Wednesday, July 17, 2013

Krugman Deconstructed

Paul Krugman seems worried that Noah Smith has been bitten by a zombie, or otherwise thrown into a discombobulated state. Noah's sin is to have suggested that what is going on in a New Keynesian model may not shed much light on what is going on in Japan. The basic Woodford framework confines attention to sticky prices as a friction that creates inefficiency that can be corrected by monetary policy. Monopolistically competitive firms can change prices infrequently - say, through a Calvo pricing mechanism - and unanticipated shocks can lead to relative price distortions that monetary policy can act to mitigate. But, those relative price distortions will disappear in the long run as prices adjust. Noah seems to think that 20 years (in the Japanese case) gets us pretty close to full adjustment, and I agree. Indeed, with the frequency of price changes we observe in the data (e.g. Klenow et al.), the same argument would seem to apply to our current predicament. It has been five years since the shock hit that produced the last recession in the United States. If we think that the current state of the US economy is somehow inefficient - i.e. there is some set of policies that could improve matters - it seems we should not be thinking about sticky prices as the source of that inefficiency.

But in another sense, Noah is incorrect:
This New Keynesian model corresponds nicely to the simple AD-AS model that people learn in Econ 102.
You can forgive Noah for thinking this, as a lot of effort on the part of Woodford (in his book and published papers) and other New Keynesians has has gone into making it appear that New Keynesian economics is somehow the same thing as AD-AS, or in fact an excuse for taking that shortcut. The basic New Keynesian model is actually a real business cycle model with Dixit-Stiglitz monopolistic competitors and sticky prices, which owes more to Ed Prescott than to Keynes. Without the word "Keynesian" attached to it, Keynes would not recognize the structure as having anything to do with him, and it's doubtful the first person who put an AD-AS model into an undergraduate textbook would recognize it either.

Krugman is certainly sold on the idea that AD-AS and Woodford macroeconomics are the same thing. According to Krugman, we can do really well with AD-AS, though sometimes we might have to "check" what we are doing by going to the full-blown New Keynesian model. In the case of Noah's post, Krugman thinks Noah's analysis is flawed because of his failure to take account of the liquidity trap, and otherwise falling short in adhering to Krugman's talking points.

Krugman has two models he likes to think about. One is AD-AS, and the other is in his paper with Eggertsson. But he wants to think of those as the same thing. So let's take the AD-AS route, as that's easier. What is Krugman trying to say? Set the wayback machine, Sherman, and off we go. I'm now imagining myself in a macro class in 1975, where I learned what was in the first edition of Branson's book. This is something like playing with an 8-track player you found in your basement. The thing works - in the sense that you can make it do what it was built to do - but you know while playing with it that you are wasting your time.

Figure 1 shows you the IS-LM, AD-AS part of the model. This is the liquidity trap case. The LM curve is flat. Equilibrium output Y0 is less than full employment output YF. Why is the AD curve vertical? That's because changes in prices don't do anything in the IS-LM portion of the diagram, as real money demand is perfectly elastic at the zero lower bound. So, in the absence of some kind of intervention, this economy remains stuck below full employment forever. The fix is fiscal policy - increase G, shift the IS curve, and we go to full employment output.

But, as I was taught in my intermediate macro class, Pigou had a fix for this. If there's a wealth effect on consumption, for example, then you get a downward-sloping AD curve, as in Figure 2. In the short run, output is less than full employment, but if we let prices adjust, then over time the nominal wage falls, shifting AS to the right, and we get full employment. But the short run fix, as in Figure 1, is fiscal policy, which gets us there faster without the bother of waiting for the price adjustment.

But Figures 1 and 2 are not what Krugman has in mind. He's barking up Kalecki's tree, apparently. Kalecki argued that a deflation would increase the real value of debts denominated in nominal terms, and that the wealth effect would go the other way. This gives you a positively-sloped AD curve, which Eggertsson and Krugman indeed coax out of their New Keynesian model.

The key question is, with the Kalecki effect in there, is the slope of the AD curve greater or less than the AS curve? Let's do it both ways. First in Figure 3, suppose AD is steeper than AS. In this case, long-run adjustment to full employment occurs through an increase in the nominal wage. This occurs because there is an excess demand for labor. But I don't remember Krugman arguing that the labor market is currently too tight (maybe you do?). He must be thinking about the other case.

So, in Figure 4, AS is steeper than AD. In this case, in the long run we have to wait for the nominal wage to fall, shifting AS to the right. Krugman says that wages are really sticky downward, so we don't want to wait that long. What's the short-run cure? Well, the problem is that there is too much aggregate demand. We want to shift AD to the left. Austerity will work nicely. Again, I don't remember Krugman pushing for that, but maybe you do.

Krugman might want to appeal to Eggertsson/Krugman as the ultimate truth with respect to issues he discusses in his blog, but that paper has its own problems. It's a superficial treatment of debt contracts and default, and uses linearization to characterize the zero-lower-bound equilibrium, which is known to be misleading.

These are the basics of Krugman's narrative:

1. Most of the macroeconomics done since 1970 was a waste of time.
2. The exception to (1) is New Keynesian economics.
3. New Keynesian economics and textbook AD-AS are essentially the same thing.
4. Wages and prices are sticky.
5. Wage and price stickiness are the primary source of current inefficiencies in the U.S. economy.
6. Relative to what is efficient, GDP is currently too low, and the unemployment rate is too high.
7. We can correct the inefficiencies that currently exist with an expansion in government expenditure on goods and services.

That narrative is a web of contradictions, and hardly internally consistent science. The AD-AS model somehow persists in textbooks, which causes it to retain its power as a tool for communicating with people who have only a smattering of economics training. Krugman exploits that, and so have New Keynesians, but that behavior doesn't serve the cause of science.

37 comments:

  1. Another part of Krugman's narrative was that more wage flexibility decreases output. I guess this is the case where AD is upward sloping and steeper than AS.

    My question is, could AD possibly be GLOBALLY upward sloping in a micro-founded model? If not, ENOUGH wage flexibility or adjustment would indeed restore full employment.

    In this case, Krugman should be arguing not that wage flexibility is bad but rather that it is bad only if it doesn't change too much.

    ReplyDelete
  2. "My question is, could AD possibly be GLOBALLY upward sloping in a micro-founded model?"

    I don't think you want to think in those terms. If you're interested in the effects of wage flexibility, you can do that in versions of New Keynesian models. Nonlinearities are a problem, which means you can't trust results like those in Eggertsson/Krugman relating to the "paradox of toil," " paradox of flexibility," etc, that come from linear approximations. Here's an example of how to deal with nonlinearities at the zero lower bound in a New Keynesian framework:

    http://economics.sas.upenn.edu/~jesusfv/ZLB.pdf

    ReplyDelete
  3. hey look: runaway inflation... great forecast stevie

    ReplyDelete
    Replies
    1. About as clever a response as those who respond to Krugman by saying "oh, want an Enron advisor."

      Maybe you should actually try to comment on the substance.

      Delete
    2. Except Krugman as an Enron advisor is nothing - some company paid him to sit on some meetings and give some lectures, all behind closed doors. Why should anyone care? He wasn't in management, he wasn't any bit responcible for the mess. He even quit the job two years before the scandal.

      On the other hand Williamson failed massively on a basic prediction macro is supposed to solve. Who are you gonna trust on macro? The Enron advisor, any day please.

      Delete
    3. Stick to the topic. We'll deal with inflation another time.

      Delete
    4. "It is better to be vaguely right than exactly wrong."

      So much about your grossly wrong inflation prediction.

      "When the facts change I change my mind. What do you do, sir?" John Maynard Keynes

      Delete
    5. RE: Anonymous
      When the facts change I change my mind. What do you do, sir?" John Maynard Keynes

      That quote is actually misattributed to Keynes; its was actually Samuelson. It likely persists because it seems to be logically something Keynes would have said given how some of his opinions... evolved... over time.

      Delete
  4. "But I don't remember Krugman arguing that the labor market is currently too tight (maybe you do?). He must be thinking about the other case."

    Surely he must be thinking about the first case. Lower wages would cause deflation which would increase the real value of debts denominated in nominal terms, and that would decrease AD and the demand for labor. Higher wages would cause inflation and have the opposite effect.

    The logic of Krugman's argument is indeed that there is currently an excessive demand for labor that needs to be addressed by forcing wages higher.

    ReplyDelete
  5. Excellent post except I'll take issue with your classification of Krugman's "basic narrative".

    He was rejecting the notion that sticky prices are causing the inefficiencies. He's saying that sticky prices are neither bad nor good in a liquidity trap, because if we deflated, the real value of debt would be higher. So while price rigidity is preventing rebalancing in some areas, in a liquidity trap that's kind of self-defeating due to debt overhang.

    ReplyDelete
    Replies
    1. Well, that may make sense in some model. But not in any of the ones that Krugman thinks about.

      Delete
    2. Yep, Williamson is wrong. Krugman could not cares less about AD-AS (in any intermediate AD-AS model with wage rigidity and price flexibility the former causes unemployment because of supply side effects ... and these hardly matter right now; i.e. the AD-AS is the wrong model at the moment) quoted Keynes who also did not consider wage and price rigidities (in Ch.19 Keynes explicitly argues that more downward wage flexibility would be detrimental) to be the key factor that causes involuntary unemployment. The key cause is to be found on the monetary side. Krugman is so annoyed because this is 80 year old stuff, Macro 101, yet many folks seem to have forgotten the basics of their discipline.

      Delete
    3. "Krugman could not cares less about AD-AS..."

      He really doesn't care, to the point where he writes endlessly about it, for example:

      http://krugman.blogs.nytimes.com/2011/10/09/is-lmentary/

      Of course that's just the "demand" side. For full AD-AS, see Chapter 28 of his intro textbook.

      You may take this as a shock, but you would be hard pressed to find serious practicing macroeconomists who think of this ancient schlock as "the basics of their discipline." Maybe you think that's a problem, but I don't.

      Delete
    4. Given that you predicted inflation and that Krugman predicted Japanese style low levels of inflation it looks like "this ancient schlock" is a better model than yours (whatever model you actually use).

      The saltwater freshwater macro fight is one between internal and external consistency. We all love the elegance and beauty of of internally consistent GE models. But when they perform badly empirically one might have to reconsider how useful they are.

      Delete
  6. Thank you for a very accurate portrayal of "New Keynesian" economics as neither new nor Keynesian.

    ReplyDelete
    Replies
    1. Actually, you could give Woodford et al. credit for doing something new. An odd thing about New Keynesian models is that they were designed as models to guide monetary policy, but the inefficiencies are ones we would more traditionally think of as amenable to fiscal policy fixes. What Woodford describes is an optimal tax/subsidy problem, not a monetary policy problem.

      Delete
    2. The funny thing is that right-wing voodoo monetarists who worry about shyrocketing inflation (and thus do not give a shit about unemployment) and make grossly wrong predictions are not even happy with the fairly classical (as Williamson correctly described, DSGE macro is not Keynesian, it is just a classical model with monopolistic competition and price rigidities) macro mainstream.
      Of course if all you care about is internal consistency and not external consistency (not that DSGE is not biased heavily towards the former) you gotta come up with some radical voodoo monetarist stuff.

      It might be fun to do ... but it has nothing to do with the real world.

      Delete
    3. 1. "right wing." What do you mean by that? I vote for Democrats. As I've stated in the past, the tools of modern macroeconomics are not bludgeons of the right wing. They're just the tools of science.
      2. "voodoo." Remember that I'm a pretty conventional academic. I publish in mainstream journals, and people in the profession seem to think I'm OK. Voodoo is what typically gets thrown around in the blogosphere.
      3. "monetarist." Remember, it's not old monetarism. Read about it here:

      http://econpapers.repec.org/bookchap/eeemonchp/3-02.htm

      Delete
    4. I might have done so if New Keynesian or New Classical models were constructed with an awareness of the implications of the Sonnenschein-Mantel-Debreu theorem for the prospects of deriving macroeconomic models from microeconomic foundations. But they weren't, and therefore what was new in both should never have been attempted in the first place.

      Delete
    5. I can do better than that. Rationality has no implications at all. All economic models are wrong. But those models with rational agents are our only hope for understanding the world we live in, and those wrong models are actually useful. Beats thinking about Hyman Minsky.

      Delete
    6. Keen is, as usual, completely wrong here. SMD doesn't specify that anything must happen, it specifies that with certain utility functions crazy things can happen. Unfortunately for dear old Keen, the utility functions that permit odd-ball dynamics are grossly inconsistent with measurements of wealth effects and elasticities of substitution. Tough break, Keeny-boy, it seems your ignorance just cannot be hidden.

      Delete
    7. "But those models with rational agents are our only hope for understanding the world we live in, and those wrong models are actually useful. Beats thinking about Hyman Minsky."

      Wrong models can not be of any use. If people behave irrationally you better build a model which captures this feature of the real world. At least if you wanna actually do science. If you just wanna play around with some equations that have nothing to do with the real world this is of course your prerogative but it is obviously not science (and I say this as an economist who prefers theoretical to empirical work any day!).

      Delete
    8. "If people behave irrationally..."

      How would you know if they were?

      Delete
    9. Previous Anonymous is clearly not an economist.

      Delete
  7. Steve, a trully excellent post! I took some notes (I must have been awfully bored) on how Krugman-Eggertsson is not just an IS-LM model as Krugman often claims, and how, because it is not, it can be evaluated quantitatively in ways an IS-LM model cannot. In my opinion, some of the model's quantitative predictions don't fare well with the data. I would work on this more if I thought the resulting paper could get published somewhere. But the point is that New-Keynesian models, good or bad, are structured theories that, as such, can be quantitatively evaluated on multiple levels, while the AS-AD (or IS-LM) model amounts to sharing stories around the fireplace over a bottle of Scotch.

    ReplyDelete
    Replies
    1. Exactly. The idea here is that even AS-AD IS-LM doesn't seem to say what Krugman wants it to. But I'm sure you can mess with it to get anything you want. What Lucas, Prescott, etc., did long ago was to put some empirical discipline on macroeconomists, and Woodford, Gali, Gertler, etc. took that to heart. It's straightforward to take New Keynesian theory apart and argue about the measurement. With Krugman-Eggertsson, we at least know what we're arguing about. It's pretty quick and dirty, but it puts something on the table at least. You can see why the standard process of writing, public discussion, peer review, and publishing works well. As opposed to blogging, which is at the level of drunk people sharing stories around (or in) the fireplace.

      Delete
    2. Funny that a guy whose model (whatever model that is) made him believe that QE will lead to inflation claims that Old Keynesian stuff is not empirical. :D

      Seriously, old-school VARs are performing better empirically than New Keynesian models (which are obviously theoretically superior) which is why some central banks use a dual appraoch: an internally consistent DSGE core plus an externally consistent periphery.

      Delete
    3. "...old-school VARs are performing better empirically..."

      At what? If you mean forecasting, I guess that could be the case, but my understanding is that a VAR, as much as any forecasting model, needs serious tweaking before it will make sense. Not sure what you mean by "externally consistent."

      Delete
  8. A model is externally consistent if it matches the data well. Old Keynesian models perform better empirically than New Keynesian ones which are obviously theoretically superior so the trade-off in one's model choice is external vs. internal consistency.

    This is basic stuff and I am surprised that a guy without a PhD has to teach it to a professor.

    ReplyDelete
    Replies
    1. Old Keynesian models perform better empirically? Can you cite the research that shows this? You can't make things up and then act surprised that people are unaware of the made-up facts.

      Delete
    2. Strange assertion given that NK models are chosen precisely to match the data. My only guess is that this nutbag really doesn't know what "fitting the data" means.

      Delete
    3. Funyn to get called a notbug by somebody who, given his obliviousness to basic stuff, seems to have not finished his undergraduate studies.

      First, take a look at what some central banks do, use a core-periphery appraoch.
      Second, structural models naturally lead to a better data fit precisely because they are structural. VARs fit the data better than anything else but are theoretically deficient.

      I am anything but an opponent of the micro foundations approach of modern mainstream macro. Actually I like it tremendously as I cherish theory. But New Keynesian models are fairly useless right now as they cannot describe the current crisis. This is hardly surprising, they do after all lack a financial sector (Stiglitz is not the only economist who has pointed out the questionable usefulness of general equilibrium macro models that lack finance and while the dozen papers or so he wrote in the nineties was a nice attempt to get away model money modelled via liquidity preferences it was one of these research projects which ended before it ever really started).

      Delete
    4. This is perhaps the most incomprehensible string of misplaced and misspelled words I've seen in quite some time. Bravo, sir, you've merely reinforced my existing belief that you are a nutbag (you may also be a notbug, but I can't be sure since I don't have any idea what that is).

      And the idea that a structural model somehow fits better because it is structural exposes a deep misunderstanding -- how can something restricted fit better than something unrestricted? It can't, you nitwit.

      Now run back to your undergrad advisor and admit you've been pretending to be an economist on the Internet, and he'll probably ask you what the Internet is since they don't have it at the two-direction state school you're attending.

      Delete
    5. Nice try but I won't bait. You clearly have no idea at all about basic macro or basic econometrics.
      I like that you revealed your disdain for public tertiary education (pretty widespread on this side of the big pond) though, it shows that economic ignorance is highly correlated with politics/ideology.

      Delete
  9. Hi Stephen,

    Good post. I think this is what Krugman has in mind (see Figure 2). What do you think?

    http://web.mit.edu/krugman/www/deflator.html

    ReplyDelete
    Replies
    1. Yes, that's Figure 1 above. But that's not his current story. He's added to it.

      Delete