Saturday, May 25, 2013

New Keynesians Don't Believe Their Models

Here's the thing about about saltwater, New Keynesian economists: They don't believe their own theory.

Via John Cochrane, here is a great example. In the NBER Macroeconomics Annual a couple years ago, Gauti Eggertson laid out the canonical New Keynesian case for the effectiveness of fiscal policy when interest rates are at the Zero Lower Bound. In the model Eggertson describes there -- the model that is supposed to provide the intellectual underpinnings for fiscal stimulus -- the multiplier on government spending at the ZLB is indeed much larger than in normal conditions, 2.3 rather than 0.48. But the same model says that at the ZLB, cuts in taxes on labor are contractionary, with a multiplier of -1. Every dollar of "stimulus" from the Making Work Pay tax credit, in other words, actually reduced GDP by a dollar. Or as Eggertson puts it, "Cutting taxes on labor ... is contractionary under the circumstances the United States is experiencing today. "

Now, obviously there are reasons why one might believe this. For instance, maybe lower payroll taxes just allow employers to reduce wages by the same amount, and then in response to their lower costs they reduce prices, which is deflationary. There's nothing wrong with that story in principle. No, the point isn't that the New Keynesian claim that payroll tax cuts reduce demand is wrong -- though I think that it is. The point is that nobody actually believes it.

In the debates over the stimulus bill back at the beginning of 2009, everyone agreed that payroll tax cuts were stimulus just as much as spending increases. The CBO certainly did. There were plenty of "New Keynesian" economists involved in that debate, and while they may have said that tax cuts would boost demand less than direct government spending, I'm pretty sure that not one of them said that payroll tax cuts would actually reduce demand. And when the payroll tax cuts were allowed to expire at the end of 2012, did anyone make the case that this was actually expansionary? Of course not. The conventional wisdom was that the payroll tax cuts had a large, positive effect on demand, with a multiplier around positive 1. Regardless of good New Keynesian theory.

As a matter of fact, even Eggertson doesn't seem to believe that raising taxes on labor will boost demand, whether or not it's what the math says. The "natural result" of his model, he admits, is that any increase in government spending should be financed by higher taxes. But:
There may, however, be important reasons outside the model that suggest that an increase in labor and capital taxes may be unwise and/or impractical. For these reasons I am not ready to suggest, based on this analysis alone, that raising capital and labor taxes is a good idea at zero interest rates. Indeed, my conjecture is that a reasonable case can be made for a temporary budget deficit to finance a stimulus plan... 
Well, yeah. I think most of us can agree that raising payroll taxes in a recession is probably not the best idea. But at this point, what are we even doing here? If you're going to defer to arguments "outside the model" whenever the model says something inconvenient or surprising, why are you even doing it?

EDIT: I put this post up a few days ago, then took it down because it seemed a little thin and I thought I would add another example or two of the same kind of thing. But I'm feeling now that more criticism of mainstream economics is not a good use of my time. If that's what you want, you should check out this great post by Noah Smith. Noah is so effective here for the same reason that he's sometimes so irritatingly wrong -- he's writing from inside the mainstream. The truth is, to properly criticize these models, you have to have a deep knowledge of them, which he has and I do not.

Arjun and I have a piece in an upcoming Economics and Politics Weekly on how liberal, "saltwater" economists share the blame for creating an intellectual environment favorable to austerian arguments, however much they oppose them in particular cases. I feel pretty good about it -- will link here when it comes out -- I think for me, that's enough criticism of modern macro. In general, the problem with radical economists is they spend too much time on negative criticism of the economics profession, and not enough making a positive case for an alternative. This criticism applies to me too. My comparative advantage in econblogging is presenting interesting Keynesian and Marxist work.

One thing one learns working at a place like Working Families, the hard thing is not convincing people that shit is fucked up and bullshit, the hard thing is convincing them there's anything they can do about it.  Same deal here: The real challenge isn't showing the other guys are wrong, it's showing that we have something better.


  1. We have something better, a criteria simple and clear... a slide rule to free your mind... Do you want to know if payroll taxes should be cut at the ZLB? Just ask one thing... Are the cuts in payroll taxes relatively more or less than the cuts in capital income taxes? If the cuts to payroll taxes are relatively less than the cuts to capital taxes, they would be contractionary at this point.
    The tide of the past couple decades needs to be reversed.

  2. J., I think you perspective is right on. Critique only takes you so far (and reliance on it often betrays an unwarranted arrogance). New knowledge is created by the scientific method and the progressive project (especially to the extent significant reorganization of society is being considered) really does need new ideas and knowledge.

  3. As a hobbyist, I struggle to understand "the economy". Discussions of "economics" stand out for me as something not useful to my pursuits. The difference is magnified by my lack of knowledge, to be sure. But you are right: Far too many analyses consider things other than understanding the economy.

    "In general, the problem with radical economists is they spend too much time on negative criticism of the economics profession, and not enough making a positive case for an alternative."


    1. reply The Arthurian,
      There is a hole in economics regarding labor share and effective demand. You rarely if ever hear these important concepts. There has not been a clear model to explain their dynamics.
      I am disappointed by the wisdom of many top economists, if not all. Yet, I now find myself developing a new model for labor share and effective demand. The model is raw in its early stages, but some things are coming together, getting clarified... new equations to test. For instance, there appears to be an effective demand limit that blocks further employment of labor and capital... and we are heading toward that limit within the next 16 months. Unemployment will still be around 7% and won't go lower. The theory is going to be tested.
      In short, I am "making a positive case for an alternative.", as you put it.

    2. Lambert, do we hear the "t" or is it like "Colbert"?

      "Positive case" not my words, but I did take a look at one of your graphs.

  4. As a hobbyist, I struggle to understand "the economy".

    Hobbyist, credentialed economist, makes no difference. The thing that really sets people like you (and me) apart from the profession is posing the problem as "understanding 'the economy.'" This is not what most economists think they are doing, at all. Economics as an intellectual activity is basically self-contained. People who are good at it do earn the right to talk in various public settings about "the economy," but they have to use a completely different set of tools to figure out what to say. That's the point here -- Eggertson is a very skilled modeler, but when it comes time to deciding what he thinks about real economies, he just erases the model and writes down something different (and in this case, more sensible.)

    I remember the first economics class I took in college -- maybe you had a similar experience. The field was defined as "the science of how rational beings make choices," not the science that deals with questions like unemployment and interest rates and inflation. Which still seems like a fair description of the vast majority of it.

  5. In other words, it would take an enormous amount of anthropological and historical work to show that studying "economics" could be any help in understanding "the economy" - work which very few "economists" attempt to do.

  6. Academic Economics is a large Mansion of Many Apartments.

    For the Friedmanite micro of "Free to Choose" or the "rent-seeking" politics of Public Choice, there's a powerful moral narrative, and a comfortably familiar framework for generating a rhetorical "analysis", which drives informal thinking about policy questions. I think the positive economics -- the reasoning -- has "objective" value, and has to be misused and miscast to get the moral results, which are touted by ideological partisans. So, a critique does not have to be totally rejectionist to be useful, either in rescuing knowledge or putting a political case.

    The DSGE models of modern macro seem to me to be a different case. They are so highly stylized, it is hard to imagine that anyone learns anything useful from them. Noah Smith tries pretty hard to give an accurate impression and appraisal, but he still gets sucked into making them sound better than they are. Friedman never left anyone in doubt about what he thought his model of a competitive market economy seeking equilibrium "meant". I think anyone, who looks as a complex DSGE model has to wonder what it is supposed to mean, at the most basic level. I've seen some of the output of DSGE forecasting models, and I wouldn't think to do what Noah did, which was to make a table evaluating their fit and accuracy, as if their stylization left them with such qualities. I think anyone would struggle, and struggle hard, to identify points of correspondence between the model and an actual economy. "What does this mean?", "What could this mean?" would echo over and over.

    (My own epistemology does not lead me to think such a correlation is to be expected from analytic models. I think one needs operational models for that. The distinction between analytic and operational models doesn't seem to be a familiar concept to many economists.)

    The actual economy is very complex, and we are seeking insight into its fundamental nature, which insights are simple enough for us to use and communicate. Maybe, it is an inherently hopeless task, in an arena of interested controversy. But, even so, this DSGE business seems bizarre, like it is intended to induce a kind of schizophrenia, in which the "practical" views are uninformed by formal analysis. And, it is not without consequences. It is hard to imagine serious people would be touting "nGDP targeting", if they hadn't had the equivalent of a frontal lobotomy.

    I've looked at Michael Woodford's big book, and I don't understand much of it, but the parts I do understand -- mostly discursive footnotes and asides and framing -- leave me suspicious that he is a complete ignoramus, as far as the actual economy is concerned. Not at all a political ideologue of any stripe, but also not inclined to engage with genuine puzzles, concerning how the epiphenomenon of aggregate behaviors emerge from, say, processes of price formation (and vice versa, in say the costs of inflation).

    It is not just that policy thinking is divorced from the kind of thinking that qualifies our most prestigious "experts", but that the "experts" may be made more ignorant and ill-equipped to think about practical policy questions, by their investment in technique and erudite model-building.

    I hate the kind of gassy emanations from theory to policy, that Barro intended to generate with his Ricardian Equivalence, but, at least, there was some correspondence, some accessible reasoning, and the ideological purpose was clear and straightforward, however misguided and reprehensible. But, this stuff is just weird. At best -- in Eggertson's recasting of Krugman's ideas, for example -- it is empty: tendentious to the point of relying on tautology; at worst, it's Alice beyond the Looking Glass.

    1. Hey Bruce, what's your defense of rent-seeking again? I sort of remember it from CT comments.

    2. Economic rent -- the margin between the income a factor generates in a particular use and the minimum income it has to generate to prevent being re-allocated to some other use -- creates the risk-attenuating structure around which we organize production and distribution activity. In a world of genuine uncertainty, securing a source of economic rent is a prerequisite to building hierarchical organization for controlling production processes and achieving technical efficiency and dynamic progress. Possession of a resource earning rents in a particular scheme of organization gives the organization structural stability and a locus on which to focus residual income, after delivering income to the other assembled factor resources.

      The political problem, imho, is maintaining conflict between firms, each in possession of rents from differing sources or embedded in differing organization strategies. Having insurance companies and savings and loans and commercial banks, each independently in possession of some defining source of economic rent isn't a problem for the political system, because politicians can play one off against the other, and isn't a problem for the economy, because there's enough competition at the margins between firms, to drive achievement of technical efficiency. Problems arise from consolidation of strategic and financial control, which eliminate conflict and consolidate rents away from the nexus of technical control. The Universal Bank or the Media Conglomerate. Then, the politicians are captured, and the product turns hackish.

    3. Thanks, Bruce. I thought there was something to do with increasing returns too.

  7. I think that this post of Krugman:

    fits this post nicely (though I suppose is more about old keynesians).

    For the record I believe the AS-AD stuff is quite stupid, because it is based on an aggregation problem:
    If you have a supply demand curve of some stuff, say shoes, and the price of shoes falls, shoe producers will lose profits, and will therefore stop producing shoes and start producing other stuff, say laptops, that give him bigger profits.
    So the fall in supply due to low prices comes through investors going into different markets.
    But, when you speak of the whole economy, defined as a closed system, investors have nowhere else to go, so how can a fall in prices cause a fall in supply? If they stop investing they either start consuming (thus increasing the AD), or save (but they have to lend to someone to save in a modern economy, and that someone either consumes or invests), or buy some particular good as a store of value, but this too would show as consumption in this kind of model (like people buyng jewels).

    Plus, what is exactly the "price level" in aggregate? Prices are usually relative to something else. The only kind of constant imput I can think of is labor, so you can speak of the price level as "how much expensive is stuff relative to an hour of work's price" but I don't think this is what "price level" is usually supposed to mean.

    1. I just read the Krugman as-ad story wonkish post, and came here instantly, with the same thought -- that there's a nice fit, though, maybe, one that Josh can articulate far better than I.

      One thread of that "fit" is the extremely crooked, disjointed path, by which economists are expected to arrive at their beliefs, leaping from analytic model to analytic model -- rather like trying to cross deep water treading on lily pads. Krugman references this in a couple of telling ways. First, there's this:
      . . . we do want, somewhere along the way, to get across the notion of the self-correcting economy, . . . that we also have a tendency to return to full employment via price flexibility . . . that money is neutral in the long run.
      then he allows at the end, that the right way to think about the present state of the U.S. economy is: (wait for it)
      IS curve plus Taylor rule plus Phillips curve

      Seen from Mars, this has to qualify as a very odd way to do economics. It is not at all obvious that the economy is self-correcting through price flexibility, or that it should be. Indoctrinating students that it is self-correcting in a metaphysical "long run" is tantamount to prescribing religious belief.

      I presume Krugman is using AS / AD as a metonym for the whole neoclassical synthesis, first created by Hansen and Samuelson to reconcile the classical model of a self-correcting economy to Keynesian ideas. Historically, the neoclassical synthesis was a first step, and the Lucas Critique a critical leap, in a long retreat from the Keynesian insight that, maybe, the economy is not self-correcting through price flexibility.

      In the classical model, where money is conceived of, as something the quantity of which matters (gold!), then the price level adjusts to allow a given "quantity" of money to be sufficient to permit all desired transactions and investments. The price level falls until the available "quantity" of money is enough, given the "velocity" of transactions necessary for full-employment. The neoclassical synthesis tries to arrive at a compromise between the classical vision of a market economy seeking general equilibrium through price formation and price adjustment, and Keynesian ideas about how things go completely wrong.

      For the neoclassical synthesis, the focal point of compromise was "sticky" prices: that the economy adjusted prices in the direction of full-employment equilibrium was conceded to the fans of classical thinking, but shortfalls in the achievement of full-employment were attributed to lags in price adjustment.

      Like a lot of political compromises, arrived at in terms of vague and abstract principles, the neoclassical synthesis has been subject to a series of revised interpretations. Krugman's "IS curve plus Taylor rule plus Phillips curve" partially traces the path of those revisions. One of the most important revisions was dumping the "quantity" theory of money, in partial acknowledgement of the reality of Central Bank credit policy. The other revision was, of course, the attempt to use "expectations" to resurrect the Classical Model, in all its uncompromised glory, in the form of methodologically orthodox DSGE models. "Sticky prices" came along for the ride, though, as a simple kludge, to, again, reconcile the classical fantasy with reality.

      Without the hydraulics of a "quantity" money, the AS / AD model doesn't make a lot of sense, though it is explained now with a lot of handwaving about lags in price adjustment, aka sticky prices.

    2. The degree to which the economy as a whole "self-corrects" through processes of price formation is, ultimately, an empirical matter, and in a world, which was serious about doing economics, theorists would try to frame analysis in ways that made the issue, measurable: just how efficient is money, the financial system, and price formation? How do we observe that degree of efficiency? But, alas!

      The focus of (political) compromise (and, therefore, the repressed dispute) in the neoclassical synthesis was price formation/adjustment: price formation doesn't do the job of self-correcting for the vagaries of money (monetary, fiscal policy, and financial regulation, or lack thereof), or the vagaries of real developments (innovation! productivity shocks!), because prices (wages, particularly, for reasons rarely articulated) are "sticky", and there are lags in price adjustments and supply and demand responses to those price adjustments.

      The nature of "sticky" is rarely the subject of serious inquiry.* (My earlier complaint about Woodford is based on my impression that he doesn't have even an elementary idea of actual price formation processes.) Austerity, among economists at least, rests on the idea that the institutions that make prices (wages, particularly) "sticky" are bad, and destroying those institutions, by requiring large deflationary price adjustments, is a desirable form of economic reform. Opponents of austerity, like Krugman, rarely acknowledge that this is a pro-austerity argument, which is an indication of the deleterious consequences of how economists have arranged to talk past each other.

      I don't know if Krugman, in his heart, genuinely believes that "money is neutral in the long-run" as a kind of religious tenet; if he takes the Brad DeLong line that money isn't neutral, but it ought to be, and good policy would make it so; or if he just wants to reassure his colleagues, after his dispute with Rogoff and Reinhart, that he's not going rogue.

      *There are a variety of "channels" by which price adjustment, which is less than instantaneous, might cause macro disturbance, which would go way beyond the sort of "slow adjustment in the 'right' direction" that seems to be the implicit consensus view of economists. You can have race conditions, in which the lag sets off uncontrolled spirals. You can create debt-deflation scenarios, and control crises from changing leverage ratios.

  8. That Krugman post was appalling. I've never seen him (or almost anyone) so baldly admit to being an ideologist, a propagandist for capitalism.

    What he says, in so many words, is that he knows AS-AD is wrong, but he teaches it anyway. Because he wants students to come out of his class thinking that capitalism is always self-equilibrating in the long run, that full employment always returns on its own, and that flexible prices are key to making this happen. Even if they are not being given any legitimate basis for this belief.

    More replies shortly.

    1. OK, I'm glad it wasn't just me misunderstanding him. I literally could not believe what I was reading.

    2. He's always been a wimpy liberal, but his saving grace is that he is empirically minded.

  9. I think thet the problem is this:
    Keynesian economics (at least Krugman's version of it) teaches wildly different models for the "long run" and the "short run". The models are so different as to be not even comparable, and as a consequence there is no idea on how to pass from the "short run" analysis to the "long run" analysis.
    In fact, nobody has the slightest idea of how long is the long run.

    The error comes from the fact that the terms "long run" and "short run" (or short, medium and long term) apparently refer to time horizons, but in reality refer to models that take in account different variables:
    - the "short run" is a model of the economy that takes in account cyclical, financial and distributional effects, but not technological improvements, that usually are very slow;
    - the "long run" is a model of the economy that only takes in account the slow technological improvements that cause a secular increase in productivity and hence in total economic output, but ignores all other aspects.

    This way it is possible to think of an economy that has "long run growth" (slow tech improvement) while remaining permanently "short run depressed", since the short run isn't actually short.

    But since Krugman thinks of the "short run" as an actual (if vague) time horizon, he is forced to think that somehow sometime the short run will revert to the long run, so he is forced to use those concepts like "money is neutral in the long run" that in reality have no relevance to the short run/long run dichotomy if you see it as a dicothomy between models and not time horizons.

    This then leads to the idea that any recession/depression will heal "naturally", though too slowly, which in pratice means to become a "propagandist for capitalism".

    1. RL-

      I think you are absolutely right, except that the situation is actually slightly worse than you suggest. There are actually three, not two, incompatible analyses. There is the short run, when the interest rate is set by the central bank and the capital stock is fixed (or at least exogenous). And there is the long run, when the interest rate is (somehow) at its natural rate, and the capital stock is optimized by private savings decision. (There is no independent investment function in the long run.) But this long run does NOT incorporate technological change. In the standard growth models that are the core of modern macro, some fixed rate of technological change is assumed -- you are just given a technology parameter A. There is no effort to explain it at all. Which is kind of hilarious when you think about it, since the main lesson of these models is that technological change contributes much more to growth (and therefore to wellbeing) than getting the interest rate right does. So the actual conclusion of these models is that the whole body of economics, with its focus on allocation, is really a diversion from the important question. And yet this doesn't seem to bother economists at all.

      Now, there is a large body of "endogenous growth theory" that tries to explain the value of A. But this work depends on a whole different set of assumptions than both the "short run" and the "long run" analysis. In particular, there's a lot of emphasis on "external economies" whereby investment produces productivity gains for other firms. Now, that might be a fine way of modeling some of what happens in the process of growth. But if growth is the most important question, and if externalities are the key to growth, and externalities (by definition) are not correctly priced by markets -- well, then why on earth are you teaching students about an economy that "self-corrects" onto the optimal path via market prices?

    2. Just to be clear -- you are totally right that there is no story for how a bunch of short runs add up to the long run, except via active policy by the central bank. And even then, it's not clear how or why the central bank always gets the economy to the right long-run path. My point is that even if you solved the problem, you'd still leave unanswered what mainstream economists themselves say is the most important question.

  10. For the Friedmanite micro of "Free to Choose" or the "rent-seeking" politics of Public Choice, there's a powerful moral narrative, and a comfortably familiar framework for generating a rhetorical "analysis", which drives informal thinking about policy questions. I think the positive economics -- the reasoning -- has "objective" value


    I agree, from today's vantage point, Friedman doesn't look so bad, at least methodologically. My teacher Bob Pollin always says, the aspiration for left economists should be to write an equivalent A Monetary History of the United States. I think that's right. And I think the reason that book was so effective, was precisely because it was narrative. Also because, as you say, there were some genuine facts about the world (the world of the book; I don't think the idea of "money" as something that has a quantity applies to the credit-money world of today), linked organically to a clear compelling moral-political stance. There is something to learn from that, substantively and rhetorically, in a way that there is not from modern macro.

    I can't speak to Woodford, but I suspect you are right. It's certainly the case that getting a graduate eduction today, and acquiring prestige in the profession, don't require knowing anything about the concrete institutions of capitalism. Altho! I do have to argue the other side a bit too. There are lots of academic economists, including quite mainstream ones at prestigious departments, who actually do know quite a lot about the real economy. It's just they do it on their own time.

    this stuff is just weird. At best -- in Eggertson's recasting of Krugman's ideas, for example -- it is empty: tendentious to the point of relying on tautology; at worst, it's Alice beyond the Looking Glass.


    1. I think your teacher is right.

      I really wish someone would update and expand JKG's "Money: Whence It Came, Where It Went."

  11. There are a variety of "channels" by which price adjustment, which is less than instantaneous, might cause macro disturbance, which would go way beyond the sort of "slow adjustment in the 'right' direction" that seems to be the implicit consensus view of economists. You can have race conditions, in which the lag sets off uncontrolled spirals. You can create debt-deflation scenarios, and control crises from changing leverage ratios.


    I think the general point - which I get from Leijonhufvud, but which you can find in lots of old Keynsian and structuralist writers, like Stephen Marglin -- is that the question isn't do prices adjust quickly or slowly in an absolute sense, but how quickly do they adjust relative to other adjustments.

    Say I reduce my expenditure, for whatever reason. This has two effects in the markets I purchase in. First, sellers reduce their prices, increasing expenditure by other buyers. And second, sellers experience a fall in their income, reducing their expenditure in other markets. The first effect pushes the system back toward equilibrium, the second away from it. So the question is not, does the first adjustment happen at all, that is, are prices "sticky" in the abstract? The question is, does the first adjustment happen so much faster than the second one that we can ignore the second? In other words, "sticky prices" is equivalent to "flexible incomes," and conversely. So there's no reason to say that sticky prices is some special assumption. We have two processes happening in time, and there's no logical reason to say one will always be infinitely faster than the other.

    The Permanent Income Hypothesis plus no credit constraints solves this problem, by making incomes completely sticky. The more interesting solution is Leijonhufvud's: for small disturbances, price adjustments dominate, but for large disturbances, income adjustments do.