Friday, February 13, 2009

Assumptions and Limitations

I recently came across this astonishing quotation, also referred to as “the F-twist”:

“Truly important and significant hypotheses will be found to have “assumptions” that are wildly inaccurate descriptive representations of reality, and, in general, the more significant the theory, the more unrealistic the assumptions (in this sense). The reason is simple. A hypothesis is important if it “explains” much by little, that is, if it abstracts the common and critical elements from the mass of complex and detailed circumstances surrounding the phenomena to be explained and permits valid predictions on the basis of them alone. To be important, therefore, a hypothesis must be descriptively false in its assumptions; it takes account of, and accounts for, none of the many other attendant circumstances, since its very success shows them to be irrelevant for the phenomena to be explained.”


I imagine I'd send this as reply to a referee report which criticizes my work on the grounds that I might have found a great dark matter candidate, but not only have I assumed unbroken supersymmetry, my model also has four neutrino generations and, oh, only 2 spatial dimensions. Assumptions that indeed qualify as unrealistic and wildly inaccurate. For not to say, bluntly wrong.

But maybe I am being unfair.

Let me guess what Friedman might have wanted to say. The more parsimonious a model, the easier it is to extract relevant features and get an understanding of its behaviour. That does not mean however, it makes for a better model the fewer and more unrealistic assumptions you have. Certainly, the standard model of particle physics would be nicer if all fermions were massless and chiral symmetry was unbroken. Unfortunately, it doesn't describe Nature then. That's why we distinguish between models of the real world, and 'toy models' meant as testing ground to increase our understanding of the general features (see earlier post on Models and Theories).

But maybe also this is unfair.

He might have meant to say that a simplifying assumption does not have to be shown appropriate for a certain range of validity, the range in which predictions derived from the assumption then can be made. Instead, one can just see whether the model works and such justify the assumption a posteriori. Unfortunately, that too is nonsense. If you don't specify the range of validity of your assumptions (typically by showing that the effect of deviations from the assumptions is negligible for the result) your model is not falsifiable and thus not scientific. If you test it and the outcome does not match your predictions, you can just go and say, well, the assumptions were not fulfilled.

Thus, I am afraid unless you want to redefine what you mean with a scientific theory, this is not a good starting point. One wonders why he felt the need to put “explain” in quotation marks.

See also: Shut up and calculate

61 comments:

  1. He seems to be trying to express Popper's thesis on bold hypotheses but muddling it a bit. Well, whaddaya expect from an economist? :D

    ReplyDelete
  2. And what if referee tells you that you "solved" something that wasn's questionable in the first place?
    A.

    ReplyDelete
  3. Bee:

    That's why Thomas Carlyle called economics the "dismal sciemce" – it rarely behaves as a science at all.

    As you and Friedman imply, Occam's Razor is the excuse for stripping down models to the barest of essentials.

    But it's 'reality' that's the umpire of whether the model has any scientific worth.

    Many economists neglect this caveat.

    ReplyDelete
  4. Milton Friedman: Noble Laureate/Economics; "Whip Inflation NOW!" US stagflation; "The Boys from Chicago" given free rein/reign by Augusto Pinochet in Chile. Merton, Scholes: Nobel Laureates/Economics (Black died). Black-Scholes derivative pricing in Long-Term Capital Management disaster and the US NINJA mortgage meltdown.

    Putting equations to paper is insufficient. Observation outside politically correct math exists. Theorists boast promiscuity while experimentalists pay child support.

    The parity Eötvös experiment opposes left- and right-handed quartz single crystals. Its net output will be zero (historic value for all pror runs) or not zero. Somebody should look.

    ReplyDelete
  5. It looks like your referee report was not as bad as this one.

    ReplyDelete
  6. Hi Bee,

    I guess then it can be safely assumed that when economists such as Friedman studied what a theory should represent as being they never heard of Einstein as he said:

    "The supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience"

    Of course it should be reminded that many economists believe they are creating reality, rather then simply explaining it. What our current experience should have them now come to grips with is that reality can only represent as being what’s possible or perhaps further restricted to say that reality itself defines what that is. Perhaps then Einstein should have said as to clarify:

    'God doesn’t play the stock market or make loans that can never be repaid' :-)

    Best,

    Phil

    ReplyDelete
  7. Bee,

    I'm sure you know Economics more than I do, so I'm sure you already know what I am about to say.

    I think Friedman has in mind such assumptions as 'people act only in their own interest' and 'people act rationally' which are wildly inaccurate descriptions of reality in simpler circumstances but have (according to Friedman)great explanatory power under 'complex and detailed circumstances'. Of course, Friedman should specify the range of validity. Perhaps Friedman takes it as implied that he's talking about the sort of range in which economics is usually interested in.

    By the way, I read that assumptions such as 'self-interested behaviour' have little empirical evidence.

    ReplyDelete
  8. Hi Count,

    I didn't actually receive any referee report (I've never written a paper on dark matter, and if, it would certainly have 3 spatial dimensions, believe me). The case you linked to is indeed bad, but it tells something I am neither surprised nor shocked. I have received my share of nonsensical referee reports without content, and when I complained to the editor I was informed that 'the referee is a well-known expert in the field' and thus knows what he is doing. Amen. On the other hand, one learns how to avoid such journals and editors with time. Best,

    B.

    ReplyDelete
  9. Hi Nirmalya,

    Yes, and economics has received a lot of criticism because of that. What I was trying to say is that the defense Friedman offers is not appropriate. You are indeed allowed to make unrealistic assumptions, but if you do, you will have to show that this is a good approximation for the situation you are considering - and that making it more realistic doesn't totally destroy the conclusions. Best,

    B.

    ReplyDelete
  10. All Friedman is saying is that a hypothesis which explains a particular set of evidence, and yet is based on assumptions that are wildly different from experience, is more significant and non-trivial than one based on assumptions that closely conform to experience.

    It has nothing to due with the number of assumptions made.

    ReplyDelete
  11. Hi Just Learning,

    I didn't say anything about the "number of assumptions made". But either way, what you say is nonsense. How "significant" a theory is depends on how well it describes Nature and not on how much its assumptions seem to match (or not to match) your experience. Best,

    B.

    ReplyDelete
  12. I'll stand by my comment, but I think you did imply that you thought that the number of assumptions were relevant in your statement:

    "The more parsimonious a model, the easier it is to extract relevant features and get an understanding of its behaviour. That does not mean however, it makes for a better model the fewer and more unrealistic assumptions you have."

    All I'm saying is that you are wrong in your interpretation of Friedman. He did not place a preferred limit, big or small, on the number of assumptions made, only that "A hypothesis is important if it “explains” much by little".

    So he is saying that the ability to describe a large collection of information with a smaller collection of information is significant, but he hasn't placed any expectation on the magnitude of assumptions or parameters. That we prefer fewer is a comment about human predilection, but not a requirement.

    With that said, our description of reality is based on our experience. Friedman is quite clear that if a hypothesis (and it is implicit in his discussion that any hypothesis is intended to describe reality and make predictions) is predicated on assumptions that are not "realistic" (or rather conform with experience), and yet that hypothesis successfully describes data and makes testable predictions, than that hypothesis is more significant than one based on assumptions that are "realistic" (eg conform with experience).

    ReplyDelete
  13. (or rather conform with experience)=

    (or rather do not conform with experience)

    ReplyDelete
  14. Hi Just Learning,

    With 'parsimonious' I was not counting the number of assumptions as you seem to imply. You could have a theory with few pretty involved assumptions or many very simple ones, I would consider both cases of parsimonious. But it is moot discussing the issue since what I find parsimonious might not be what you do, or Friedman does. This is equally fuzzy as Friedman's notion of 'explaining much by little'.

    What I said above is that I generally agree it is desirable to have a model that does indeed explain 'much by little', but that desirability is not sufficient to justify the model.

    Friedman is quite clear that if a hypothesis (and it is implicit in his discussion that any hypothesis is intended to describe reality and make predictions) is predicated on assumptions that are not "realistic" (or rather conform with experience), and yet that hypothesis successfully describes data and makes testable predictions,

    I think you did not understand what I was saying. If you have a hypothesis and you make a prediction that is based on an assumption you know is unrealistic, you have to explain why - in the range that you make predictions - this assumption is justified. The way you do that is by showing that making the assumption more realistic would not change the predictions dramatically. If you fail to do that your model does not make any falsifiable predictions, since - if your predictions turn out to be wrong - you could always say that's because the assumption was not fulfilled.

    Best,

    B.

    ReplyDelete
  15. "If you have a hypothesis and you make a prediction that is based on an assumption you know is unrealistic, you have to explain why - in the range that you make predictions - this assumption is justified. The way you do that is by showing that making the assumption more realistic would not change the predictions dramatically. If you fail to do that your model does not make any falsifiable predictions, since - if your predictions turn out to be wrong - you could always say that's because the assumption was not fulfilled."

    My immediate interpretation of this comment would be that you are conflating initial conditions with assumptions.

    ReplyDelete
  16. Dear Bee,

    Subsequently Friedman writes:

    "To put this point less paradoxically, the relevant question to ask about the “assumptions” of a theory is not whether they are descriptively “realistic,” for they never are, but whether they are sufficiently good approximations for the purpose in hand. And this question can be answered only by seeing whether the theory works, which means whether it yields sufficiently accurate predictions. The two supposedly independent tests thus reduce to one test.
    The theory of monopolistic and imperfect competition is one example of the neglect in economic theory of these propositions. The development of this analysis was explicitly motivated, and its wide acceptance and approval largely explained, by the belief that the assumptions of “perfect competition” or “perfect monopoly” said to underlie neoclassical economic theory are a false image of reality. And this belief was itself based almost entirely on the directly perceived descriptive inaccuracy of the assumptions rather than on any recognized contradiction of predictions derived from neoclassical economic theory. The lengthy discussion on marginal analysis in the American Economic Review some years ago is an even clearer, though much less important, example The articles on both sides of the controversy largely neglect what seems to me clearly the main issue - the conformity to experience of the implications of the marginal analysis - and concentrate on the largely irrelevant question whether businessmen do or do not in fact reach their decisions by consulting schedules, or curves, or multivariable functions showing marginal cost and marginal revenue. "

    ReplyDelete
  17. Friedman follows up with a reasonable argument which can be summarized as follows:

    The law that a body falls in time t a distance s given by

    s = 1/2 t^2

    works for a lead ball, but not for a feather. It is this circumstance which leads us to know that assuming that we are close enough to a vacuum for a feather is false, i.e., the theory doesn't work for a feather leads us to realize that the assumptions behind it are false for a feather.

    quote: "This point needs emphasis, because the entirely valid use of “assumptions” in specifying the circumstances for which a theory holds is frequently, and erroneously, interpreted to mean that the assumptions can be used to determine the circumstances for which a theory holds, and has, in this way, been an important source of the belief that a theory can be tested by its assumptions.

    Or to put it in my words, the theory of the falling body specifies that we must be close enough to a vacuum, but the theory does not a priori tell us whether we are actually close enough to a vacuum in a particular case (we'd need additional theory for that).

    So, trying to make sense of the F-twist, at best I can say that "we are in a vacuum" is a assumption of the falling body theory that is false in most experience, but is a reasonable assumption for many cases in our atmosphere is something that we find by testing. But I'm still not led to "Truly important and significant hypotheses will be found to have “assumptions” that are wildly inaccurate descriptive representations of reality", and so must admit my belonging to the ranks of small-brained people who cannot understand Luboš Motl's science.

    ReplyDelete
  18. Reading further, I see Friedman is trying to pull a fast one: If the "efficient markets" assumption leads to a fruitful theory, then the falsity of "efficient markets" need not concern us.

    ReplyDelete
  19. Hi Just Learning,

    My immediate interpretation of this comment would be that you are conflating initial conditions with assumptions.

    You can have additional assumptions for initial conditions, but that's not the relevant point here. It really isn't so hard to understand. You are making assumption A,B,C,D for a model from which you follow D and make a prediction for a system. The system however doesn't fulfill condition A. Now you can still look whether your prediction is fulfilled. Great if it is. But if it isn't you can say, well, that's because A was not fulfilled. You thus can never proof the model wrong. What I am saying you have to do is you need to justify why A is a good assumption even if the system does not fulfill A. This then tells you in which circumstances your prediction should hold, and when not. If you can't do that, your theory is not scientific. Best,

    B.

    ReplyDelete
  20. Dear Arun,

    Well, yes, it certainly does not follow any relation between 'significance' and 'wildly inaccurate' assumptions.

    Regarding the example with the feather and vacuum, let me repeat what I said above. As far as free fall without friction is concerned, you can determine under which circumstances this is a good approximate assumption to describe a system. If you make a prediction, you thus know whether or not the assumption is fulfilled to good enough accuracy to make for a good prediction. If you did not know that, you could drop an object and if it does not as you predicted you can always go and blame it on air resistance, thereby saving your model.

    And that, basically, is what has saved the efficient market hypothesis again and again. Whenever shit happens that shouldn't happen, or whenever data doesn't confirm the predictions, well, then probably people weren't acting really rational, or maybe markets are not sufficiently perfectly competitive. Or maybe not deregulated enough? The problem is, as long as you can't tell how large an impact deviations from the idealized assumptions have on the predictions, and these idealized assumptions are pretty much never fulfilled, your model is unfalsifiable, and you will never make any progress on it. Best,

    B.

    ReplyDelete
  21. Typo:

    "A,B,C,D for a model from which you follow D" should have been "from which you follow E"

    ReplyDelete
  22. "You are making assumption A,B,C,D for a model from which you follow D(E) and make a prediction for a system. The system however doesn't fulfill condition A. Now you can still look whether your prediction is fulfilled. Great if it is. But if it isn't you can say, well, that's because A was not fulfilled. You thus can never proof the model wrong."

    I don't agree with your construction of the issue. Friedman was arguing that we should not disregard a successful model simply because the assumptions were "unrealistic". His concern was that people were attacking models that used idealized, perfect assumptions as a starting point.

    Perfect conditions can never be met in principle, so your construction is misleading. The perfect conditions can never be met and therefore can never be tested. That doesn't undermine the falsifiability of predictions. Our ability to test certain predictions might be limited, but as long as they can be tested in principle, or at a later time when our ability to do so has improved, then the theory is still valid.

    Although it might be true that certain idealized conditions can never be met in practice, and therefore we are limited in what we can test, it is also true a predictive theory will have some ability to be tested within the range that is testable, otherwise it isn't really a theory. As long as this is true, than the theory is still falsifiable.

    Friedman argued that we shouldn't be obsessed about "unrealistic" assumptions if our theory made successful predictions about the world we lived in. I believe for him the idea that we could make connections from our world to an idealized one was significant and insightful.

    ReplyDelete
  23. Hi Just Learning,

    Friedman was arguing that we should not disregard a successful model simply because the assumptions were "unrealistic". His concern was that people were attacking models that used idealized, perfect assumptions as a starting point.

    I never said you should disregard a successful model because the assumptions are unrealistic. I have explicitly acknowledged several times assumptions can indeed be unrealistic. What I disagree with, and it seems you still didn't understand that, is the absoluteness with which he puts forward this opinion, and the lacking of criteria under which circumstances such unrealistic assumptions are okay.

    Perfect conditions can never be met in principle, so your construction is misleading.

    There is nothing 'misleading' about my construction, and I am well aware perfect conditions can never be met in reality, but this is not the point. To repeat it once again, the point is if you have used an unrealistic assumption you will need to show in which circumstances it is appropriate to use it as an idealisation of the actual world. If you fail to do that, your theory is not falsifiable because one of the assumptions you have used to draw conclusions is always violated and leaves you an emergency exit.

    The perfect conditions can never be met and therefore can never be tested. That doesn't undermine the falsifiability of predictions

    Oh yes, it does. Scroll up and read again what I wrote. It really is not so hard to understand. I will say it one more time: You use an idealized assumption A, next to B,C,D to construct a model of the real world. You derive from A,B,C,D conclusion and prediction E. Then you look at the real world system which does not fulfill A. And let us be clear here, Friedman is talking about 'wildly inaccurate' assumptions, that are de facto known to be crudely false. Now if the system meets your prediction nevertheless, great. But if it doesn't, you can go and say that's because A was not fulfilled. Thus, the theory is not falsifiable. What you have to do instead is to say why A is a good, though idealized, approximation for the real condition, say A', and why the difference between A' and A does not have a large impact on your prediction E under some circumstances that you better be careful to specify. Best,

    B.

    ReplyDelete
  24. Dear Bee,

    It is very difficult to make sense from the F-twist. Let's assume someone can, and can even explain it in fewer words than Friedman. That means the F-twist is a lousy piece of writing. Since, of course, we cannot accuse the Nobel Laureate grand-pooh-bah of being obscure, it means any explanation must of necessity be longer than Friedman's paper that you linked to.

    The real question to ask, however, is not whether the F-twist makes sense or not; but whether it has been influential.

    ReplyDelete
  25. I think you agree with my comment.

    If I say that condition A can never be met, yet I provide an appropriate corrective factor, my theory is still falsifiable because I can test to see if my corrective factor is always more significant than any effect implied by A. Once I get to a point where the effects of condition A is less significant than chance, then A is clearly not relevant and my theory is falsified. However, never in this argument does the impossibility of condition A itself undermine the falsifiability of the theory.

    ReplyDelete
  26. This comment has been removed by the author.

    ReplyDelete
  27. Hi Bee,

    I’ve been following all the back and forth exchange in regards to the logic of Friedman and his followers, with one of the principle followers being of course Alan Greenspan. I though I’d offer just a few admissions he’s made in this regard since the onset of the crisis garnered from pieces in Bloomberg's and the U.K. Guardian

    ‘Greenspan said he was ``partially'' wrong in opposing regulation of derivatives and acknowledged that financial institutions didn't protect shareholders and investments as well as he expected.’

    "I made a mistake in presuming that the self-interests of organisations, specifically banks and others, were such that they were best capable of protecting their own shareholders and their equity in the firms,"

    ` “Part of the problem was that the Fed's ability to forecast the economy's trajectory is an inexact science”

    And my personal favourite:

    `If we are right 60 percent of the time in forecasting, we are doing exceptionally well; that means we are wrong 40 percent of the time,'' Greenspan said. ``Forecasting never gets to the point where it is 100 percent accurate.''

    If that was to be counted as exception I wonder what the real tally was. Perhaps 50/50 :-)

    Best,

    Phil

    ReplyDelete
  28. Hi Just Learning,

    I do agree with your most recent comment. However, Friedman apparently 'forgot' to mention the necessity of justifying an idealized assumption by considering the effect deviations towards less ideal but more real situations have, and making sure such deviations have no devastating consequences on the predictions. Even if that necessity was clear to Friedman, it seems to me it was not clear to many people who have used this statement as a free ticket to not care whether or not assumptions are realistic. Clarifying this was the essence of my post. When it comes to economic models, I am still waiting for people to realize that. I would really like to know why real markets should be 'efficient' if this conclusion is build on unrealistic assumptions and it is not clear how much an effect deviations from these idealized assumptions does have on the 'efficiency'. That is to say, show me an analysis that proves the conclusions holds for deviations from the idealized model that can include the real circumstances. Best,

    B.

    ReplyDelete
  29. The "free ticket" not to care whether the assumptions of a theory are "realistic" was indeed the very point of Friedman's F-twist and I am sure that every good theoretical physicist understands that this principle is very important.

    It is scientifically illegitimate to attack a theory because its assumptions look "unrealistic" - in the sense of "(too) idealized" - to someone. Only if actual evidence emerges against a theory, one has a reason to modify it, abandon it, or replace it.

    If the goal is e.g. to explain the behavior of the society and many corporations, neoclassical economics assumes things like perfect competition etc. And it gives the more-or-less right predictions. Some people found the assumptions about "perfect competitions" that were needed for those predictions "unrealistic" or "too idealized" and created alternative theories of "imperfect competition" and all this crap.

    Friedman explains that these alternative theories were never driven by actual evidence but just by a dissatisfaction with the "realism" of the assumptions - and such a motivation for a new theory is simply unscientific. There is no evidence whatsoever that these modified (and thus more awkward) theories give better predictions than the idealized, neoclassical theory.

    Even more obviously, the local blogger's requirement that theories should know their own range of validity is beyond ridiculous. Should Newton's theory have been rejected as "unscientific", as the blogger boldly claims, because it didn't know (in the 17th century) that it breaks down at speeds close to 300,000 km/s or for angular momenta as small as Planck's constant (i.e. because Newton didn't know relativity and quantum mechanics)?

    It is obvious that it was not possible to know the limitations in advance. Only a more complete theory - or its sketch - can estimate the corrections and limitations of all of its (previously successful) approximations. Hossenfelder's dictum to know the range of validity in advance violates causality, common sense, logic, and pretty much everything else that is completely essential in science - and any other rational thinking, for that matter.

    ReplyDelete
  30. Hi Lubos,

    As usual you are assigning opinions to me I have never put forward, only to then make fun of them. This behaviour is really tiresome.

    I have never said that a theory should be 'attacked' because its assumption are unrealistic. I have instead explicitly acknowledged, several times if you'd care to read, that there is a priori nothing wrong with unrealistic assumptions. But you have to justify their use is appropriate for a prediction you have made.

    the local blogger's requirement that theories should know their own range of validity is beyond ridiculous.

    Just that the 'local blogger' has never said anything of that sort. What I have said is if you make an unrealistic assumption that is not fulfilled in reality, you need to explain in which range your assumption is justified and specify the range of validity.

    That does not exclude however your model breaks down as a good description of reality because you are leaving its range of applicability without even knowing, in most cases because you have made an assumption you were not really aware of. But in such a case what you do is you can falsify your model, and hopefully make progress by building a better one. That progress however can only happen if you have a chance to falsify the model. And that you can't do if you haven't shown it should have been a good description of reality in the first place. It really isn't so hard to understand, Lubos.

    Hossenfelder's dictum to know the range of validity in advance violates causality, common sense, logic, and pretty much everything else that is completely essential in science - and any other rational thinking, for that matter.

    Your inability to extract meaning from the simplest statements I make is always entertaining, but rarely insightful.

    Best,

    B.

    ReplyDelete
  31. PS: If the goal is e.g. to explain the behavior of the society and many corporations, neoclassical economics assumes things like perfect competition etc. And it gives the more-or-less right predictions. Some people found the assumptions about "perfect competitions" that were needed for those predictions "unrealistic" or "too idealized" and created alternative theories of "imperfect competition" and all this crap.

    Friedman explains that these alternative theories were never driven by actual evidence but just by a dissatisfaction with the "realism" of the assumptions - and such a motivation for a new theory is simply unscientific. There is no evidence whatsoever that these modified (and thus more awkward) theories give better predictions than the idealized, neoclassical theory.


    There is a whole plethora of evidence neoclassical economics can not explain, for example high and clustered volatility, you find more examples in this nice paper.

    ReplyDelete
  32. Sabine: "I have never said that a theory should be 'attacked' because its assumption are unrealistic."

    LM: Of course that you have not only written it but you have actually offered such attacks yourself, and you are doing so in your newest answer once again:

    Sabine: "But you have to justify their use is appropriate for a prediction you have made."

    LM: No, I don't have to - and no one else has to - justify unrealistic assumptions. The agreement of the predictions of my theory (or a theory) with the observations is the only justification that a theory needs. That's Friedman's point, it's very essential for all of science because it is what distinguishes science from the promotion of pre-conceived prejudices (e.g. in religions), and you're clearly still not getting this paramount point.

    Sabine: "Just that the 'local blogger' has never said anything of that sort. What I have said is if you make an unrealistic assumption that is not fulfilled in reality, you need to explain in which range your assumption is justified and specify the range of validity."

    LM: I just explained to you - using a very transparent and "canonical" example of the relativistic and quantum corrections to classical mechanics - that your opinion that scientists have to know the range of validity of their theories in advance, when they are proposed, is completely absurd. I can't believe you still don't understand it.

    Only a more accurate or complete theory that will supersede the present theory e.g. in 50 or 200 years may know the limitations of the present theory.

    Sabine: "That does not exclude however your model breaks down as a good description of reality because you are leaving its range of applicability without even knowing, in most cases because you have made an assumption you were not really aware of."

    LM: This is a vacuous, meaningless declaration because no scientist can ever know all assumptions that he is making and that will be challenged in the future. It's very clear that if you allow for this loophole, it really destroys everything you have written above because every scientist proposing any theory must realize that some of his assumptions will be challenged in the future even though it's not quite clear which of them they are.

    Only if there is an actual evidence that the theory doesn't work well, one has a scientific reason to challenge the assumptions. Challenging the assumptions of a theory without evidence - because someone considers them "unrealistic" or "too idealized" - is rationally unjustifiable.

    In fact, "idealized" assumptions are often correct exactly. That's almost certainly the case of the postulates of quantum mechanics, among a few other universal principles.

    I am pretty certain that everyone can check - by looking at your article and this thread - that you have written every single silly assertion that you are now denying to ever have written.

    ReplyDelete
  33. Hi Lubos,

    It is pathetic that you once again (as in many previous exchanges) argue by ignoring my replies and insisting that I said things I never said.

    I never said scientists have to know the complete range of validity of their model in advance. I have explained this point, which you seem to have misunderstood, already above. The point is not that you need to know in advance all assumptions that specify the range of validity of your model, some of which you might not even know. The point is you should *not* have assumptions you know to be in conflict with reality without explicitly showing why they are justified.

    Challenging the assumptions of a theory without evidence - because someone considers them "unrealistic" or "too idealized" - is rationally unjustifiable.

    Lubos, we were talking here about the case when assumptions are *known* to be wrong (or call it 'wildly inaccurate') and the case in which there *is* evidence that the predictions are not fulfilled. Best,

    B.

    ReplyDelete
  34. That's an interesting epistemological discussion. I have only two remarks, as a logician:
    1. Can we count economy as a science like, say, physics is?
    2. In physics, an unrealistic assumption, whatever the success of the theory derived from it, is always scrutinized for falsification, and if falsified, the theory is rejected. A theory based-on unfalsifiable assumptions is just not scientific.
    For instance, the constance of the speed of light for various inertial observers, has been thoroughly tested by numerous experiments. It seems on the contrary that some "unrealistic" AND "uncheckable" assumptions in economic theory have been taken for being true because theories based on them were supposedly describing adequately some piece of observation at a given time. But this is just a sophistic pseudo-scientific approach, confusing A implies B with B implies A, which is the very signature of sophism. So, we come back to my first question: is economy a science or a limited engineering ad hoc theorization of some specific and time-stamped data? And if we go further, taking as granted uncheckable assumptions that have been successful to derive some observed piece of time-stamped data, how nonscientific and ideological is any derivation aiming at future prevision beyond these restricted sets of data?
    That's all the problem of Friedman's theories,which have been the basis of the western-world ideology for thirty years, and have just crashed last year...

    ReplyDelete
  35. Incidentally, I guess that the paper you consider "nice" would be just another example that Friedman would consider to be bad science that is in violation of his principles.

    They seem to be full of "normative economics" - the science trying to plan social changes - and talk about the equilibrium in various conditions.

    But there's no real equilibrium in systems that deliberately try to restrict the human potential because these systems will eventually be abandoned, or the societies that won't abandon them will be superseded by others.

    Please feel free to take the Soviet Union as an example. Every country that suppresses the human freedom much more than other, similarly large countries in the world will be squeezed and replaced in a similar way as the Soviet Union.

    Your statement about "high and clustered volatility" as being in violation of neoclassical economics (but explained by something very particular, which you have vaguely described) is extremely bold, but I don't think that you (or the paper) have presented any evidence that these phenomena are created by the free markets rather than by "central interventions" and that your explanation is actually correct, rather than an ad hoc method to support your preconceptions.

    More seriously, I am not even sure whether you're interested in the evidence that would be useful to settle this question or similar questions. Neoclassical economics explains "much by little" while your attempts to poke holes into it are explaining "little by much" because you seem to be supplementing your non-neoclassical explanation of the real world by a new random meme for every situation.

    That's exactly what Friedman and I consider bad science.

    ReplyDelete
  36. Fine, Lubos, it is great to hear you are with Friedman when it comes to the question what 'bad science' is. Have a nice day. Best,

    B.

    ReplyDelete
  37. Dear Sabine,

    you seem to completely lack any trace of scientific integrity. What's the point of your constant denial that you have written what you have explicitly written and you didn't even bother to delete it to hide the traces?

    You have attacked Friedman's very deep and important wisdom - it was clearly the very main point (and arguably the only point) of your whole posting and everything you have written in this comment thread - and instead, you have presented all the nonsense that theories must know its limitations in advance and scientists have to justify unrealistic assumptions, and all this stuff.

    When I just repeat that you have written these things, you start to be irritated. Why? Can't you see that you behave like a formless piece of jellyfish that can't be talked to? You are just like Lee Smolin, it's pretty distasteful.

    Sabine: "Lubos, we were talking here about the case when assumptions are *known* to be wrong (or call it 'wildly inaccurate') and the case in which there *is* evidence that the predictions are not fulfilled."

    This is completely absurd. Neither Friedman nor I was ever talking about theories making wrong predictions (outcomes!) because theories making wrong predictions are falsified, Friedman, I, or anyone else has always appreciated this trivial fact, and I don't believe that you actually believe that Friedman didn't understand this point. I don't know whom you want to fool with your idea that Friedman didn't realize that falsified theories were falsified. Friedman was one of the most brilliant and scientific economists ever, and a Nobel prize winner. Who is gonna believe you that he didn't know that falsified theories are falsified?

    Friedman was talking e.g. about neoclassical economics whose predictions impressively agree with reality but some people don't like the assumptions because they're "too idealized" and don't match their ideas about the "diverse" driving forces behind the economy etc. Except that this perceived "unrealism" is completely irrelevant.

    In the very same way, some people don't like quantum mechanics because its assumptions are literally "unrealistic" because they reject the existence of the "real state of the system" prior to the measurement. Except that the theory is correct and predicts everything it should, so the complaints are irrational and the alternative (Bohmian...) theories are scientifically unmotivated.

    Another example is your criticism of the correct solution to the black hole information puzzle (with the Penrose diagram (1) supplemented by purely quantum, macroscopically nonlocal very weak processes that preserve the information). You have criticized the correct solution as "not sufficiently conservative" - which is the very same thing as Friedman's "unrealistic". In this case, what you're actually bothered by are - again - the assumptions, not the agreement with the evidence after the (nontrivial) comparison of the theory and evidence is made. If Friedman learned all these things about the black holes, he would surely view your criticism of the correct solution to the black hole information puzzle as a violation of his F-twist principle.

    Lubos

    ReplyDelete
  38. Lubos last post is coming to the real bottom of the truth of the ideological intent of his defense of Friedman's epistemologically "suspect" stances.
    Well, the sad thing is that, really smart, brains like Lubos, are blind to the crash of their economical/political ideology, like communists took years to understand the failure of theirs.
    I am not always agreeing with Bee scientific assumptions, especially with black holes, but really, Lubos is becoming a psychiatric case of reality denial, close the one of the GOP people in the US congress...

    ReplyDelete
  39. Hi Lubos,

    The problem seems to be that you are not able to extract meaning from words that are aligned to arguments exceeding one or a few sentences. You take a single sentence or maybe two, or a picture, or maybe even only a part of a sentence, then you invent a story around it, and decide to disagree with it. Just that the point you are then attacking has nothing to do with the point that was made to begin with. This is a logical fallacy also known as the straw man argument and is pretty common, though rarely so obvious as in your writings. I am really not interested in that kind of exchange, because it is completely unconstructive and merely an annoyance. As you say, everybody who has half a brain can easily find out what I actually wrote. I don't bother to 'delete and hide traces' because I have nothing to hide.

    Best,

    B.

    ReplyDelete
  40. Hi Anonymous,

    To answer the question whether or not economy is a science you'd first have to explain what you mean with science. It is certainly not a science in the way physics is for the obvious reasons that the things physicists describe don't care what theories we have about their behavior. But people do. Best,

    B.

    ReplyDelete
  41. Dear Anonymous,

    1) economics is analogous to some disciplines of physics, especially thermodynamics. Everyone who has studied both introductory textbooks to economics as well as thermodynamics is going to agree that the character of the general principle and their "emergent" yet "universal" character is what these disciplines share.

    2) thermodynamics is not a complete theory of everything - it has to be supplemented by some phenomenological data etc. And occasionally, new phases of a material and new instabilities can be found as well.

    However, I think that you have completely misevaluated the reasons behind the downturns - either the recent one or the Great Depression. Friedman had a much more robust explanation of the Great Depression.

    He unfortunately died before the recent mortgage-started crisis but when one looks into his writings and TV programs, he can see that Friedman did understand what problems can be created by the violation of the freedom of markets - e.g. by violations of the principle that only wealthy enough people should be able to afford their house (instead of offering a mortgage to every loser just because she's black because black is a popular color).

    Irrational mortgages driven by anti-market emotions were one reason behind the U.S. downturn. Another reason were complex derivatives etc. - but these things only caused the mess because traders have lost the urge to properly evaluate the risks, and they have lost it because they have learned in the past that if they're in a deep trouble, the father state will save them.

    Unfortunately, the governments did the same mistake too many times in the recent downturn, once again, which has created seeds for yet another downturn in the future. All big downturns are caused by governments' interventions into the free markets.

    Best
    Lubos

    ReplyDelete
  42. Hi Bee,

    IMHO, the big mistake of these last 30 years, is to have believed that economy was a science. This false understanding led to teach a whole generation of firm and political executives to be trained in thinking through Friedman/reagan/Thatcher/Bush's Weltanschauung. Why economy cannot be an exact sciences? Basically because any prediction believed by the agents can affect the outcome. This kind of backreaction needs a fixed-point theory to be taken into account, which economy cannot do. Well, economy is no more an exact sciences than futurology or Marxism! That's why it is an imperfect and highly ideological tool, and no more.
    To go back to Black Hole scientific discussion between you and Lubos, both of you started from the assumption that GR BH do exist, one arguing that the lost information paradox was in the singularity and the over in the horizon. For me, the dubious assumption is that GR BH do really exist. It's like the rational economical agent assumption from Friedman... Look at this contradiction: classical GR BH do esist only in an open universe with an infinite future, because the information losing horizon is a backward shadow of the infinite future. But, by Hawkins radiation an BH cannot exist forever. That's a simple logical contradiction that shows that the real existence of GR black hole, instead of asymptotically would-be black hole is quite dubious as an assumption... We are back to the discussion of Friedman F-twist, and the need to check experimentally and logically any assumption underlying a theory;

    ReplyDelete
  43. Dear Lubos,


    you are talking about scientific integrity? That's the biggest joke in a long time.

    But as you are obviously not able to understand plain English - everyone who is interested in your opinion and wants to applaud your deep insights knows where to find your asylum. It's not necessary that you come here. Actually, with respect to your health, it is a good idea that you don't do it ;-).

    To put it bluntly:

    Were it not for Sabine's patience, I would since long have banned you completely from here. There is no need to welcome you, and I don't understand how anyone with some intelligence could ever assume there is any sene in "discussing" with you. In fact, the only sensible way to deal with you is to stay away from you.

    Next time you hike through the Czech woods, do me a favour and get lost.

    Have a nice day, Stefan

    ReplyDelete
  44. I am with James Grant:

    "I will concede that markets are just as efficient as the people who operate in them. How efficient is that? Let us turn to mortgage finance to find out. Here is a living laboratory in complexity gone wrong. During the long upswing, lenders were happy to credit a highly counterintuitive proposition. They accepted that a clump of low-rated mortgages could be transformed into a highly rated security. Only very smart people could dream up such a wonderful idea. And only educated people could accept it (they are taught to respect great minds, no matter what cockamamie conclusion those minds arrive at). What was the source of this alchemy? The reordering of cash flows to assure continuous payments to the holders of the senior claims. Risk of nonpayment, such as it was, was borne by the holders of the mezzanine and equity tranches. They bore it willingly. The ratings agencies lent their imprimatur to the assumptions and calculations on which the projections were based.

    Investors bought with confidence. It reassured them to know that the ratings agencies deployed default probability generator models featuring a Monte Carlo multi-step default probability apparatus -- whatever that was. Besides, the mortgage scientists and ratings agency quants believed that diversification was their shield and armor. A given residential mortgage-backed security would contain mortgages from every region of the country. The diversity of collateral delivered low correlation in credit performance; they could hardly all default at once. As for house prices, the mortgage scientists observed that they almost invariably went up. And all agreed that home ownership was an unalloyed social benefit.

    And if the lenders, borrowers, ratings agency analysts, investment bankers, appraisers, etc., hadn't been human, the story might have a happy ending. But it won't, because people in markets neglected to judge the effects of their actions on others. Uniquely uninhibited underwriting practices pushed up house prices and thereby coaxed more borrowers to employ greater leverage. Not wanting to be left behind, lenders completed with one another to offer the easiest terms. Loans tumbled out of the origination mills into the securitization factories, emerging as ABS or CDOs, investment-grade for the most part and thereby suitable for widows, orphans and foreign central banks. Thankfully, too, there were hedge funds and proprietary trading desks to absorb the residual credit risk of the lower-rated tranches. Warsh paid tribute to these social benefactors in his talk the other day: "By serving as willing counterparties in a variety of contracts, these institutions, in my view, are serving as a critical linchpin in the development of more complete markets."

    ReplyDelete
  45. I'll also point out that
    - Hedge funds exist
    - Hedge funds operate by finding arbitrage opportunities - i.e., situations where the market has not priced a security "correctly".

    The market is inefficient enough for hedge funds to thrive.

    ReplyDelete
  46. Hi B,

    I have been more lucky with my papers. Only one of my papers received a bad referee report (I did manage to get that paper published in another journal).

    Perhaps Lubos was the referee of a few of your papers?

    ReplyDelete
  47. I completely agree with Friedmanns points.

    As others have pointed out, the motivating example is neoclassical economics which has rather dubious assumptions at its core. Otoh it explains the data better than any other theory on the market, including the situation we are currently in (which is just straightforward Keynsian liquidity trap scenarios which is incorporated in neoclassical economics by construction).

    Other in principle dubious assumptions exist in physics that turn out to be true. For instance that quantum mechanics is predicated on probabilities or that light takes the path of least time rather than least distance etc. Sometimes you can understand these assumptions from a deeper theory, but its almost always the case that these nontrivial assumptions typically imply deeper more penetrating observable consequences.

    ReplyDelete
  48. Hi Haelfix,

    it explains the data better than any other theory on the market, including the situation we are currently in

    Could you maybe point me to a reference where I can find which data is explained by neoclassical economics to which accuracy? This is not a rethorical question, I would really like to know and I couldn't find anything very useful on this so far. Also, to repeat what I said earlier, don't you think that predicting volatility to be a factor 10,000 smaller than what is actually the case is not a particularly great prediction? Best,

    B.

    ReplyDelete
  49. Hi B,
    Try a standard textbook on macroeconomics, for instance:

    http://www.amazon.com/Principles-Economics-Student-Gregory-Mankiw/dp/0324224729

    Which has the benefit of being very readable unlike other classical textbooks like Samuelsons.

    As for which problems exist in neoclassical synthesis' well, the example you listed is probably one of about 100. Its far from complete, macroeconomics is a complex system. However it is the consensus view, and has successfully modeled and predicted more scenarios than any other theory on the market. For instance, the Great depression and stagflation are typical phenomenons where previous theories all failed to adequately model.

    ReplyDelete
  50. Hi Haelfix,

    Thanks. It wasn't the intention of this post to criticize neoclassical economics, just the way of thinking that underlies this quotation. I would agree that neoclassical economics works well in some cases, but has shortcomings in others. What I was pointing out here was the need to specify in which limits idealized assumptions hold. Your above comment about 'dubois' assumption misses the point. We are here not talking about unintuitive assumptions, but assumptions that are known to be just wrong. Best,

    B.

    ReplyDelete
  51. Friedman was not interested in deep theoretical scientific understaning. Friedman was interested in using "economic science" to justify economic policies hated by millions of people across the planet. And if they don't grasp the assumptions of Friedmans's theories, the CIA -and if this didn't work -the US Marines can be used to educate these people about Friedman's free market truth's.

    Prior experience with Friedmanian economics by ordnary humans is not included as one of Friedman's scientific assumptions:the more experience with Friedmanian economics the less enthusiasm for it.

    The Russian people recently had a large dose of it. Strangely, they didn't seem to appreciate it. Maybe it had something to do with the shortened life expectancies and population collapse of the Russian population. What a bunch of cry babies.

    ReplyDelete
  52. To make a second answer to the question that B. asked me: "why is not economy a science",I'd say that not only economy is wrong in predicting long term trends because of the back reaction effect of predictions on the outcome, but also its analysis on past events is rather a matter of ideological interpretation than scientific analysis. Just look at the current unwrapping crisis. The main explanations coming from economists is that it is primarily a crisis of liquidity because of bad loans. The two interpretations are competing: first not enough regulation on the baking system has let it gone awry (say, democrats interpretation). Second, too much regulation has not helped the banking system to cure itself (republicans interpretation. Well, some other are going deeper in the past: the Friedman/Reagan/Thatcher/Bush deregulation + globalization/delocalization has been favoring dividends against salaries, so working people have not had their share of the global increase in wealth, and to keep on have had to jump into risky borrowing which led them to unsolvable debts. And why salaries have been sacrificed with respect to dividends and bonuses: well to help fight against work cost inflation. Well, the fight against inflation is a pre-Friedman ideology that was set-up in Europe by the club of Rome in the early 70's. The problem, is that this global policy led to have porrer people in the otherwise rich countries and no resumption of the national or private debts by their weakening through inflation. SO, what's the truth? According to your ideological belonging, you'll choose one of these explanations. Science? Economy is no more a science than politics or sociology!

    ReplyDelete
  53. Hi SFJP,

    I didn't ask why economics is not a science, I asked what you consider a science, because without that the question is ill-defined to begin with.

    But to reply to your alleged argument, you are making the same mistake as Horgan who has put forward a similar opinion in several cases. You are assuming that whatever the system, there have to be backreaction effects that lead the system to break down. Now that might be if you look at the question as an 'in principle' question, because 'in the very long term' the sun will blow up into a red giant and all systems on earth will break down one way or the other. There is however no reason why a system can't be stable for a long time. The reason is simply that people might not have any reason to spoil it. Instead, a decent system should be such that people have a reason to protect it.

    Now look at the current economic system. It has a lot of hiccups. Well, to be cynical, this is only consequential. Because if our economy is off equilibrium chances to make profit are much higher. Equilibrium and rational traders is a very boring thing, and it is very difficult to be smarter or that tiny fraction of a second faster than everybody else. But imagine the economy is not in equilibrium. Now everybody who knows how to twiddle the wheels and has the necessary resources can win. Just that it goes on the expenses of the people who don't play in that game, and somebody eventually has to pay the bill.

    According to your ideological belonging, you'll choose one of these explanations. Science? Economy is no more a science than politics or sociology!

    It is not quite as bad as you make it sound, but not as good as it should be either. For more on this point, see this post. Best,

    B.

    ReplyDelete
  54. Dear Bee,
    Imagine that our economy is a Madoff or Stanford—financiers of fraud—like construction, except with a decay time of a few decades. Maybe what keeps it ticking is wars and new technology. But at its heart it is a Ponzi scheme. That might explain the hole that we are in.

    ReplyDelete
  55. I imagine. And it isn't hard to do. If you haven't done so, you should read Naomi Klein's "Shock Doctrine" (It is somewhat repetitive. If you read the first 100 pages, you know pretty much what she's saying.)

    ReplyDelete
  56. "We are here not talking about unintuitive assumptions, but assumptions that are known to be just wrong. "

    Well I wouldn't say the efficient market hypothesis is wrong, or that rational expectations is wrong. In fact, i'd say their veracity is greatly justified by the success of the neoclassical model to predict and describe complex phenomena with a great deal of accuracy.

    Instead we can say that they are merely incomplete. For instance people try to deform the theory to incorporate irrational behaviour and so forth (treating it as a sort of game theory perturbation) and you can get quite far with that. Whats interesting is that the assumptions are like a sort of thermodynamic limit, and work well in that sense and are surprisingly stable. Indeed, its really a miracle that it works so well at all and doesn't simply output junk.

    Thats the nontrivial content of what Milton Friedman is getting at, nothing more. I'm sure he wouldn't dispute the need to test assumptions rigorously.

    ReplyDelete
  57. Hi Haelfix,

    Whats interesting is that the assumptions are like a sort of thermodynamic limit, and work well in that sense and are surprisingly stable.

    Do you have any reference for that? I think it is pretty much without doubt that real people do not fulfill the requirements of rational expectations. Not only are their preferences often inconsistent, more importantly they might just not have preferences. Note that the standard economics approach requires households to have preferences for all possible goods, at all times, under all possible 'states' the world could be in. This is just plain nonsense. Then there is the rising marginal cost of production which is typically assumed, despite the fact that evidence says most firms operate with constant or falling marginal coast. Etc. But despite that I would agree that such unrealistic assumptions can work, if one can show that deviations from these assumptions do not completely spoil the predictions. Just that I at least don't know of any such study. Best,

    B.

    ReplyDelete
  58. Hi Bee,

    To try answering your last question, I'd suggest we separate "Predictive quantitative" sciences, say physics, chemistry, biology from "interpretative sciences". History, politics, economy and somewhat climate prediction are in this second category. They are mainly based on trying to gather and interpret old data to fit them in some model, and to extrapolate these models to try some fuzzy limited predictions in order to inspire decider's actions. Among these interpretative sciences one would have to distinguish purely interpretative ones from normative ones. There falls economy. Economy does not try only to interpret old data and do some extrapolation into the future, but also tells you how to drive a firm or country economiy. That's were normative ideology comes into play. Well, normative ideology could be a good engineering tool if economic system where in equilibrium, but, like history, all that we know for now is that this is not the case. Real economy is a complex, chaotic and human system, where normative inputs alter things in an unpredictable way, while it might not be the only cause of fundamental unstability.
    Note that some models are more prone to instability than others. Purely free markets ideology leas to very unstable struggle of the fittest to become richer, while socialist ideology tries to stabilize things in a way more predictable and more suited for human needs, but even this desirable last approach breaks down on human greed and corruption.
    However, I prefer socialist approach than free market ones, at least from an ethical point of view. True "hard quantitative" sciences have nothing to do with ethics: see, after billions of euros put in the LHC, it could well be that it is the Tevatron that will discover first the Higgs, if it ever existed.....

    ReplyDelete
  59. Neoclassical economics predicts things within an economy organized around neoclassical model. Or to put it another way,an economy organized around the needs of psychopathic corporations. Not exactly an interesting prediction. Of course, Economies can be organized in different ways.

    Why are the endowed economic chairs at Harvard,Rochester,Princeton and Chicago funded by psychopathic corporate entities? See the documentary "The corporation as a psychopath". Neoclassical econimcs=worship of the corporation"

    ReplyDelete
  60. Hi SFJP,

    I tend to agree with what you say. However, that does not mean that you can't address economic questions scientifically. The issue you are bringing up with normative sciences is that you change the system you are examining. Thus, you better be very careful with stating the assumptions you are dealing with. The problem that ideological convictions can change the system is present mostly because the system considered is incomplete (as I have argued excessively in an earlier post, people's opinions on their economic system do matter, but the present models don't pay any attention to them). The questions here are however not very unlike in physics, that is what are the right observables to look for that are independent of things you don't know how to fix. Here as always, you need to ask the right questions to get useful answers. Best,

    B.

    ReplyDelete
  61. Dear Bee,

    Your remark that ideological assumption can change the system because the system is incomplete strikes me, who is a logician, as a very valuable insight. In fact, incompleteness (in the sense of Gödel) seems to be intimately related to the power of expression of a symbolic system, when it is able to represents itself. In other words, when a language/system can be its own meta-language/system, you're pretty sure to run into incompleteness troubles, which is the case of most of the mathematics, and philosophy as well. This is surely also the case of economy, in which besides basic observational rules, which rules are themselves related to the observables an ideology can define, you get in addition normative know how on how to change or modify these rules. I believe physics is more like a first-order language. You can be wrong in your questions, but you can't change nature's rules by whatever means except if you are God Himself. That's a big difference, economists try to play as if they were God, physicists are just humble researcher of invariant rules, which does not prove they are always asking the right questions. Well, for what I know, because I read recently that some people think that by observing the universe astrophysicists can unwillingly accelerate the rate of the universe's expansion, which is anyway beyond my capabilities of understanding.... Best

    ReplyDelete

COMMENTS ON THIS BLOG ARE PERMANENTLY CLOSED. You can join the discussion on Patreon.

Note: Only a member of this blog may post a comment.