Tuesday, June 2, 2015

On Missing Minsky

Yes, we miss the late Hy Minsky, especially those of us who knew him, although I cannot claim to be one who knew him very well.  But I knew him well enough to have experienced his wry wit and unique perspective.  Quite aside from that, it would have been great to have had him around these last few years to comment on what has gone on, with so many invoking his name, even as they have in the end largely ended up studiously ignoring him and relegating him back into an intellectual dustbin of history, or tried to.

So, Paul Krugman has a post entitled "The Case of the Missing Minsky," which in turn comments on comments by Mark Thoma on comments by Gavyn Davis on discussions at a recent IMF conference on macroeconomic policy in light of the events of recent years, with Mark linkhttp://economistsview.typepad.com/economistsview/2015/06/the-case-of-the-missing-minsky.htmling to Krugman's post.  He notes that there seem to be three periods of note: a Minsky period of increasing vulnerability of the financial system to crash before the crash, a Bagehot period during the crash, and a Keynes period after the crash.  Krugman argues that, despite a lot of floundering by the IMF economists, we supposedly understand the second two, with his preferred neo-ISLM approach properly explaining the final Keynes period of insufficiently strong recovery due to insufficiently strong aggregate demand stimulus, especially relying on fiscal policy (and while I do not fully buy his neo-ISLM approach, I think he is mostly right about the policy bottom line on this, as would the missing Minsky, I think).  He also says that looking at 1960s Diamond-Dybvig models of bank panics sufficiently explain the Bagehot period, and they probably do, given the application to the shadow banking system.  However, he grants that existing official models do not sufficiently explain the Minsky period, the runup, how things got so fragile that they could collapse so badly.

Now I do not strongly disagree with most of this, but I shall make a few further points.  The first is that in effect Minsky provided a model and discussion of all three stages, although his model of the Keynes stage is not really all that distinctive and is really just Keynes.  But he probably did a better job of discussing the Bagehot stage than did Bagehot, and more detailed, if less formal, than Diamond and Dybvig.  I suspect that Bagehot got dragged in by the IMF people because he is so respectable and influential regarding central bank policymaking, given his important 1873 Lombard Street, and I am certainly not going to dismiss the importance of that work.  But the essentials of what go on in a panic and crash were well understood and discussed prior to 1873, with Minsky, and Kindlegerger drawing on Minsky in his 1978 Manias, Panics, and Crashes, quoting in particular a completely modern discussion from 1848 by John Stuart Mill (I am tempted to produce the quotation here, but it is rather long; I do so on p. 59 of my 1991 From Catastrophe to Chaos: A General Theory of Economic Discontinuities), which clearly delineates the mechanics and patterns of the crash, using the colorful language of "panic" and "revulsion" along the way.  Others preceding Bagehot include the inimitable MacKay in 1852 in his Madness of Crowds book and Marx in Vol. III of Capital, although admittedly that was not published until well after Bagehot's book. 

One can even find such discussions in Cantillon early in the 1700s discussing what went on in the Mississippi and South Sea bubbles, from which he made a lot of money, and then, good old Adam Smith in 1776 in WoN (pp. 703-704), who in regard to the South Sea bubble and the managers of the South Sea company declared, "They had an immense capital dividend among an immense number of proprietors.  It was naturally to be expected, therefore, that folly, negligence, and profusion should prevail in the whole management of their affairs. The knavery and extravagance of their stock-jobbing operations are sufficiently known [as are] the negligence, profusion and malversation of the servants of the company."

It must be admitted that this quote from Smith does not have the sort of detailed analysis of the crash itself that one finds in Mill or Bagehot, much less Minsky or Diamond and Dybvig.  But there is another reason of interest now to note these inflammatory remarks by Smith.  David Warsh in his Economic Principals has posted in the last few days on "Just before the lights went up," also linked to by the inimitable Mark Thoma.  Warsh discusses recent work on Smith's role in the bailout of the Ayr Bank of Scotland, whose crash in 1772 created macroeconomic instability and layoffs, with Smith apparently playing a role in getting the British parliament to bail out the bank, with its main owners, Lord Buccleuch and the Duke of Queensbury, paying Smith off with a job as Commissioner of Customs afterwards.  I had always thought that it was ironic that free trader Smith ended his career in this position, but had not previously known how he got it.  As it is, Warsh points out that the debate over bubbles and what the role of government should be in dealing with them was a difference between Smith and his fellow Scottish rival, Sir James Steuart, whose earlier book provided an alternative overview of political economy, now largely forgotten by most (An Inquiry into the Principles of Political Oeconomy, 1767).

I conclude this by noting that part of the problem for Krugman and also the IMF crowd with Minsky is that it is indeed hard to fit his view into a nice formal model, with various folks (including Mark Thoma) wishing it were to be done and noting that it probably involves invoking the dread behavioral economics that does not provide nice neat models.  I also suspect that some of these folks, including Krugman, do not like some of the purveyors of formal models based on Minsky, notably Steve Keen, who has been very noisy in his criticism of these folks, leading even such observers as Noah Smith, who might be open to such things, to denounce Keen for his general naughtiness and to dismiss his work while slapping his hands.  But, aside from what Keen has done, I note that there are other ways to model the missing Minsky more formally, including using agent-based models, if one really wants to, these do not involve putting financial frictions into DSGE models, which indeed do not successfully model the missing Minsky.

Barkley Rosser

Update: Correction from comments is that the Ayr Bank was not bailed out.  It failed.  However, the two dukes who were its main owners were effectively bailed out, see comments or the original Warsh piece for details.  It remains the case that Adam Smith helped out with that and was rewarded with the post of Commissioner of Customs in Scotland.


Thornton Hall said...

What, exactly, is the value added of formal (or even informal) "models" in all this?

That is to say, if a historian were to describe the events and responses outlined above, what would he leave out that an economist would put in?

Keep in mind, there are an infinite number of models that fit the data. Science requires more that a fit. It requires that the model correspond with reality in a way that it can fill in observable data before it is observed.

Meanwhile, Simon Wren-Lewis dismisses the policy-maker who listens to the historian as using mere "intelligent guess work", strongly suggesting that economists clearly do better. But if trying to figure out whether the current moments is Minsky, Keynes or even Keen, isn't "guesswork" then I don't know what is. Put "intelligent guess work" policy next to model guided policy in your history above. Where's the value added from modeling? It has to be useful AND the policy maker must have a scientific reason for knowing it will be useful IN REAL TIME.

Krugman frequently defends "textbook" modeling with a "nobody else has come up with anything better" response. But that's a classic "when did you stop beating your wife".

What if the economy can't be modeled? Claiming to do the impossible is deluded, even if you can correctly say: "no one has ever improved upon my method of doing the impossible."

"But we have learned so much!" People say that, but what, exactly, are they talking about?

Thornton Hall said...

Here's a theory (not a model): the true and revolutionary insights of Veblen, Keynes, and Minsky have all failed to significantly alter the trajectory of economic thought because the discipline expects "the truth" to do the impossible.

Newton faced this when his theory of universal Gravity was criticized for failing to explain the distance of the planets from the sun. The Aristotelian tradition said that a proper theory of the heavens would do this.

And so Keynes has his Aristotelian interpreter Hicks and Minsky has his Keen. Requiring the revolution to succeed in doing the impossible means that the truth gets misinterpreted or ignored. Either way, no revolution despite every generation producing a revolutionary that sees the truth.

What is the value of this theory? If true, it explains how economics can be filled with smart people seeking the truth and yet make zero progress in more than a century.

chrismealy said...

I get the impression that mainstream economists are generally resistant to any kind of boom-and-bust models (at least while getting a BA in econ I was never taught any). Is this the case? It's too bad, because models like Lotka–Volterra are not that hard. Just from messing around with agent-based models it seems like anything with a lag or learning generates cycles. Is it because economists are fixated on optimization and equilibrium? Are they worried about models that are too sensitive to initial conditions?

rosserjb@jmu.edu said...


Maybe they should not be, but the discussion among IMF economists, Davis, Thoma, and Krugman has involved models, and in particular, conventional models. So, Krugman declares that there are conventional models as noted above to cover two of the stages, the latter two, but not the first one identified with Minsky. I think there are better models for all this, but they are not the conventional ones.


The DSGE and other conventional models are able to model booms and busts, although they generally do not use the Lotka-Volterra models that such people as the late Richard Goodwin (and even Paul Samuelson) have used for modeling business cycle dynamics. The big difference is that the conventional models involve exogenous shocks to set off their busts, with cyclical reverberations that decay then following the exogenous shock, with some of the lag mechanisms operating for that.

It is not really surprising that this sort of thing does not model Minsky or the Minsky moment, which involve endogenous dynamics, the very success of the boom as during the Great Moderation itself undermining the stability and even resilience of the system as essentially endogenous psychological (and hence behavioral) factors operate to loosen requirements for lending and to use Minksy language, lending and borrowing increasingly involves highly leveraged Ponzi schemes (and I note that some more conventional economists have emphasized leverage cycles, notably John Geanakoplis, although avoiding Minsky per se in doing so).

New Deal democrat said...

This is a good post snd discussion so far. So here's my $.02:

1. Maybe the behaviorists like Thaler have already explored this, but it seems to me that economists still need to learn learning theory from psychologists. Most importantly, "bservational learning," { http://psychology.about.com/od/oindex/fl/What-Is-Observational-Learning.htm ), or more simply "monkey see, monkey do." We constantly learn by observing behavior in others: our parents, our older siblings, the cool kids at school, our favorite pop icons, our professors, our business mentors, and so on. As to which,

2. Some people are better at learning than others (duh!). Some learn right away, some more slowly, some never at all. And further,

3. Some people are more persceptive than others, recognizing the importance of something earlier or later. If you recognized how important the trend change was when Volcker broke the back of inflation in the early 1980s, and simply bought 30 year treasuries and held them to maturity, you made a killing. If you discovered that in the early 1990s, you made less. And so on.

All we need, to pick up on chrismealy's comment, are time periods and learning. Incorporate variations in skill and persceptiveness into the population, and you can get a nice boom and bust model. As more and more people, with various levels of skill, learn an economic behavior (flipping houses, using leverage), they will "push the edge of the envelope" more and more -- does 2x leverage work? Yes, then how about 4x? Yes, then how about 20x? -- until the system is overwhelmed.

4. But if you don't want to incorporate imitative learning models from psychology, how about just using appraisals of short term vs. long term risk and reward. Suppose it is the 1980s, and I think treasury yields are on a securlar downtrend. But this book called "Bankruptcy 1995" just came out, based on a blue ribbon panel Reagan created to look at budget deficits. That best selling book forecasts a "hockey stick" of exploding interest rates by the mid-1990s due to ever increasing US debt. So let's say I am 50% sure of my belief that treasury yields will continue to decline for another 20 years, and I can make 10% a year if I am right. But if I am wrong .....

Meanwhile, I calculate that there is an 80% chance I can make 10% a year for the next few years by investing in this new publicly traded company named "Microsoft."

Even leaving aside behavioral finance theories about loss aversion, it's pretty clear that most investors will plump for Microsoft over treasuries, given their relative confidence in short term outcomes.

Historically, once interest rates went close to zero at the outset of the Great Depression, they stayed there for 20 years, and then gradually rose for another 30. How confident are investors that the same scenario will play out this time?

Either or both of the learning theory or the short term-long term risk reward scenario are good explanations for why backwards induction ad absurdum isn't an accurate description of behavior.


BTW, a nice example of a failed "backwards induction" is the "taper tantrum" of 2013. Since investors knew that the Fed was going to be raising interest rates sooner or later, they piled on and raised interest rates immediately -- and made a nice intermediate term top at 3%.

reason said...

I think New Deal Democrat has it here. This surely, can be covered with a simple model of asynchronous adaptive expectations with stochastic (Taleb type - big tail) risks. I wouldn't think you would even need a sophisticated agent based model. There must be plenty of ratchet type models out there to chose from.

Sandwichman said...


"...the true and revolutionary insights of Veblen, Keynes, and Minsky have all failed to significantly alter the trajectory of economic thought because the discipline..." sacrifices to the God, Equilibrium.

New Deal Democrat,

WRT "monkey see, monkey do" see Andred Orlean's The Empire of Value which articulates his mimetic theory of value.

rosserjb@jmu.edu said...

To New Deal Dem and reason,

Sorry, I am not on board with this at all. Sure, I am all for incorporating learning and lags. No problem. This is good old adaptive expectations, which I have no problem with.

The problem is back to what I said earlier, that Minsky's apparatus operates endogenously without any need for exogenous shocks, although it can certainly operate within those, as his quoting of Mill shows, although I did not provide that quote, but Mill starts his story of how bubbles happen with some exogenous initial supply/demand shock in a market.

Why is what you guys talk about an exogenous shock model? Look at the example: Volcker does something and then different people figure it out at different rates. But Volcker is the exogenous shocker. If he does nothing, nothing happens.

In Minsky world, there does not need to be an exogenous shock. The system may be in a total anf full equilibrium,, but that equilibrium will disequilibrate itself as psychological attitudes and expectations endogenously change due to it. This is what the standard modelers have sush a problem with and do not like. They have no problem wiht adaptive expectations models. This is all old hat stuff for them, with only the fact that one does not know for sure what all those lags are being the problem, and what opened the door to the victory of ratex because it said there are no lags and thus no problem. Agents now what will be on average.

Anonymous said...

Warsh's history of the Ayr Bank has errors. The Bank of England offered it a "bailout" in 1772 but required the personal guarantees of the two Dukes which were not forthcoming. The Ayr Bank struggled on without lender of last resort support until August 1773, when it closed for good. (This is all in Clapham's history of the Bank of England.)

What Warsh is calling a "bailout" was not a bailout of the bank, but of its proprietors who had unlimited liability and were facing the possibility of putting their estates on the market (which would have affected land prices in Scotland).

As I understand Warsh's description, Parliament granted the two Dukes a charter for a limited liability company that would sell annuities. It is entirely possible that contemporary sources would describe such an action as "indemnifying" the promoters of the company. But what is meant by this use of the term is only that Parliament authorized the formation of corporation. The actual indemnity is provided in the event the corporation fails by the members of the public who are creditors of the corporation.

In short, it is an error to claim that there was a "bailout" of the Ayr Bank.

rosserjb@jmu.edu said...

Point taken, csissoko. I accept your corrected version. Parliament bailed out the dukes, not the Ayr Bank, which indeed failed.

Thornton Hall said...

I still don't understand what information is added by "models". Krugman has a job he has created for himself where everything he does is with an eye toward policy.

So I'm a policy maker. Explain why I need a model. in the 1930s austerity caused recessions and WWII ended the Depression. A little history of Japan's lost decade and some thinking about the implications of fiat currency, and, voilia, Krugman's policy suggestions, with no models and therefore no need to listen to economists like Mankiw or the Germans currently destroying Europe (third times a charm).

Thornton Hall said...

By the by, I have thought this thru. The head of Duke's Philosophy Department agrees: Krugman's method for using models is empty hand-waving. However he comes to his conclusions, it is not logically possible that ISLM, or any other model, has anything to do with it.

AXEC / E.K-H said...

Sitcom economics
Comment on ‘On Missing Minsky’

Since Adam Smith economists have told rather enthralling stories about speculations, manias, follies, frauds, and breakdowns. The audience likes this kind of stuff. However, when it comes to how all this fits into economic theory things become a bit awkward. Of course, we have some modls -- Minsky, Diamond-Dybvig, Keynes come to mind -- but we could also think of other modls -- more agent-based or equilibrium with friction perhaps. On closer inspection, though, economists have no clue at all.

Keynes messed up the basics of macro with this faulty syllogism: “Income = value of output = consumption + investment. Saving = income - consumption. Therefore saving = investment.” (1973, p. 63)

From I=S all variants of IS-LM models are derived including Krugman's neo-ISLM which allegedly explains the post-crash Keynes period. Let there be no ambiguity, all these models have always been conceptually and formally defective (2011).

Minsky built upon Keynes but not on I=S.
“The simple equation ‘profit equals investment’ is the fundamental relation for a macroeconomics that aims to determine the behavior through time of a capitalist economy with a sophisticated, complex financial structure.” (Minsky, 2008, p. 161)

Here profit comes in but neither Minsky, nor Keynes, nor Krugmann, nor Keen, nor the rest of the profession can tell the fundamental difference between income and profit (2014).

The fact of the matter is that the representative economist fails to capture the essence of the market economy. This does not matter much as long as he has models and stories about crashing Ponzi schemes and bank panics. Yes, eventually we will miss them all -- these inimitable proto-scientific storytellers.

To have any number of incoherent models is not such a good thing as most economists tend to think. What is needed is the true theory.

“In order to tell the politicians and practitioners something about causes and best means, the economist needs the true theory or else he has not much more to offer than educated common sense or his personal opinion.” (Stigum, 1991, p. 30)

The true theory of financial crises presupposes the correct profit theory which is missing since Adam Smith. After this disqualifying performance nobody should expect that some Walrasian or Keynesian bearer of hope will come up with the correct modl any time soon.

Egmont Kakarot-Handtke

Kakarot-Handtke, E. (2011). Why Post Keynesianism is Not Yet a Science. SSRN Working Paper Series, 1966438: 1–20. URL http://ssrn.com/abstract=1966438.
Kakarot-Handtke, E. (2014). The Three Fatal Mistakes of Yesterday Economics: Profit, I=S, Employment. SSRN Working Paper Series, 2489792: 1–13. URL http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2489792.
Keynes, J. M. (1973). The General Theory of Employment Interest and Money. The Collected Writings of John Maynard Keynes Vol. VII. London, Basingstoke: Macmillan.
Minsky, H. P. (2008). Stabilizing an Unstable Economy. New York, NY, Chicago, IL, San Francisco, CA: McGraw Hill, 2nd edition.
Stigum, B. P. (1991). Toward a Formal Science of Economics: The Axiomatic Method in Economics and Econometrics. Cambridge, MA: MIT Press.

New Deal democrat said...

Prof. Rosser:

In addition to my hypo re Volcker and interest rates, I also mention flipping houses and leverage.

Person A flips a house, makes $100k. Person B learns of it, figures s/he can do just as well, and flips a house. Eventually enough people are doing it that news stories are written about it. By now 1000s of people are figuring, "if they can do it, I can do it.":

So long as the trend continues, the person using financial leverage to flip houses makes even more profit. Person B uses more leverage, and so on. And since 2x leverage worked, why not 4x leverage. And if that works, why not 10x leverage?

Both the number of people engaging in the behavior, and the financial leveraging of the behavior, are endogenous, unless you are going to hang your hat on existing trend (note, not necessarily a shock - of rising house prices0.

All you need is more and more people of various skill sets at various entry points of time engaging in the behavior, and testing increasing leverage the more the behavior works.

Secondly, as to stability breeding instability, stability itself is the existing trend. Increasingly over time, more and more leverage will be used to profit off the existing trend. All it takes is learning + risk-takers successfully testing the existing limits. The more stable the system, the more risk-takers can apply leverage without rupturing it -- for a while.

New Deal democrat said...

Let me try to express my position as a series of axioms:
1. Assume that no system, no matter how stable, can withstand infinite leverage.
2. Assume that there is a certain non-zero percentage of risk-taking individuals.
3. Assume that risk-takers will use some amount of leverage to attempt to profit within a stable system.
4. Assume that risk-takers will use increasing leverage once any given lesser percentage of leverage succeeds in rendering a profit, in order to increase profits.
5. Assume that others will learn, over various time periods, at varying levels of skill, to imitate the successful behavior of risk-takers.

Under those circumstances, it is certain that any system, no matter how stable, will ultimately succumb to leverage. And the more stable, the more leverage will have been applied to reach that breaking point. I.e., stability breeds instability.

Bruce Wilder said...

Even using history as an analogy is implicitly introducing a model. You're saying, here's my model of this history and I crank my little model to show the behavior of the model simulates the historical record, then I adapt the model to present circumstances, and crank again arguing, again by analogy.
What I would object to is the reliance on "analytic models" as opposed to operational models of the actual institutions. Economists love their analytic models, particularly axiomatic deductive "nomological machines", DSGE being the current orthodox approach. Not that there is anything wrong about analysis. My objection would be to basing policy advice on a study of analytic models to the exclusion of all else -- Krugman's approach -- rather than an empirical study of institutions in operation (which would still involve models, because that's how people think, but they might be, for example, simulation models calibrated to observed operational mechanisms).
There are reasons why economists prefer analytic models, but few of those reasons are sound. In the end, it is a matter of bad judgment fostered by a defective education and corruption or weakmindedness. Among other things, reliance on analytic models give economics an esoteric quality that privileges its elite practitioners. Ordinary people can barely understand what Krugman is talking about in the referenced piece, and that's by design. He does his bit to protect the reputations of folks like Bernanke and Blanchard, obscuring their viewpoints and the consequences of their policies.
I am not sure what can be done about it. Economists like Krugman are as arrogant as they are ignorant -- there's not enough intellectual integrity to even acknowledge fundamental errors, and that lack of integrity keeps the "orthodoxy" going in the face of manifest failures. For the conservatives on some payroll, the problem is even worse.
I am not confident that shooing economists from the policy room and encouraging politicians to discuss these matters among themselves improves the situation. In doubt, people fall back on a moral fundamentalism of the kind that gets us to "austerity" and "sacrifice" and blames the victims -- pretty much what we have now.
Re-doing Minsky as an analytic model is an impossible task almost by definition. Minsky's approach was fundamentally about abstracting from careful observation of what financial firms did, operationally. It made him a hero with many financial sector denizens, who recognized themselves in his narratives, even when he cast them in the role of bad guys. (No one is ever going to recognize himself as a representative agent in a DSGE model.)
Perhaps the hardest thing to digest from Minsky is the insight that business cycles can not be entirely mastered. The economy is fundamentally a set of disequilibrium phenomena, the instability built-in (endogenous, as they say). The New Keynesian idea is that the economy is fundamentally an equilibrium phenomenon, that occasionally needs a helping hand to recover from exogenous disturbance. These are antagonistic world views, which cannot be reconciled with each other, and the New Keynesian view can be reconciled only minimally with the observable facts of the world, by a lot of ad hoc fuzzy thinking ("frictions").
It's very ugly.

Thornton Hall said...

Bruce, I disagree with your view of politicians. The current GOP crop are essentially following the moral philosophy that, in the end, is the only content generated by economics. But it was not always thus.

I once watched Senator Kit Bond of Missouri (very-R) try to round up a quorum in the Small Business Committee. It was quite clear that the man enjoyed people. He liked the company of just about everybody. Without the strong interference from economists, that's who ends up in politics. People like that are pragmatic. They try things. They aren't there for the purpose (contra Ted Cruz) of breaking things.

You're right, my problem really is with analytic "models" which aren't really models but rather metaphors or analogies. But I don't think that's the only way reasoning from history can work.

There are lots of areas of policy, some of which continue to resist conversion to economic religion. In education policy we try interventions and see what happens. It's inductive and mostly correlation, but thru trial and error we do progress toward better policy (although schools of education are only slowly moving away from their notoriously anti-scientific past).

Politicians don't have to think about the budget like a household and tighten belts. They know that business borrow money all the time. It's actually the language of the academy that leads to "tightening belts" instead of investing in the future. Economics is the science of claiming that if you need something, and you can afford that something, you still must consider "multipliers" or "the philosophers stone" or some other nonsense before you can decide to buy what you need.

Thornton Hall said...

Mark Buchanan's book "Forecast" describes just this process.

Thornton Hall said...

What I mean to say about history: don't confuse theories with models. I have a theory about what caused what in the Great Depression. But I don't model the economy.

reason said...

I think I should clarify, I think endogenous and exogenous are a little bit besides the point here. I think the exact trigger that starts a "state change" in the system has a stochastic component. But the increasing vulnerability of the system is endogenous, in a very Minsky sense. What I am saying is that increasing vulnerability could be modelled without using agent based modelling (a bit like modelling landslides or earthquakes if you like). I'm not saying that the model is just being driven by exogenous shocks.

I think there is a bit of tendency to mischaracterise what Paul Krugman is saying. He is the last person you should be accusing of mistaking the map for the territory. He is saying that EVEN relatively simple models can make sensible suggestions about policy in some circumstances.
Yes, though I tend to agree with you that general equilibrium is the original sin in macro-economic modelling and that the system is in fact a disequilibrium system. But that doesn't imply to me at all that you can't use analytical approaches.

Thornton Hall said...

Krugman has no map. My link above demonstrates this quite conclusively.