Scott Sumner’s macroeconomics: my thoughts

He writes,

I’ll use “(e)” to denote a (market) expected value.

NGDP(e) is the single most important variable in macro; it should be the centerpiece of modern macro.

How can we work with a central concept that is purely mental? Nominal GDP, or NGDP without the (e), is a measure derived from the market exchange of goods and services. Most concepts in macroeconomics, such as interest rates, or prices, are observable when goods or securities change hands.

NGDP(e) is not an observable result of goods or securities changing hands. It is something in people’s heads.

But it’s even more problematic than that. At least 99.9 percent of all people do not even have an NGDP(e) in their heads. Even most economists do not have one.

Even if you had a futures market in NGDP, NGDP(e) would still be a purely mental concept. In the wheat market, if you sell 1000 bushels of wheat forward, on the day that the contract expires you could deliver 1000 bushels of wheat to fulfill the contract. (Actual delivery does not take place in such markets, because buyers and sellers unwind their positions at expiration.) But nobody can deliver NGDP against an NGDP futures contract. It is a pure bet.

In any event, an NGDP futures market currently does not exist. So the most important variable in macroeconomics is something that exists in people’s minds, and yet in truth it exists in almost no one’s mind. I worry that this is like saying that in physics the most important variable is the ether, even though no one can observe it.

Sumner writes,

Low and stable NGDP growth minimizes the welfare costs of “inflation”, and also leads to approximately optimal hours worked.

NGDP is an observable variable, and Sumner argues that low and stable NGDP growth is associated with good performance of inflation and employment. So why bother with NGDP(e) at all?

At the risk of putting words in Sumner’s mouth, I think he would say that NGDP(e) is important because the Fed affects NGDP by manipulating NGDP(e). How did he get there?

Monetary theorists used to say that the Fed manipulates NGDP by manipulating the quantity of money. The problem with this is that it is impossible to find a definition of money that can satisfy two conditions at the same time: (1) that the Fed can control it; and (2) it closely correlates with NGDP. The former requires a narrow definition of money, and the latter requires a broad definition.

Old Keynesians said that the Fed manipulates NGDP by manipulating the short-term interest rate. When the short-term rate gets stuck at zero, it has to manipulate the long-term rate. Or it becomes impotent. But even when it is not stuck at zero, the Fed’s manipulations often seem ineffective. For one thing, long-term interest rates sometime do not respond, or they respond perversely.

Then there are the New Keynesian types who say that the Fed manipulates NGDP by manipulating expected inflation. But to me that is another ethereal concept. At the risk of putting words in their mouths, the New Keynesians are saying that the Fed can mysteriously change expected inflation through “quantitative easing” even if short-term interest rates and long-term interest rates are both impervious to Fed actions, or even if long-term rates react perversely.

From the New Keynesian view, it is a relatively small step to Sumner’s view. Just swap out the ethereal expected inflation for the ethereal NGDP(e).

Got it? In modern macro, we have everybody working in the GDP factory. And we have everybody forming expectations about the price of the output from this GDP factory, or about the total nominal value of that output. And booms and recessions are caused by changes in these expectations. And the Fed can manipulate these expectations through an immaculate process that cannot be measured using interest rates or the money supply.


I know that almost nobody who reads Specialization and Trade buys into my view that movements in aggregate price indices mostly reflect habits and inertia, rather than central bank operations. But when you see the contortions that monetary theorists have gone through over the years, I think I have a fair case.

This entry was posted in Monetary Economics, PSST and Macro, Scott Sumner is Coherent. Bookmark the permalink.

13 Responses to Scott Sumner’s macroeconomics: my thoughts

  1. Dave says:

    I have never understood the enthusiasm for Sumner’s ideas, which are completely speculative. At least it seems like other economists have lost interest; this is the first time I’ve seen his name mentioned in a long time.

  2. Matthew Young says:

    NGDP minus RGDP is the expense of price discovery. The expense eventually is written off as a loss, becoming a bond default on government which insures against inflationary losses on government programs.

    • Les Cargill says:

      There is never a reason for a politically sound government to default. It’s simply never seen in the wild – default or hyperinflation are extremely closely correlated with regime failure or the like.

  3. Doc at the Radar Station says:

    “…movements in aggregate price indices mostly reflect habits and inertia, rather than central bank operations.”

    I think that is the most common sense explanation that I’ve ever read. I can see why you are so skeptical about macroeconomics. I suppose the market monetarists and New Keynesians would say that habits and inertia can be modeled and influenced. The trick is whether they can *effectively* be influenced or not I suppose.

  4. Dave M. says:

    If the Fed wanted 2018 US NGDP to be more than $20T (for example), it could say, “whenever betting markets generally predict 2018 US NGDP under $20T, we will start to buy and hold (non-US-currency-like) assets of our choosing, at a rate of $1B the first day and doubling the rate every day thereafter, until betting markets generally predict 2018 US NGDP over $20T.” (A symmetrical promise could be used to try to keep 2018 NGDP under, for example, $21T. It is possible but unlikely in this example that the Fed would run out of assets to sell.)

    Large negative NGDP(e) surprises are bad regardless of their cause and regardless of measurement difficulties. When there is a large negative NGDP(e) surprise, the mismatch between downward-sticky wages and much more plastic hiring/firing/layoffs aggregates into a large short-term misallocation of real resources. Sumner calls this the “musical chairs model.” (Large NGDP(e) surprises, positive or negative, may also be bad because of sticky long-term nominal contracts, menu costs, etc.)

    Therefore, the Fed should buy whatever it takes to reverse large downward surprises in NGDP(e). (It should probably also sell whatever it takes, unless/until it runs out of assets, to attempt to reverse large upward surprises in NGDP(e).)

  5. Morgan Warstlet says:

    The power of NUDE in Sumners world is the Level Target LT.

    In fact in Scott world, the Fed would get most the bang, just from switching to a Level Inflation Target.

    It puts the Fed I’m automatic pilot, it becomes an algorithm. We know for a fact high frequency trading works. Of course this would work too.

    The BANG of this, is Chuck Norris, after a few dudes get their heads caved in as the machine ratchets up buying or selling to hit that monthly number as predicted by 2% inflation to infinity…. nobody fights, they may hedge against tiny swings, up you know for sure what dollars are worth forever.

    A simple look up table removes variables in business planning.

    All of this stabilizes expectations to RGDP alone.

  6. Andrew' says:

    I still don’t understand why we won’t overestimate NGDP resulting in a crash. Except then we would KNOW recessions had been cured.

  7. BC says:

    “movements in aggregate price indices mostly reflect habits and inertia, rather than central bank operations”

    Doesn’t long-term growth in base money supply (a narrow enough definition of money that the Fed can control it) across countries correlate closely with long-term inflation across countries? See for example Sumner’s table here:

  8. Dude says:

    Isn’t the S&P 500 just a measure of expectations? Why would NGDP(e) be a “purely mental concept”, when other indices are seen by most people as real things?

    Also, Morgan is right.

  9. Les Cargill says:

    I’m not following you.

    NGDP is simply RGDP plus inflation. There are significant lags. I’m not sure of the futarchic “NGDP market” either, but the idea is at least for a mechanism where when RGDP is down, inflation is used to balance the scales.

    Take a look at corporate indebtedness. It’s rising faster than most other forms of debt. Many, many significant companies are dead (wo)men walking, zombies.

    The alternatives to NGDP targetting seem to be a return to the gold standard, or whatever it is we now have. And isn’t part or much of Tyler Cowen’s “complacency” simply that people in a low-growth economy will choose more-rentier investments rather than suffer the slings and arrows of actually producing something?

    Perhaps ( I am no expert, so … ) NGDP(e) is simply the then-common ( after an NGDP targeting regime is adopted ) knowledge to expect higher inflation when RGDP falters. At least maybe then velocity will stay up.

  10. Handle says:

    Let me see if I can explain my understanding of it via a metaphor.

    Let’s say you own a huge dam on a major river feeding California’s demand. You have one tool, which is a big wheel that adjusts the flow valve. The supply into the reservoir is volatile and uncertain from year to year. One never knows how much precipitation there will be, or when the snow will melt, or how fast. One also doesn’t know exactly how much water or electricity will be wanted that year. If it’s hotter or cooler than average, people will want more or less, respectively.

    Before the dam, a few people had to guess about the weather, which they were very incompetent at doing. Most people don’t pay attention to precipitation or weather or have any idea about them, but they notice the information tricking downstream (as it were) in the form of the prices of the things they do care and know about, which become the local knowledge about the things that affect their particular business activities.

    Some of these businessmen may not even be aware that some of the volatility they experienced in the “business climate” in a yeear was due to the actual climate, as the effects on the weather and water-dependent sectors rippled throughout the spiderweb of the broader economic ecosystem.

    What the dam does is help to stabilize that volatility. The dam owner can promise to deliver a much more reliable and predictable stream of water, in wet or dry years, and to try to manage average flows to match average precipitation, while also delivering quantities that more-or-less match demand and cause the least disturbance to the overall patterns of specialization and trade. As with actual dam owners, even the long-term climate models are only so good at forecasting, but the existence of the reservoir gives a lot of slack to these bad estimates with the ability to compensate for any past errors with future adjustments.

    How should the dam owner decide how wide to open the flow valce at any one time? One way could be to look at the prices in the futures markets for the most flow-dependent industries. The dam owner could say he intends to keep the prices of those products stable and predictable from year to year, and will adjust the flow accordingly. If the futures markets aren’t cooperating, then the wheel could be turned one way or the other to accomodate for whatever issue is causing the disturbance.

    With this one source of stabilization and predictablity, the rest of the economic ecosystem is buffered against, and much less susceptible to the formerly natural volatility. They know they can depend, more or less, on a a narrow band of results for one of the most important features of their economy, and without having to carry around any climatic details in their heads, the information flow via prices will still allow them to coordinate their economic activities in a more stable environment, where they have to worry about risk less, and have to allocate fewer resources to securing against risk.

    • Les Cargill says:

      But we have stabilization. Oodles of it. What we’re short on is growth.

      So in signals, you can never improve the entropy of a signal. You can only move that entropy somewhere where it’s less bothersome.

      What you describe is like putting a big “integrator” on the economy. We probably need that, but when we’re in a state where inflation targeting is used and we still can’t even hit low targets ( 2% for a long time ), something else is going on.

      It is as if we’re still fighting the war on inflation, purely out of habit.

      • Handle says:

        One funny thing with trying to pass the “Intellectual Turing Test” and with expressing someone else’s views in the form of the argument I guess they might make themselves, is that it often turns into a Devil’s Advocate experience, trying to deal with a bunch of criticisms, even though I don’t actually hold the view I’m trying to explain.

        As for wanting higher real growth, I think Kling and Sumner would both agree that Central Banks have little or no effect on that over the long run.

        The point of the dam metaphor was to illustrate that it isn’t necessary for most participants in an economy to have any “views” or estimates regarding their future expectations of some particular aggregated number. That was a key aspect of Kling’s criticism in the original post. It’s sufficient for everyone to know that some important price – a historic source of volatility – is being managed to be as predictable and stable as possible. It gives everyone a common “pole star” of reliably-shared expectations, which allows them to plan and coordinate their activities with a lower amount of inefficient hedging for uncertainty, volatility, and risk. The information about this one major stabilization will trickle through the rest of the economy as “reaction consequence information” in the form of different prices that are relevant to each particular sector. Now, as to why NGDP in particular is on the A-list for stabilization targets vs anything else, well, that’s a different argument. But again, the point is that some number be managed to be predictable as common knowledge in the Aumann sense.

        As it happens, my views are somewhere between Kling and Sumner, and I don’t know if that makes me more or less heterodox than either of them alone. I’ll articulate those views more fully when another relevant thread is posted.

        At any rate, Sumner would definitely agree with you that central banks all over are still fighting the war against inflation out of very misguided motives and reasoning. In his view, central banks never run out of ammunition and can always and easily hit any arbitrary target, if not with exact accuracy in the short term, then to any level of precision of the long term average. If they say they want to hit X% and fail to do so consistently, then it’s either because they’re lying about what they really want to do, or have bad reasons to be too nervous to use the tools they have.

Comments are closed.