Saturday, January 19, 2013

More new-Keynesian paradoxes

Last week I saw Johannes Wieland's paper "Are negative supply shocks expansionary at the zero lower bound?"  A side benefit of the job market season is that we see interesting new papers like this one, and it contributed to my project of trying to better understand new-Keynesian models.

Though starting academic papers with blog quotations is usually a bad idea, Johannes starts with a great and very appropriate one,
As some of us keep trying to point out, the United States is in a liquidity trap: [...] This puts us in a world of topsy-turvy, in which many of the usual rules of economics cease to hold. Thrift leads to lower investment; wage cuts reduce employment; even higher productivity can be a bad thing. And the broken windows fallacy ceases to be a fallacy: something that forces firms to replace capital, even if that something seemingly makes them poorer, can stimulate spending and raise employment.” -Paul Krugman
I endorse this quote, because it is an accurate and pithy description of the properties of many careful new-Keynesian analyses in the academic literature.

 Johannes explains
Does destroying productive capacity raise output when the zero lower bound (ZLB) binds? [ZLB: When interest rates are zero, the Fed can't lower them any more in response to shocks -JC] While this question may seem absurd, in fact it is a common prediction of many macroeconomic models: In these models, temporary negative supply shocks raise inflation expectations and lower expected real interest rates at the ZLB, which stimulates consumption and output. While some prominent economists have subscribed to this view and its policy implications (e.g., Eggertsson and Woodford [2003], Eggertsson and Krugman [2011], Eggertsson [2012]), there is wide disagreement over such a radical and unintuitive proposition.
Indeed there is.

These are just the beginning of the strange predictions new-Keynesian models (or modelers) make.

"Fiscal stimulus" is the prediction that even completely wasted government spending is good for the economy. Paul Krugman recommended, with refreshing clarity, that the US government fake an alien invasion so we could spend trillions of dollars building useless defenses. (I'm not exactly sure why he does not call for real defense spending. After all, if building aircraft carriers saved the economy in 1941, and defenses against imaginary aliens would save the economy in 2013, it's not clear why real aircraft carriers have the opposite effect. But I'm still working on the nuances of new-Keynesianism, so I'll let him explain the difference. I'm not a big fan of huge defense spending anyway.)

Furthermore, all the new-Keynesian models are "Ricardian." They predict the same stimulus whether spending is financed by borrowing or by lump-sum taxes  today. Good, we don't need to argue about "Ricardian equivalence," but to believe their predictions for spending borrowed money, you have to believe that taxing you and me a trillion dollars and spending it on a trillion dollars of alien defenses will raise overall output by 2, 3, or 4 (you can get really big multipliers in these models) trillion dollars.

Actually, stimulus financed by temporary payroll taxes can be even better than from borrowing money. These are a negative supply shock, which causes inflation and lowers the real interest rate. Sand in the gears is good. Stimulus financed by temporary consumption taxes is worse, because that encourages saving. Promises of higher future consumption taxes, anathema in the standard view of the world, are good, as they get people to consume today.

Super-weirdly, many new-Keyensian paradox predictions get worse as the central friction, price stickiness, gets better.

Johannes again on the new-Keynesian paradoxes:
First, according to the “Paradox of Thrift,” a rise in the desire to save is self-defeating at the ZLB, because it reduces output so much that aggregate savings fall (Keynes [1936], Krugman [1998], Eggertsson and Woodford [2003], and Christiano [2004]). Second, according to the “Paradox of Flexibility,” output volatility may rise at the ZLB when prices and wages are more flexible (e.g., Werning [2011], Eggertsson and Krugman [2011]).
My empirical results concern primarily the “Paradox of Toil” (Eggertsson [2010]), whereby a temporary increase in desired labor supply at the ZLB reduces the equilibrium employment level in standard models. .... Following this logic, payroll tax cuts are contractionary at the ZLB because they lower expected inflation (Eggertsson [2011]), and allowing collusion among firms is expansionary because it raises expected inflation (Eggertsson [2012]).
A pause in praise of economic models: They tie ideas together. You can't pick and choose. If you like stimulus with borrowed money, but suspect that tax-financed stimulus might not work so well, you can't just waive your hands and refer to new-Keynesian models to defend you. These models predict the two policies have the same effect. If you like your stimulus, but think that maybe hurricanes wiping out a bunch of the capital stock isn't great, sorry, you can't refer to new-Keynesian models to defend you. If you don't buy one of Krugman's assertions, you don't buy any of them. (At a minimum, you have to build a new variant of model -- you can't refer to existing new-Keynesian models to defend you.) To taste fish, you have to swallow the whole whale, hook, line, and sinker.

So, back to Johannes. He notes that the models predict quite different behavior away from the bound than at the bound, so conventional estimates don't really tell us that much about whether these predictions are true. But we have enough experience with economies at the lower bound now, that we can begin to test some of these astonishing predictions.

(Minor suggestion for PhD students. The key requirement for these predictions is that the Fed does not change the nominal interest rate in response to shocks. There have actually been other periods of time when central banks have fixed nominal interest rates, for example between 1945 and 1952 in the US. More generally, the general new-Keynesian view is that interest rates did not respond enough to shocks before 1980. So in fact, versions of the paradoxes should be visible in data away from the zero bound.)

Johannes looks at the earthquake in Japan, and oil price shocks. Surprise, surprise, earthquakes are bad for output. More subtly, the new-Keynesian prediction flows through inflation: "Supply shocks" should raise expected inflation, which lowers real interest rates, and lower real interest rates should raise consumption and output. (As I explained last week, new-Keynesian models anchor expected consumption in the far off future. Then real interest rates determine the growth rate of consumption, and higher growth means a lower level today. In the models consumption=output. See Johannes' equation 2 page 7.) Johannes finds that the supply shocks led to higher expected inflation, and hence a lower real rate. But the lower real rate just didn't have the predicted effect on output. In fact, he finds that oil shocks have worse negative effects on employment at the zero bound than in normal times!

Like all provocative empirical work, I'm sure this one will be picked over. The Booth Macroeconomics workshop did its usual good job of exploring nooks and crannies. But let us also pause in praise of serious empirical work. Rather than blurt "this is ridicuous!" let us go see if indeed earthquakes, hurricanes, labor market restrictions, oligopolization and other normally adverse "supply" shocks actually help the economy. The sun might just come up in the West at the zero bound.

Where to go from here? If I had this great introduction, and results that rather decisively reject a central night-is-day new-Keynesian proposition, clearly linked to all the others, I would obviously have been tempted to write it up as "this model is wrong," and dig deep into which key assumptions of the model drive its basic mistakes. Johannes takes another tack, and adds credit constraints to the model. Whether this is a successful repair or a clever epicycle I will leave for another time -- and frankly I haven't studied it closely enough to opine yet.  How many of the paradoxes it overturns is another good question. It seems to overturn quite a few. But the paradoxes are also the sexy policy implications.  It may save new-Keynesian models from their prediction that hurricanes are good, by destroying the new-Keynesian multiplier.


  1. John, Johannes should take comfort from the findings by Anton Braun et al. (2012) who show the bizarre results (e.g. paradox of toil) found in New Keynesian models are an artifact of log-linearizing at the ZLB. Here is their abstract:

    "Does fiscal policy have qualitatively different effects on the economy in a liquidity trap? We analyze a nonlinear stochastic New Keynesian model and compare the true and loglinearized equilibria. Using the loglinearized equilibrium conditions, the answer to the above question is yes. However, for the true nonlinear model, the answer is no. For a broad range of empirically relevant parameterizations, labor falls in response to a tax cut in the loglinearized economy but rises in the true economy. While the government purchase multiplier is above two in the loglinearized economy it is about one in the true economy."

  2. John Cochrane: "If I had ... results that rather decisively reject a central night-is-day new-Keynesian proposition, ... I would obviously have been tempted to write it up as "this model is wrong," and dig deep into which key assumptions of the model drive its basic mistakes."

    Consider the type of science which can predict what will happen, real science. It is not primarily up to the reviewer (John Cochrane) to find the evidence which supports and contradicts the proposed theory. In real science, the theorist examines all of that, especially the contradictory evidence.

    Cargo Cult Science
    1974 by Richard P. Feynman - Commencement speech at CalTech

    The late particle physicist Richard Feynman was a plain-spoken genius in his field. This speech considers why we continue to not know the truth about many things, hundreds of years after people discovered how to do good science. An enjoyable must-read.

    === ===
    [edited]   Details that could throw doubt on your interpretation must be given, if you know them. You must do the best you can to explain them, if you know anything at all wrong or possibly wrong.

    If you make a theory, for example, and advertise it, or put it out, then you must also put down all the facts that disagree with it, as well as those that agree with it. There is also a more subtle problem.

    When you have put a lot of ideas together to make an elaborate theory, you want to make sure, when explaining what it fits, that those things it fits are not just the things that gave you the idea for the theory; but that the finished theory makes something else come out right, in addition.
    === ===

    Prediction is everything, and it must work more than once. Explaining everything after the fact is merely making up complicated stories.

    That is what this complicated theorizing of New-Keynesianism seems to be. And, old Keynesianism. It is a lot of story-telling combined with math models which have not been shown by experience to predict anything.

    Then, these stories are presented without being tested against the known supporting and contradictory evidence. It isn't science, and it is not reliable. Yet, it is used to justify massive experiments on the lives of the peasants.

  3. John, I am curious how do you explain the following

    My guess is that your anti-demand and anti-stimulus story is based on an implicit assumption you make which says that private sector will make all optimal investments at all times, including at zlb and including right after a major financial crisis.

  4. You are making the same mistake as the FED made in the 70's. Supply shocks like oil price shocks raise inflation, but that's different from inflation due to the economic cycle (overheating). The same is true for inflation expectations. Note that Krugman speaks of "something that forces firms to replace capital". Oil price shocks don't do that, quite to the contrary. Now you may think that Japan's earthquake and tsunami does fit the bill. But that is not completely right, because a very important effect was that they had to shut down all their nuclear power plant, a big effect for a country like Japan, and was forced to import lots of oil. This means the price of resources for energy production has increased considerably. This is basically the same as a massive oil price shock. So I don't think that New Keynesians, like Krugman, are talking about supply shock *in general*, but about *certain types* of supply shocks. Namely, those that force firms to replace capital.

  5. "In the models consumption=output."

    But in reality output consists of investment, export, and import, other than consumption. If consumption rises when supply is constrained, import should rise, which leads to decline in GDP by definition. And negative supply shock of course leads to decline in investment and export. Overall decline in GDP means decline in income, which leads to decline in consumption.
    In this entangled reality, one easy test of new-Keynesian models is to see whether the decline of consumption is smaller than that of GDP or IP. According to the paper, it seems that is actually the case.

  6. Show me a DSGE macro model of any sort and I'll show you a model that makes some pretty bizarre predictions for certain variables under certain conditions.

    And if you're serious that, "To taste fish, you have to swallow the whole whale, hook, line, and sinker," then no one has any business eating fish, period.

  7. John, it shouldn't surprise you that a model gives opposite results from "conventional" ones, when it at the ZLB changes the slope of the AD curve. All these "paradoxes" more or less come from two assumptions: 1) At positive interest rates, the central bank mechanically follows an active Taylor rule leading to a negatively sloped AD curve (which together with a positively sloped AS curve lead to conventional effects of supply shocks). 2) At zero interest rates, researchers mostly focus on the determinate equilibrium (linearized around the positive interest-rate steady state of the two in existence), which is mathematically equivalent with focusing on the case of a positively sloped AD curve. And one which is _steeper_ than the AS curve. Presto! "Paradoxes" and big spending multipliers emerge.

    A nice paper that explicitly acknowledges the existence of two steady states, and shows that fiscal policy multiplier may be smallest at the ZLB, is Mertens and Ravn (2012): "Fiscal Policy in an Expectations Driven Liquidity Trap". (they look at the indeterminate equilibrium where the AD curve becomes positive, but flatter than the AS curve: Negative "multipliers" cannot be ruled out).

    So even within New Keynesian models types, work are so diverse that one cannot talk about "The New Keynesian Model" anymore as I see it.

    Kind regards, Henrik Jensen (University of Copenhagen)

  8. John, I would be very interested in your explaining how, at the ZLB, an increase in private savings will lead to an increase in investment?

    In particular, I would further be interested in your explanation of why businesses would increase investment when demand is falling?

    And, I would appreciate your explaining why Buffett and Munger are wrong in their explanation that businesses invest where, put simply, they either have or can create "moats," which has nothing to do with savings whatsover. I.e., if no moat is possible, a business ought not to invest.

    1. In certain versions of Keynesian accounting, savings is defined as investment. Presuming that you are John D/Hamilton, I think you might have commented on the dust-up between Sumner and Wren-Lewis involving that.

    2. It is not in 'some versions of Keynesian accounting'. It is an aspect of the basic macro model, i.e. the national accounts. For starters, all our macro definitions (investments, households, non-financial companies, MFI's, whatever) are based upon these very accounts which are defined much better than any other macro model and which are also fully estimated, unlike any other macro model. Subsequently, these accounts are, unlike DSGE or new Keynesian models, complete and show the ENTIRE flow of consumption, final expenditure, investment, exports, income, whatever. The phrase 'savings' is however used in multiple ways, alas. An example how this works is the USA income account for december 2012, which shows a large increase in income and a whopping increase in the savings rate, and the production accounts for the fourth quarter, which show a small decline in production. How does this square? National accounts are defined around income generating activities, income being wages, profits, interest income and the like. Spending counts as final spending when it generates income. Investments in real stuff create work and income as people has to produce them. This are macro-economic savings in the national accounting sense. Financial savings, i.e expenditure on existing stuff like stocks or bonds or whatever, does not generate income and is not saving in the national accounting, income generating sense. It's just swapping one kind of asset for another kind of asset. On a micro level this pays off, consider the island with 1.000 coins where one household starts to save coins. This household will assemble wealth. The number of coins will however not increase, the island as a whole won't get richer (the paradox of thrift)- while planting a coconut instead of consuming it will make the island richer. That's saving in the national accounting sense: (Y-C) or the amounts of coconuts produced minus the amount consumed (which makes an increase of the inventory of coconuts an investment, too). We really have to learn to make a distinctin between real saving (producing stuff for use in another period) and financial saving, i.e. swapping assets. Now we come to the paradox of thrift, however, which is about financial savings.

      Merijn Knibbe

  9. It had been my impression that these paradoxes only occurred in a subset of NK models, and not a very representative subset either (I believe Casey Mulligan refers to them as "extreme Keynesian"). I don't recall NKs like Greg Mankiw or John Taylor arguing on behalf of them. I might also cite Quine on the hard "core" of a theory vs the peripheral aspects we are more likely to reject when the model fails to correspond to empirical evidence.

    Still, thanks for pointing out the paper.

  10. Where to go from here? Well, we could look at the data for public spending and GDP growth and ask the question: Does public spending ever increase private sector growth?

    If you do this, if you look at the actual data, it is surprising see: Sydenhams Law of public expenditure and GDP growth

    The answer is NO, furthermore the New Deal is a myth, there was a strong private sector recovery in the USA before the New Deal.

    1. I looked at your argument, and it's very unconvincing. Not that it may be incorrect, just that thousands of economists have argued the same thing before you and done much more rigorous and credible statistical studies (as well as theoretical, for that matter).

      For instance, although you make the claim that "Public spending is almost certainly the causative agent", you don't support it with anything rigorous (despite a few historical scenarios you mention in which you give no depth, and aggregate statistics that essentially don't control or find any causality). I could name multiple recessions (hell, even seasonal adjustments) where the fall in private growth precedes government expenditure growth due to falls in employment and increases in automatic stabilizers.

      So maybe you shouldn't go naming a law after yourself just yet?

  11. John, deliberately misrepresenting Krugman's tongue-in-cheek recommendation puts everything else you say into question. May I suggest you stop doing that?

    1. What didn't you like here? I thought I represented his views with complete accuracy. It was a serious and very insightful example of how new-Keynesian economics works. The equations of the models say that building alien-invasion defenses raise GDP. He honestly said so. For once I thought I was being nice!

    2. Oh, please. He said it would raise GDP. It would.

      That's not at all the same thing as "recommending" it. Obviously there are far better ways to do the same thing -- ones that don't involve deceit and silliness.

      You're committing a basic logical fallacy. Just because an absurd thing would produce X doesn't mean X is absurd.


    "The remarkable thing right now is this:

    I do not know if Cochrane doesn't understand that Krugman's point is that, right now, under these special conditions, expanding government purchases to buy even useless things passes the benefit-cost test so expanding government purchases to buy useful things really is a no brainer.

    I do not know if Cochrane understands this point that considering the consequences of a useless government expenditure is intended to put a lower bound on the benefit-cost ratio, and yet hopes to mislead his readers into not understanding this point.

    I do not know whether Cochrane has thought about the issues at all."

    1. 1. I know new-Keynesian models make this prediction. I think those models are wrong.
      2. I know that some government spending has additional benefits, and have said so many times in print.
      3. There's a lot Brad doesn't know. If he wants to know what I think about something, he can (heavens) ask me before posting what he thinks I know. My email is easy to find. I know that reading what people write or asking them what they think before posting violates the Krugman-DeLong modus operandi.

  13. Krugman has linked to this post, calling our host "remarkably dense", so that's no doubt why JamTheCat and others consider it to be a misrepresentation of Krugman's views.

  14. Krugman's model predictions have been pretty good, while the Austerians have been wildly mistaken. Market efficiency has taken a black eye from the Wall Street shennanigans, but the political alignments of leading economists seem to be unaffected by real world events. Mr. Cochrane has been on the losing side of this debate for a long time.

    1. I let this one through so that regular blog readers can enjoy some of the intelligent, well-informed and tightly-reasoned fan mail filling my inbox this morning. Most of the rest are getting deleted along with the ads for viagra and such.

      Did you notice how Krugman, though shamed now into linking to the piece he's slamming, carefully skates around the question whether he personally read any of it, vs. just cheering Brad on?

  15. "Fiscal stimulus" is the prediction that even completely wasted government spending is good for the economy.

    oh is that the definition of fiscal stimulus

    now i know thanks for that

    1. You're welcome. Actually that's something Krugman and I agree on.

    2. Good for the economy under certain circumstances, you mean. Nobody in their right mind thinks additional government spending is good an economy at (or near, I suppose) full capacity.

  16. Mr. Cochrane: As a German economist, I wonder why American discussions rotate around the ZLB. I think even if one adheres to ISLM, the "interest rate" is that interest rate which must be paid by investors. Otherwise specification of an investment function I(i) would make no sense.

    Now, the US primary bank rate is 3.5 percent. The conventional mortgage rate, relevant to home buyers, stands at 4.5 percent.

    The only interest rate which is in fact close to zero, the federal funds rate, isn't relevant for any real investment decision because only banks can lend at that rate. Hence, why don't you flatly reject the Keynesians's fancy of a ZLB?

    1. I would like to read an answer to this question, it has always puzzled me.

    2. The ZLB does operate in the case of the Prime Rate these days since a flat 3% is added to the Federal Funds Rate to "calculate" the Prime Rate. What has always puzzled me is why there has been no effort to look at this as a way to stimulate private borrowing.

    3. The ZLB claim isn't that interest rates are zero throughout the economy. It's that the Fed has hit the ZLB in terms of its ability to lower interest rates. Normally, countercyclical monetary policy consists of lowering interest rates to spur investment. But when the interest rates controlled by the Fed hit zero, then what? That's the ZLB.

      Plus, the ZLB claim is about risk-free rates. Your 3.5 or 4.5 figures include credit spreads. I don't know offhand exactly the real interest rate on short term U.S. debt right now, but not too long ago, it was slightly negative.

    4. "If you don't buy one of Krugman's assertions, you don't buy any of them. (At a minimum, you have to build a new variant of model -- you can't refer to existing new-Keynesian models to defend you.) To taste fish, you have to swallow the whole whale, hook, line, and sinker. "

      This is an odd assertion -- one that ignores the possibility that an existing model can be roughly correct in most aspects but incomplete or inaccurate in some others. And no, a person with a flawed model doesn't have to shut up completely until the model is fully worked out, as long as he or she acknowledges the limitations.

      Let me give you an example from physics. Both Newton's physical mechanics and Maxwell's model of electromagnetism turned out to be incomplete. And yet, even today, even knowing that they are incomplete, physicists still teach and use those models in certain circumstances -- specifically BECAUSE you don't have to throw out an entire model just because it is wrong in some set of conditions.

    5. My understanding is that the ZLB is important because the Fed can not easily lower the rate. Therefore, the Fed lacks the power to increase consumer spending through the usual method. The specific rate isn't what matters, the relevant point is whether or not the Fed can induce increased spending.

    6. As the originator of this sub-thread, I am grateful for the various answers and wish to put my question more precisely:

      1. Politically, it is clear that the fed cannot reduce the federal funds rate below zero.

      2. But my point is a theoretical one, namely, why *should* the fed stimulate demand at present? This pertains to two sub-questions:

      a) The US have (moderate) inflation and (moderate) growth since mid-2009. Where does the perceived need for further "stimuli" come from? You may point to unemployment but this is a different matter (see Mulligan's redistribution recession).

      b) More importantly, Keynesianism rests on the assumption of a market failure of the following form: If S>I and i=0, excess savings depress Y. But at the moment, we have S=I and i>0 if we take bond yields or mortgage rates at relevant interest rates.

      To sum up, I simply cannot see a market failure in the capital market as long as bond prices are finite and bond yields positive. S>I would mean that savers cannot find bonds in the market, a perfectly absurd imagination. Without such a market failure there is no need to stimulate the economy.

  17. People don't need Krugman to tell them that his views are being misrepresented. They can figure that out for themselves, having read what Krugman has written. By the same token, we don't need Milton Friedman to come back from the dead to tell us that he's not an Austrian.

  18. Can you clarify the following statement?
    "all the new-Keynesian models are "Ricardian." They predict the same stimulus whether spending is financed by borrowing or by lump-sum taxes today."

    I thought that new-Keynesian models considered tax-cuts a type of stimulus. Isn't that why the 2009 stimulus bill contained tax cuts? If tax cuts count as stimulus, then raising taxes would have the opposite effect. Shouldn't new-Keynesian models predict different effects for stimulus spending financed by borrowing or by taxes?

  19. I think the criticism from Delong is aimed at one specific sentence. You wrote, "After all, if building aircraft carriers saved the economy in 1941, and defenses against imaginary aliens would save the economy in 2013, it's not clear why real aircraft carriers have the opposite effect."

    But nobody said that real aircraft carriers have the opposite effect. Wouldn't a build up to defend against a fake alien invasion involve building real defenses?

  20. ["Fiscal stimulus" is the prediction that even completely wasted government spending is good for the economy]

    So obviously true that it's hard to see why it's worth commenting on.

    A minor example from last year:


Comments are welcome. Keep it short, polite, and on topic.

Thanks to a few abusers I am now moderating comments. I welcome thoughtful disagreement. I will block comments with insulting or abusive language. I'm also blocking totally inane comments. Try to make some sense. I am much more likely to allow critical comments if you have the honesty and courage to use your real name.