in crowding out, etc. - disproven by MMT logic, etc. He's made progress over the years but is not there yet. Thanks.

    3. Paul Segal,

      That is the "MMT is nothing new" or maybe the similar "we knew it all along" claim.

      Those claims do not proceed. Bill Mitchell, in his blog, has already discussed and addressed those criticis, and I believe he is able to explain much better than I could (he is, after all, one of the "founders" of MMT).

      If you want more details, you should look for his blog entry entitled "Modern Monetary Theory – what is new about it?" (with 3 parts) or maybe another one entitled "The mainstream old guard tell it as it is – and how different that is to MMT". Those posts should answer your questions...

  2. Anonymous8:36 PM

    I really do wonder if MMT is String Theory for economics: silver bullet to answer all the hard problems, but they keep moving the goal posts and can't really specify how you would experimentally verify whether or not it is true.

    1. Anonymous11:22 AM

      Except String Theory is usually based on rigorous math, not cultish "we're the ones with revealed truth" attitude.

    2. Mathematics is neither a necessary nor a sufficient condition for good scientific theory.

      For example, evolution theory, although very scientific, doesn't rely on mathematical models. You don't need to write a mathematical model to define it or to explain it (although you could). So no mathematics here and still an excellent scientific theory.

      Another example: an economist made a (more or less) complex mathematical model with differential equations to study the economical dynamics of a population of vampires and non-vampires. The math is perfect and totally coherent. However, it doesn't matter: you couldn't use the model to explain real world phenomena, because there are no vampires in the real world. So here you have a lot of rigorous math and no actual scientific theory.

    3. Though not an economist (maybe a good thing?), I liken MMT to Supply Side theory. MMT supporters just need the little paper napkin with a nice curve that they can use in support of the theory; it seemed to really help Professor Laffer explain his!

    4. The rigor of String Theory math is irrelevant in the absence of testable hypotheses. Is the math in ST anything more than the candle flame that draws the moth?

  3. There's something else very odd in the model. The model appears to last for one period at the end of which taxes must be paid. This means that the velocity of money is assumed to be constant (as in a cash in avance constraint). This is very old fashioned for traditional monetary theory and really odd in a monetary theory which claims it and only it is modern.

    It doesn't fit recent data well (to put it mildly). In the actual USA people don't hold just enough money to pay their taxes. Also that amount would be (roughly) not much more than $ 350 billion as most taxes are withheld monthly.

    The model seems to me to consist entirely of extreme and unusual assumptions.

    1. People don't hold much money for taxes because of withholding and required quarterly tax payments. Most mortgage companies have escrow for property tax payments. So, I am a bit confused by this statement.

  4. One possible benefit of the JG: By deceptively putting 'full employment' (lipstick) on our unemployment rate (pig), would it at least make these MMTers stop saying that the government is lying about the labor market (?)

  5. plgaldon6:59 AM

    Maybe this model helps you answer your questions
    A Simple Model of a Currency Union with Endogenous Money and Saving-Investment Imbalances
    Berlin School of Economics and Law Working Paper No. 16/2012

    1. Observe how that paper/model links the "financial" world to the "real world": by assuming that the velocity of money is constant. The author doesn't even give importance to that. He simply assumes that and doesn't even discuss or criticise it.

      It seems just a little detail, but it is a big, enormous thing. Empirical evidence shows that money velocity is not constant. And it is actually quite easy to see that, you don't need to apply any sophisticated methodology to arrive at that conclusion. You just need to calculate the money velocity (with whatever definition of "money" you want) and see that it is never close to something you could call stability.

      In the real world, the money velocity is a endogenous variable influence by all the others, including C, I, G, S, etc. Also, unused capacity (like unemployment) is not in the model, and it is also influenced by all those variables and it also influence the money velocity.

      Hence, that paper/model is not useful to analyse the real world...

  6. "Another question arises: Why is anyone willing to work as a firefighter in this model? According to Tcherneva, they need dollars to pay their taxes. But who pays the taxes? Tcherneva specifies that the tax bill "for the entire community" is $10. But how is this tax bill paid? Does the entire community file taxes jointly?"

    It is a property tax. Everyone who has property pay taxes, or they will face the consequences of not paying it (enforced through the monopoly of violence enjoyed by the central authority).

    Of course, Tcherneva could have detailed more: everyone has an identical property so everyone is taxed equal ($1 per person), or maybe two of them have bigger houses and they pay $2.5 each, while the others pay $ 0.625, etc. That doesn't change the general concept. Also, the concept is not changed no matter how many goods you put in the economy (one, ten or millions) nor how many periods you put there. Also, it doesn't need to be about property taxes. Other kinds of taxes would do, but they would also interfere in people's decisions. Houses could be goods, and hence property taxes would affect the decision of people wanting to buy or sell houses.

    It is a toy model, used for didactical purposes. You cannot hope anything beyond that. Most MMT proponents do not like toy models because they are too limited, and I believe that is why you don't see many MMT mathematical models out there...

    1. drive-by commenter4:52 PM

      Most MMT proponents don't like toy models because they're not even competent enough to work a simple one out, as this case exemplifies.

      Yes, she could have detailed more. That's the point of writing down a model: To force you to explain the necessary details.

    2. They are very competent.

      In Consumer Choice Models, you don't need to detail the color of the hair of the agents, because it doesn't change the outcome of that toy model. Also, the general concepts remain the same no matter whether you put two or a million of goods there, and you don’t even need to name the goods. Of course, if you want to do some actual analysis of the real world, you would need to detail more and add complexity (like multiple periods, durable goods, production, storage, externalities, asymmetrical information, gaming and strategy, risk, long term commercial relationships, transaction costs, market power, political and military power, law, imperfect enforcement, trust, death, ignorance, irrationality, etc etc).

      In that MMT toy model, you also don´t need to detail exactly the name of the goods, you don´t need to work with thousands of goods, and you don´t need to be specific about the taxes.

  7. Bumper3:57 PM

    It seems like the key difference between the "African example" and the alternative (2b) is the lower real value of G and T. Whether they attribute that to organized labor or corruption or counterfeiting or liberal democracy, the forced labor isn't eliminated, only reduced.

  8. Bumper4:45 PM

    "Because Susan has $10 and everyone else has $0, so the only way the government can get its $10 is to tax Susan $10."

    If we're being charitable and assuming there is economic activity outside of the model then Susan can exchange her $10 for goods and services from whoever the taxpayers are. Or to someone else who would do the same.

    The forced labor is still there of course, but only hits those who bear the tax burden after being laundered through the out of model economy, so I guess that makes everything okay.

  9. Anonymous2:26 PM

    Noah, taking the piss out of the economy only produces fire fighting might not be the most constructive way to approach this... quite funny though

  10. People who are not able to be employed in the private sector would be willing to work as "fire fighters." Drop the neoclasscial assumption that the economy is always at or headed toward full employment and you will see the purpose of the job guarantee.

  11. MMT is a disaster! Catastrophic! The end of the world!

    Side note: the MMT crowd has been running Washington DC since about 1960. Since then, real incomes per capita in the US have about tripled.

    The current day academic MMT crowd and the present-day GOP-MMT crowd are distinguished in two ways: the academics are powerless, and the academics prefer to discuss additional spending as opposed to tax cuts. The academics appear to understand little about public opinion.

    I think MMT would work if they would just go to money-financed fiscal programs, rather than borrow and spend.

  12. The discrete model has an interesting implied result: only existing property owners can be firefighters and be paid by the government as posited.

    To see it, remember that money can only exist as a creation of the government. As of January 1, no money exists; by construction, December 31 following, all property owners together owe 10 dollars.

    Thus, a gov't wage of 1 dollar implies ten property owners, 2 dollars 5 owners, and so on, with a ten dollar wage implying only a single property owner.

    This assumes a legitimate government: one that does not seek to force a tax onto people who, by construction, have no means of paying it via the model mechanisms postulated.

    Yes, you could subsequently generalize to include a wider economy, but that isn't included in this particular model.

    I think, while you could probably come up with other social structures that can result consistent with the model, in formal terms they may all need some assumption about further economic structures.

  13. oops, followup to previous comment: the year time period I used is misleadinh, isn't it? Note simply that the single period could in principle be one day. In fact, by construction the period is the time period it takes to fight a single fire.

    In which case, positing extended economuc structures that allow money to flow between the whole population, i.e. that the gov't is putting a tax on property owners but providing the means to pay it to only a subset of those property owners becomes a little more strained; having the subset of the population being paid as firefighters as the only tax owing subset seems more parsimonious.

  14. One more comment: my previous two assumed that the population are possible taxpayers, and the gov't wage reveals how many actual taxpayers there are.

    Suppose instead that we posit that all of the population are actually taxed, but the gov't pays only a single one of them the 10 dollars, the only money available by construction.

    The interesting question is then how long the tax period must be, for a legitimate gov't that doesn't force someone to steal in order to fulfill a physical constraint of no legal money available.

    In this case the answer is that the tax period must be long enough for each of the other nine taxpayers to convince the firefighter to voluntarily give them a dollar for their taxes owed.

    Two firefighters would roughly halve this tax period for the remainder, and so on.

  15. I don't know much about MMT except it sounds bogus, however, in mitigation, so do most of the other schools of economics. So what if there's a model or not? For example, the monetarist model states money is short term non-neutral and prices, employment are sticky. But in reality they do not appear to be very neutral or sticky, and if they're not, all of monetarism goes out the window. Yet to this day you have diehards who believe (faith based economics) that money is non-neutral short term and prices are very sticky. Same with fiscal policy economics, the "Keynesian multiplier" which may in fact be zero or even negative. But it's hard to measure it. Same with the "natural interest rate" or "natural rate of employment" and other such occult variables in various models. So what if you can 'understand' the model? It does not make it less true. The advantage for MMT is precisely that, like communism, they don't need a model nor do they want a model. Since economics is a political process, not a science, they want you to believe. And once you 'believe' their unstated priors, in a way you're no different than a believer in the metaphysics of Milton Friedman or Lord Keynes. Faith based economics is MMT, but without the models that the other schools have.

  16. MMT and the canonical macroeconomic model
    Comment on Noah Smith on ‘Examining an MMT model in detail’

    For the full (8576 character) text see

  17. "Note the part that I've marked with a red arrow. This economy produces only one service, which is firefighting. But you can't eat firefighting. So if that's literally the only thing anyone does in this economy, everyone will starve to death."

    And unless they're going to spit or urinate on the fires, they'll probably need some equipment to actually fight the fires...before they die of starvation.

  18. The people fighting fires now have money (tax credits) that are in demand by others. Others can do non-fire services to earn those from firefighters. So long as the government doesn't tax out of existence all the tax credits and there are enough to circulate around, the system can work.

  19. Seems people are worked up over a simple model that one can read a little or a lot into.

    Ok, Government says firefighters only fight residential fires, and corporations will not be serviced. Now corporations need firefighting staff. What will they pay them? Will the public firefighter pay scale be used in determining those salaries? So couldn't government possibly be setting the price of a firefighter?

    I have friends that are volunteer firefighters, and many locales pay nothing for these workers. That is interesting too, but the fire incidence is less, and there are tax credits, plus many corporations support these employees in this volunteer time off. I think it is an interesting model - not saying it is right, just interesting.

    Regarding the colonial tax, that may only be necessary to "jump-start" a currency - get it accepted. Once accepted it can become more entrenched as a mechanism to facilitate trade, so maybe the taxes are not as important at that point.

  20. IMO, jobs aren't intrinsically good, though some people may genuinely like their jobs. For many others, and perhaps for most workers, jobs are good only in a world where one either has a job or dies of starvation and exposure. Hence the argument for the wonderfulness of colonialist labor conditions. A person who takes a job because she or he must prevent their family from starving, works in desperation.

    IMO, also, mathematics has an intense dulling effect on the science of economics, a science that otherwise would be far more enjoyable (and I write as a mathematician).

    An interesting read, Noah. Thank you.