Monday, July 23, 2012

Jonathan on Keynesian Uncertainty

Jonathan makes some excellent point on Keynesianism, the Austrian school, and uncertainty here. I don't know why some people feel this need to make things up about the other side to make them appear completely naive to less knowledgeable readers. It's not a way to raise the discussion to a higher level.

I think the best and most obvious way to put it is that uncertainty and expectations have been a critical part of both the Keynesian and Austrian visions. Uncertainty has probably been more central to Keynesians longer. It's the core of Keynes - and he contributed, along with Frank Knight, our most sophisticated understanding of uncertainty, as opposed to risk*. Both of these men wrote a lot about how uncertainty was critical for understanding the role of the entrepreneur in the economy. This concern for uncertainty continued implicitly and explicitly through other Keynesians. I think you saw uncertainty enter the Austrian discussion somewhat later with Lachmann, Kirzner, Higgs, and Rothbard (I'm happy to be corrected - I just never hear people citing earlier work on Austrian uncertainty the way they cite Keynes).

As Jonathan points out, concern with expectations has been there from the beginning for both. I think I've noted before that in this sense New Keynesianism is really coming back to the Keynesian core by emphasizing things that the "Old Keynesianism" that came from Hicks and Samuelson glossed over. Part of this is just modeling sophistication. Hicks made a tremendous contribution. We can't expect him to formalize everything in the General Theory. Uncertainty, of course, is notoriously hard to model for reasons that Jeff Friedman outlined in a recent Critical Review article. I disagree with Friedman on the point of whether economists who work with search and risk models ignore these points. I don't think they do at all. But he's right on the modeling difficulties.

So that's a lay of the land on uncertainty and expectations I think. You would be safe to steeply discount anyone who says this isn't central to the way Keynesians think. You also ought to steeply discount anyone who says that of Austrians.

That's not very salacious blogging, I know. "We both think X but we may emphasize somewhat different things about it" is a lot less exciting than "they don't pay attention to X and that's why they're awful". It's less interesting blogging to some people, but it is... you know... accurate.

Jonathan notes:

"Whether or not these economists are talking about uncertainty in their blogs and op-eds is irrelevant, because uncertainty is implicit in their models.  Consider, for instance, Paul Krugman’s work on Japan (“It’s Baaack,” Brookings Papers on Economic Activity 1998, 2 [1998]): expectations, and thus uncertainty, is a major factor behind the advocacy for high inflation targeting.  That they target different causal factors doesn’t make it any less of a use of uncertainty."

I agree with this. A lot of the understanding of uncertainty is implicit. Uncertainty is a big part of why financial crises reduce demand. It's not just an income effect (although the income effect matters too!). Sometimes both of these are subsumed and people just observe that financial crises cause large demand shocks. So I think Jonathan is right that even if we don't see this stuff in blog posts, that doesn't mean it's not there. But as the last four Keynesian Uncertainty posts of mine pointed out - it's in the blog posts, spelled out very clearly! I'm also glad he linked to Krugman's Japan paper, because despite any claims by Krugman to the contrary, he's not just a hydraulic, circular flow Keynesian - and the Japan paper proves that. It discusses uncertainty and long-run expectations on a more sophisticated level than you'll see most Post-Keynesians discussing the issue, certainly.

Jonathan ends with this, and I am intrigued to hear more: "So, when discussing on what causes the differences in policy advocacy between Austrians and the rest, the real answer ought to target the decades (almost a century now) of divergence in understanding of the market process."

Maybe. I'm not quite convinced that's the real difference. I keep hearing that Austrians have a sophisticated understanding of the "market process" that nobody else shares but I've never in my life heard an Austrian say something about the market process that isn't agreed to by most economists (including, of course, Keynesians). Sometimes Austrians may like to provide verbal expositions and they think that the modeling of others is unsophisticated. Sometimes the feeling is mutual, even. Different styles emphasize different things. But I have yet to come across an explanation of the market process or exchange relations from an Austrian that goes beyond emphasizing something I might not talk about as much and actually clashes with my view of thing.

So I'm doubtful. This stuff often veers into that "I'll tell you your position, which I will then proceed disagree with" territory that really bugs me.

I think the differences are a lot less exciting than that. First, we emphasize different processes. Austrians talk a lot about the impact of macro phenomena on the capital structure - an issue Keynesians often ignore in the name of parsimony. Keynesians discuss why liquidity preference implies that the economy is demand-constrained, and that the interest rate is not just the price of loanable funds. Divergences between the interest rate implied by a clearing loanable funds market and one satisfying demand for liquidity can generate swings in economic activity. Austrians often ignore these issues, less in the name of parsimony (because there is less formal modeling among Austrians), and more as a result of the way they think about credit markets. There's nothing "anti-Keynesian" about talking about the capital structure (in fact Keynes agrees with Bohm-Bawerk's capital theory in the General Theory, he just says it's inconsequential for the macroeconomy). Likewise there's nothing "anti-Austrian" about talking about liquidity preference (see Bob Murphy). They just ended up talking about different things.

I think the directions they each took Wicksell is key to understanding the difference between Austrians and Keynesians. You will hear internet Austrians say that Keynes didn't understand Wicksell. Ignore them. They are barking up the wrong tree and probably doing it because they have a chip on their shoulder about Keynesians. Hayek added Cantillon and Ricardo and Bohm-Bawerk to Wicksell and came up with a theory that emphasized certain processes. Keynes added Jevons and Marshall and Malthus to Wicksell and came up with a theory that emphasized certain process. Which Wicksell would have preferred isn't knowable because the man was dead by then. I'd bet he would have liked both.

The point is, pretending that the difference between Austrians and Keynesians as being (1.) political (e.g. - Keynesians are statists and Austrians are the ones that appreciate liberty), (2.) a differential appreciation of something that both demonstrably have an appreciation for (e.g. uncertainty), (3.) a somewhat different way of talking about something we all agree on (e.g., the market process) is all wrong.

The difference, really, is that human economic and social behavior is tremendously complex and there are a lot of processes at work. No one can talk about all of them. Austrians and Keynesians talk about slightly different processes (with at least one common antecedent: Wicksell). There are two questions: (1.) does their theory make sense (in most versions of Austrian and Keynesian economics I think the answer is more or less "yes"), and (2.) is their theory empirically important (I tend to think the Keynesian answer wins out on this one)?

* - This is another example of "making things up" that people do. Some people will only cite Knight on this development and act like Keynesians are clueless or don't understand the role of entrepreneurship and profits. Why? They published their most substantial contributions in the same year. They're both famous books and most people cite them together. The only reasons are genuine ignorance (which is easily corrected) or pettiness (which is less easily corrected).


  1. Alongside Keynes and Knight you have to add Mises. Mises incorporated uncertainty into Austrian economics around the same time as it began appearing in the work of other economists. Mises also differentiates between risk and uncertainty, although influenced by his brother's views on the theory of probability. I don't know who influences Lachmann on uncertainty, but Kirzner was influenced by Mises.

    On different views on the market process, by multiplier versus Ricardo Effect discussion is an example of that. While writing the most I think I might have been distracted, and so I didn't write on the Ricardo Effect, at all. I might go back and edit that in. Other differences include the theory of price formation; Böhm-Bawerk's theory of price formation is not the same as the neoclassical's, and it's not the same as the post Keynesian's. So, there are important differences, and these differences shape the structure of a way of thinking, because more advanced ideas are derived from these principles that are often different between schools of thought.

    1. I would be very interested in hearing more about risk v. uncertainty type distinctions in Mises. I feel like even from Austrians I see Knight cited, not Mises. Very interesting - thanks.

      Also interested in hearing any more thoughts on the market process. I think what some people claim are fundamental differences in understanding the market is - in my opinion - just different ways of modeling it. Economics is not like physics - the model is not reality. The model is a way of talking about reality. Certainly people model these things differently, I'd agree with that.

    2. Dr. Michael Emmett Brady has a paper dealing with John Maynard Keynes's decision theory and the von Mises brothers. However, he has yet to publish it, for he has no desire to do so. Which is a shame, because the way Austrians deal with uncertainty seems to be different with that of the economics of Keynes or Daniel Ellsberg's contributions to decision theory. Dr. Brady's paper could be constructive to the Austrian School with regard to decision theory.

    3. I've been trying to weed out the differences myself -- and you might be right in some respects, since we can't say everything is different.

  2. One unresolved piece of scholarship in the history of economic thought would be why didn't J.M. Keynes cite Frank Hyneman Knight's Risk, Uncertainty, and Profit? See the following review for reference.

    Also Daniel, haven't I told you about Keynes's conventional coefficient of risk and weight, and the fact that Dr. Michael Emmett Brady has published research on it extensively? Uncertainty can be easily modelled based on Keynes's equation in Chapter 26 of A Treatise on Probability. I sent you an e-mail about this a while ago. Perhaps I ought to get Dr. Michael Emmett Brady to do a guest post on Facts and Other Stubborn Things or something...

    1. Cite it in what? The Treatise on Probability was written about the same time, which is why he didn't cite Knight in that. He doesn't deal as directly with these issues. That's not to say uncertainty isn't there - it's just there in a much more applied way, and the scope for citation doesn't seem as obvious to me. Did he not cite Knight in the GT? I don't personally recall.

      Yes, perhaps the coefficient of risk and weight would be useful to incorporate more. It certainly seemed sensible when I read that paper you sent me a while back on it. I still wonder, though, if it really gets at the issue. We all know there are unknown unknowns... kind of funny, but "unknown unknowns" as a category are themselves "known unknowns" insofar as we know such things exist. To deal with that we act with the understanding that unknown unknowns could crop up and we take appropriate cautions. Part of that is discounting what we think we do know. This is, I think, where the coefficient of risk and weight may be useful for talking about uncertainty. But it still doesn't model the unknown unknown itself.

      I have other thoughts on this - maybe we'll get into it more in the future. I think papers like Jeff Friedman's vastly overcomplicate the problem. I made this point extensively in the Coordination Problem comment section and then was mocked by Steve Horwitz for the effort, so I haven't really revisited it. But essentially the way to talk about radical ignorance and radical uncertainty in a model is simply to not model agents to act on the reality of the world. Specify a model of an economy where X holds true and then model your agents such that they have no knowledge of the existence of X whatsoever. Then have the agents act, and bingo - they are going to experience "surprise". They will be acting under radical ignorance.

      Modeling uncertainty in that sense is trivial. The difficulty is figuring out what uncertainty we care about modeling. Presumably an uncertainty we care about is not an uncertainty we know about ourselves.

      We can easily model generic uncertainties, though. The trick is to make sure it's not so generic that it's not useful for generating insights.

    2. I meant "cite it in the General Theory." Keynes does not cite Knight or Schumpeter in the General Theory. Even Dr. Brady thinks that it was a blunder on Keynes's part not to do so.

      As for Keynes's conventional coefficient of risk and weight...I don't know how well it gets at the problem of unknown unknowns, either, but I think Dr. Brady would say that complete ignorance is also a special case, like evidential weight equating to 1. As I have said before, often the weight of evidence is in between a lower bound and an upper bound.

      I think you ought to e-mail your thoughts on the conventional coefficient of risk and weight to Dr. Brady. Perhaps he can help you out.

    3. Is not "animal spirits" just how people frame uncertainty?

  3. Lorenzo: According to Dr. Michael Emmett Brady, optimism and pessimism are incorporated only informally (in the form of "animal spirits") by John Maynard Keynes in The General Theory of Employment, Interest, and Money. It would be twenty-five years after the publication of General Theory and a little more than forty years after the publication of A Treatise on Probability that Daniel Ellsberg would complete his doctoral dissertation, Risk, Ambiguity, and Decision. It was regrettably published decades later after it was completed in 2001. Ellsberg has technicalities that incorporate animal spirits, and he improves on Keynes's decision rule in some respects.

    1. Thanks, that's informative. Though I take the short answer to be "yes".

    2. Yes, the short answer would be yes. For more information, I recommend reading the works of Dr. Michael Emmett Brady. Search "Michael Emmett Brady" on Google Scholar or go to his SSRN account (see link below).

  4. Here's a set of questions to get people thinking about this more....

    * Is the principle effect of uncertainty on money holdings, or is there more to it than that?

    All of us, and the Market Monetarists can agree that uncertainty affects money holdings. Mises wrote about that way back in 1924 in "The Theory of Money and Credit", and I think he borrowed a lot of that discussion from 19th century economists. But the question is: is there more to it than that? Lots of mainstream economists don't seem to think that there is, at least not in practice.

    * Does the amount of risk businesses take vary with the interest rate?
    * Does the amount of risk businesses take vary over without a simple driving factor?

    Austrian economists would say that risk taking can vary with the interest rate. Keynesians would point to "animal spirits", and to Minsky's explanations indicating that risk taking can vary over time without that driver.

    * Does uncertainty over government policy play a significant role?

    As Daniel pointed out earlier, Keynesians don't deny this can happen, they just don't believe that's it's important in practice.

    "But essentially the way to talk about radical ignorance and radical uncertainty in a model is simply to not model agents to act on the reality of the world. Specify a model of an economy where X holds true and then model your agents such that they have no knowledge of the existence of X whatsoever. Then have the agents act, and bingo - they are going to experience "surprise". They will be acting under radical ignorance."

    Lots of very simple equilibrium models essentially do this. The assumption of equilibrium at the start means that the actors have a clear expectation of what will happen in the future (which is normally the same as what will happen in the past). Then a change occurs, which according to the setup of the model was not expected.

    1. "As Daniel pointed out earlier, Keynesians don't deny this can happen, they just don't believe that's it's important in practice."

      If Dr. Michael Emmett Brady can be considered a Keynesian (though I don't think he would use this label, despite his great respect for Paul Anthony Samuelson), I have seen him say that uncertainty and ambiguity does affect the public sector/government as much as it affects the private sector.


All anonymous comments will be deleted. Consistent pseudonyms are fine.