In the comments to my piece on Janet Yellen the hypocrisy of my position was pointed out, as it so often is, by a certain reader of this blog. What was my hypocrisy on this particular occasion? It was the fact that I complained about Yellen’s obsession with ‘closing’ models but, in other circumstances, champion Godleyian Stock-Flow Consistent (SFC) modelling which, of course, contains models that have ‘closures’ of various forms.
I think that it’s worth talking about this in a little bit more detail. I should say right off the bat that my endorsement of SFC modelling is purely opportunistic. I know that there are many people out there who make their living by ‘closing’ models and I doubt that I will persuade them to stop doing this. SFC models seem to me a nice compromise in this regard because they are actually didactically useful in that the person who explores their properties absorbs key lessons about how the macroeconomy actually functions and the importance of the national accounts. Those doing marginalist analysis, say DSGE, on the other hand are simply absorbing ideology. In short, SFC models actually convey something about the real-world, while DSGE models are a form of brainwashing that basically lobotomise those who study them and render them useless as economists.
Anyway, I think a good way of approaching the broader question here is through the work of GLS Shackle. In his wonderful book A Scheme of Economic Theory Shacle discusses two types of approaches. The first is Keynes’ own approach which he describes as such,
The General Theory of Employment, Interest and Money is the most paradoxical of books. Constructed on a purely static and equilibrium frame of formal argument, it clothes this frame in a rich and suggestive mantle of ideas about expectations and their precarious basis and their extreme unstable sensitiveness to ‘the news’. In the present book I have had to find a special term to indicate this quality of Keynes’s though, and I have called it kaleido-static. The patterns in a kaleidoscope change abruptly to something totally different, yet, given the exact character of the twist imparted to the instrument, these patterns no doubt have their own internal logic. Keynes saw in the business world a succession of highly unstable equilibria. In his formal construction he described the equilibrium of each situation, in his powerful gloss upon this formal argument he explained their almost explosive instability. (p5)
Shackle notes that a similar approach can be found in Roy Harrod’s work on growth theory and the business cycle. Another great macroeconomist whose work exhibits these characteristics was, of course, Hyman Minsky. There is an argument to be made that Joan Robinson’s work also tended in this direction but I do not want to deal with this here because poor Robinson was a very conflicted theorist, constantly at war with herself in this regard.
Shackle also notes that there is another type of economic theorising. This is the type of theorising that he attributes to Keynes’ other followers. He writes,
The image of the leaping cataracts, with pools of stillness between them, which is suggested to me by the General Theory, or the analogy with the kaleidoscope already proposed, are far from the deterministic, mechanical, cyclical and above all self-contained models which sprouted in such profusion in the fertile seed-bed of Keynes’s work… Kalecki, Kaldor, Samuelson and Hicks have in succession opted for a ‘business cycle machine’ complete in itself, to which regular and therefore predictable oscillation is as natural as the tides or the seasons. These latter models give no place at all to expectation and have no use for expectational time, but exemplify in the purest form the concept of a historical pattern seen ‘from without’ by a detached observer in whose mind it exists as a simultaneously valid whole. These, therefore, lie at the extreme of the ‘mechanical time’ ranking. (pp5-6)
I think that Shackle is basically correct. When he speaks of ‘self-contained’ models he is referring, of course, to what I was calling models with ‘closures’; that is, closed models. Note that while Keynes’ discourse puts the reader in the drivers seat, in that it includes the reader in the economic world that is being thought through, the ‘business-cycle machine’ theorists try to take a sort of God’s eye view of the whole situation. This is precisely, in my opinion, how not to train working economists. Economists should be able to consider their own position within the framework of the theory they are using and if they cannot they will not make exceptionally good policy economists.
However, I think that we can also name a third type of model. And it is under this heading that the likes of the DSGE and representative agent models fall. It is these models that are generally referred to as ‘microfounded’ but which might more properly be called ‘hermetically-sealed myopia machines’. These push the ‘business-cycle machine’ models in a far more extremist direction. The ‘business-cycle machine’ models at least dealt with aggregates. This gave the theorist some scope, even in the God’s eye position adopted, to take into account contingencies in the environment and think these through. The ‘hermetically-sealed myopia machines’ close this off entirely. They give the user a set of a priori, dogmatic list of behaviors that people can engage in.
Once formulated, these behaviors are never questioned. They must be rigid and deterministic otherwise they cannot be defined and formalised. But these completely blind the person using them to the real world in all its variety. In fact, I would argue that anyone who takes such models seriously is not actually an economist. Or, at least, they are not doing economics. Rather they are engaged in pure thought experiments with literally no connection to the real world. What these people are doing is something halfway between engineering and theology.
So, what can we now say about models? Well, those that make the best economists and the most fertile theorists — that is, those that are best able to spot novelty which is essential for any good working economist and any progressive theorist — should adhere as best they can to Keynes’ kaleido-static approach. Or, if they so feel, they should engage in a form of kaleido-dynamics if they can manage it. This requires a great deal of irony on the theorist’s part and a distinct ability to recognise that Truth is a slippery concept and that all we can do is try to weigh up rather general arguments and see which is, ordinally, the most likely. (Yes, readers of Keynes’ Treatise on Probability will see something of his underlying philosophy of knowledge here…).
Those who are better at writing textbooks and engaging in instruction should try to build the best ‘business-cycle machines’ that they can. In my opinion, the Godley approach — which is an outgrowth of Kaldor’s and Kalecki’s work — is the most promising field of inquiry. Such theorists should be very careful not to stray too far from the real world. I think, for example, that Samuelson and Hicks committed this particular sin (Hicks largely recognised this, Samuelson was blind as a bat). But with some discipline I think that there is little problem if economists build business-cycle machines and use these for didactic purposes. (But don’t try to test these thought-experiments against the data, please!).
Finally, those who are inclined to the ‘hermetically-sealed myopia machines’ should pack up and move on. There are other disciplines where the skills involved in constructing complex, rigidly deterministic systems are required. Engineering is the most obvious example. People inclined toward this sort of theorising would be better placed in these disciplines. They pay quite well and if you fall into this skill-set you will probably find some satisfaction there, but you will not make a good economist. Indeed, as I said, I don’t think that what you will be doing can adequately be called ‘economics’ and you will likely just spoil the discipline for the rest of us and annoy potential students.
Fixing the Economists 
I think the best way of approaching economics is to draw from any discipline but understand the limitations in any model or relevance. No model is correct all the time, all models are correct sometimes. The problem is people that rigidly adhere to one ideology like religious fundamentalists.
Identities are correct all the time. And they are often more useful than models for empirical work.
Not at all. Here’s a model: “Every time that the price of apples rises a unicorn appears in my living room”. Although I cannot definitely prove that this will never be true, I think that I can safely say that it will very likely never be true. I could make the same case for DSGE models because they, like my unicorn model, make absurd assumptions that do not hold in reality.
“Identities are correct all the time. And they are often more useful than models for empirical work.”
Can you give me an example please?
“…all models are correct sometimes.”
What I meant was models like IS LM or other established economic models. I didn’t mean to say any model that pops out of your head at random.
http://en.wikipedia.org/wiki/Sectoral_balances
And I clearly said that DSGE models are never true. Nor are RBCT models. Nor are EMH based models. I can go on and on.
“And I clearly said that DSGE models are never true.”
In DSGE models an increase in net wealth through helicopter money issuance would lead agents to increase their current and future consumption. That is true isn’t it?
That is one effect, yes. But you have to take the Truth of the model as a whole.
Also, I can set up equally untrue models of the economy that have that effect. Example:
Same effect. But obviously the model is a load of old rubbish.
I understand that models like IS LM can be ridiculous and make ridiculous assumptions to create a simplified economy. But saying they are never true is just as extreme as saying they are always true. Im definetely not a big fan of models but I think its a bit extreme to say they never work. AS AD model can give you a pretty good idea of the effect of policy in a simplified and basic way. Models have some functions but know their limitations which are many I agree.
I think the problem is extremists. Some people have extreme faith in particular schools of thought or models which I think is ridiculous. But to go to the other extreme and say they never work is just as crazy.
Saying that models are sometimes true and that they sometimes work are two entirely different things. The duck model may work but it is certainly not true. I would argue that all the mainstream models I noted above are definitively not true and are unlikely to “work” — when they do at all — in any substantive sense better than a far simpler and more realistic model or open-ended argument (which I prefer, obviously). That makes them redundant.
As to something like SFC models, are they true? Not really. But they have a higher information content than the mainstream ones because they impart actual truths about the national accounts and stock-flow consistency and so forth.
“Saying that models are sometimes true and that they sometimes work are two entirely different things. The duck model may work but it is certainly not true. I would argue that all the mainstream models I noted above are definitively not true and are unlikely to “work” — when they do at all — in any substantive sense better than a far simpler and more realistic model or open-ended argument (which I prefer, obviously). That makes them redundant.”
I guess the terminology is confusing. Saying a model is not true can come across as saying it never works. I think models can work half the time so they are good for making simple arguments or getting some point across. But yes I would prefer simple models and open ended arguments like you say. Its ridiculous to think that you can reduce reality down to a model and then use the model as a substitute for reality to make policy decisions.
Phil, I so enjoy coming here to learn even though I am not an economist (nor do I have anything near an economics degree altho’ I have plenty of other degrees). Just wanted to express an opinion about the value I attribute to your writings, and how I appreciate how you take the time when I ask to explain things I don’t understand. I know that doesn’t come across very often.
My only quibble would be for you to expand your alphabetized shortcuts like “And I clearly said that DSGE models are never true. Nor are RBCT models. Nor are EMH based models. I can go on and on.”
I know it’s boring, but regular folk DO come here and you would do well to accommodate their ignorance, as well as mine.
Sorry. DSGE are Dynamic Stochastic General Equilibrium models. These are what so-called New Keynesian theorists use. They have recently come under fire for their lack of realism by people like Paul Krugman.
RBCT models are Real Business Cycle Theory models. They are New Classical models that argue that recessions are due to technology and supply shocks. So, the Great Recession was due to… um… I dunno… Godzilla destroying plant and equipment I guess.
EMH is the Efficient Markets Hypothesis. This holds that financial markets are perfectly balanced in line with new information and that no one person can consistently beat the market. No comment! 😉
Great. Thank you. Now add those three to the DEFINITIONS tab that you’re missing. 😉 [With SFC.]
Nice to see you to describe your preferred “methodology” of economics explicitly. Apparently it’s a methodology that eschews formulating economic theories with precise language in favor of impressionistic “kaleidoscopes”. Tools of logic and empirical testing are replaced by literary criticism, ideas and arguments that should stand on their own by endless quoting of old authors. Discussing inconvenient implications or assumptions of your favorite theories is dismissed as “boring” intellectual game, empirical evidence is limited to eyeballing cherry-picked bar charts, and in place of serious attempt to engage opposing views one can only find juvenile psychoanalysis and thinly-veiled insults.
It’s a methodology that’s very convenient if one is interested in playing rhetorical games reinforcing his ideological prejudices and/or sense of superiority. Of course, for doing actual science, it’s useless.
Aww… I thought you were going to seriously engage. All I got was bile. After all, I know that you read my pricing paper which, of course, uses what you consider to be “precise” language (i.e. algebra). So, I can only conclude that you read the above post, got mad because I said DSGE modelling was silly and then vented in the comments. Predictable.
Also, you might want to read Shackle. He made extensive use of differential calculus and matrix algebra when it suited the argument. But he still held kaleido-static reasoning as the best tool for economic analysis.
And with that, ivansml pisses his name in the sand.
Doing math and concocting a model that uses that math but has nothing to do with reality is not science (and following a school of thought that builds auxiliary assumptions to justify their method of theorizing is a degenerate research program, ie no longer science). I suggest you read some philosophy of science if you dare to go the route of talking about what is or is not science. As for this comment:
“Discussing inconvenient implications or assumptions of your favorite theories is dismissed as “boring” intellectual game, empirical evidence is limited to eyeballing cherry-picked bar charts, and in place of serious attempt to engage opposing views”
This is what mainstream economists do. You’re not the one with the criticisms, we (heterodox) economists are.
It’s just standard badly thought through shouty rubbish, don’t pay attention. And if the mainstream are really doing science I’m very amused because they all seem to fundamentally disagree with each other and they all hold idiosyncratic views. I would think that one aspect of science would be, at the very least, some sort of consensus about fundamental issues.
This was quite a legendary comment ivansml. And very very accurate. Of course, Phil’s reply is as vacuous as ever, citing old authors (the irony!) and amateurishly diagnosing (ibid) *why* you might have written your comment.
Yeah, I got totally pwned! I read books and I do girly things like literary criticism! Oh why, oh why, can I never aspire to be a true man — a true scientist even (self-defined, naturally) — and build a DSGE model!?
I suppose we could play the game where I say “but no, you’re mischaracterising my argument” and then you or ivansml insist that you’re not and this is my argument. Then I say that I’m being mischaracterised again and round and round we go. We insist on not communicating because we feel, deep inside, an anger at the other person. They have said something bad about the tribe so they must be a Bad Person. Forgive me if I don’t engage. This is where we descend into internet argument mode, which is all the mainstreamers really have to offer when they’re called out.
@Phil:
“All I got was bile.”
No bile, just a few observations. And while my tone may be hostile (matching your own), I don’t think I’m mischaracterizing you. Case in point – this post is meant to explain why “closed” models are not useful. Yet you don’t explain anything – all we get is a quote from somebody else, restating the same claim. Arguments from authority are invalid, as I’m sure you know, and thus your whole post can be simply dismissed by linking to https://www.youtube.com/watch?v=pWdd6_ZxX8c
Moreover, even if we try to reconstruct some kind of reasonable argument from the short quote you provide, it becomes clear that Shackle is most likely referring to a particular class of models with periodic deterministic business cycles (think Samuelson’s investment accelerator). Indeed it’s hard to see how such models could capture the true nature of business cycles, which are irregular and random-looking. Which is why more recent models (including those so maligned DSGE ones) DO NOT rely on such simple deterministic dynamics, but instead usually include random shocks. And while I fully expect you’ll dismiss the difference as irrelevant technicality, these are in fact fundamentally different approaches to modelling business cycles, so your quote is not only insufficient to support your claims, but also mostly irrelevant.
And by the way, I used the term “precise language”, not math or algebra, on purpose. One can be precise also without math, or incomprehensible even with math. But it’s hard to see how one can be precise when aiming for “kaleidoscopic” thought.
@deusdarkjaws:
“This is what mainstream economists do. You’re not the one with the criticisms, we (heterodox) economists are.”
No, they don’t. And yes, you have plenty of criticisms – but quantity doesn’t imply quality, as we all know.
@pontus:
Thanks. Your fight with Phil few days back was pretty epic too 🙂
(1) The post is not “meant to explain why “closed” models are not useful”. The title and the opening paragraph clearly indicate that the idea of the post is to outline two approaches to economics and one approach to pseudo-economics. Nowhere do I make the claim that I am endeavoring to explain why closed models are not useful. Furthermore, I expressly claim that closed models CAN be useful for didactic purposes. I write:
So, you simply have read what you felt like reading into the post. None of what you claim is there.
I do say that I prefer not to use closed-models and think that the Keynes approach is better. The post is not meant to highlight WHY I think that this approach is better but I provide a gloss here:
This is tied to the fact that, as Shackle and I say, closed models take a God’s eye view of the world that is inappropriate to the nature of doing applied economic work (even a Bayesian would recognise this…).
(2) I did not use Shackle to discuss DSGE. That would be stupid because DSGE did not exist when he was writing. Again, you are reading what you want to read into the piece. Rather I said that there was a third type of model and then discuss this purely from my own perspective. My key criticism is that ‘hermetically-sealed myopia machines’ (my phrase, not Shackle’s, please note) assumes a priori deterministic human behavior which is not how the real world works.
(3) Finally, there is a difference between citation — which is meant to indicate where you get an argument from so that the reader can follow it up — and “argument from authority”. The level of scholarship in economics is exceptionally poor in this regard but most other disciplines recognise this method of doing academic work.
Step back. Take a deep breath. And read it again. Then engage.
Oh really? So you’re not discussing limitations of closed models? Well let’s check your own words:
“I complained about Yellen’s obsession with ‘closing’ models but, in other circumstances, champion Godleyian Stock-Flow Consistent (SFC) modelling which, of course, contains models that have ‘closures’ of various forms. […] I think that it’s worth talking about this in a little bit more detail. […] I know that there are many people out there who make their living by ‘closing’ models and I doubt that I will persuade them to stop doing this. […] Anyway, I think a good way of approaching the broader question here is through the work of GLS Shackle. […] I think that Shackle is basically correct. When he speaks of ‘self-contained’ models he is referring, of course, to what I was calling models with ‘closures’.”
Any reasonable person will understand your post as an explanation of your previously stated position that closed models are not useful (or of limited, merely pedagogical usefulness, if you will). Yet now you’re saying you’re just discussing “two approaches to economics and one approach to pseudo-economics”, which have nothing to do with closed models.
Whatever. Such backtracking and changing of goalposts is typical in discussion with you, so there’s no point in continuing.
Ivansml,
I don’t know why this is difficult for you. Let me state it clearly.
Pontus said that I was a hypocrite because I said that I didn’t think that closed-models were the way to go but then supported SFC modelling. So, in the post, I lay out the two approaches and explain why, while I do not like closed models, I nevertheless support SFC modelling as it has didactic usefulness.
As I wrote above:
That’s what you’re doing here. You’re kicking up dust and pretending that I’m being evasive or whatever so you can feel like you’ve “won”. That’s Pontus’ game. I’ve come to expect more from you.