Tuesday, March 03, 2009

The Trouble with Economics

I recently enriched my vocabulary with expressions like “marginal product of labor,” “price elasticity,” and “isoquants”. The Edgeworth Box. Partial equilibrium. Certeris Paribus. Hutzelpoh.

And why is that? Because, well, we have an increasing group of people here at PI sharing an interest in complex systems. For reasons you find in your daily newspaper, these meetings have turned into more or less regular discussions about economic models, and I thought opening a textbook on the matter might be a good idea. See, we'll even have a conference here to save the world economy. I gracefully declined being among the local organizers. Not until I can say “asset-backed securities” and “collateralized debt obligations” without blushing.

Thus, one could find last week on the quantitative finance arXiv a paper by Samuel Vazquez on Scale Invariance, Bounded Rationality and Non-Equilibrium Economics, and Time and symmetry in models of economic markets by Lee Smolin. Among other things, Lee discusses the question of gauge symmetries in economic models. For an introduction into the topic, I recommend Eric Weinstein's talk on “Gauge Theory and Inflation,” that you can find at PIRSA: 06050010.

A remark on footnote 11 of Lee's paper. Of course there is private health insurance also in Europe. The difference to the USA is that health insurance companies are required to provide a governmentally set minimum of coverage, and it is mandatory to at least have this coverage (which in Germany is referred to as “insurance duty”- Versicherungspflicht.) The USA is besides South Africa the only developed nation without universal health coverage.

Further, it is unclear to me in exactly which sense education in Europe is supposedly more or less “socialized” than elsewhere. As far as I know, the USA too has a public school system. And though France and Sweden traditionally have not had and still don't have tuition fees for universities, Italy, Spain and the Netherlands do. In Germany tuition fees for public universities have been discussed on and off over the last years, and presently regulations differ from state to state. But there are of course private universities also in the European Union that will happily charge you money for a degree. (I'm not an expert on the European Union just because I have a European passport, please correct me if I'm wrong.)

That having been said, I think what Lee meant is that there are a lot private universities in the USA, the tuition fees are the highest worldwide, and that the health insurance coverage is despite being entirely privatized and not mandatory also more costly and less efficient than in the EU. The consequences of that and the conclusions to draw could fill books, and are hardly appropriately taken care of in that footnote.

51 comments:

Uncle Al said...

Economics is heteroskedasticity: "We were horribly wrong. One more parameter and the model works." Heteroscedasticity is equally good.

Take 200 years of US economic data and assign (n-1) variables for n points - or more. It's a positive feedback system. Delta-epsilon decomposition fails at all scales. Perfect interpolation gives meaningless extrapolation.

Kurzweil, Vinge et al. demand a first derivative Singularity. What if the second derivative is a trough not a peak?

Michael F. Martin said...

The premise for Dr. Smolin's article is wonderful. And his analysis of the problems with neoclassical theory trenchant. But do we need gauge theory to make improvements?

Because economic models include agents who act in accordance with economic models, new economic models will be rate-limited in their impact on economics to the extent that they are difficult to learn.

Why not take some of the insights from this wonderful article and apply them in a less ambitious way to reforming accounting theory?

For example, if prices are not good observables because of gauge invariance, doesn't that imply that financial statements should report both the dollar value of accounts AND the amount of time that it took (on average within the period of the statement) for each dollar in the account to turnover?

In other words, we don't have to teach economists gauge theory to help people run happier companies.

By the way, there is a movement in accounting and management called "lean," which essentially has been emphasizing the importance of frequency-averaged measurements of profitability for a long time. They have had a rough time because this is a systemic problem.

Bee said...

Hi Martin,

Why not take some of the insights from this wonderful article and apply them in a less ambitious way to reforming accounting theory?

I think you'd first have to convince somebody it's good for something. Saying because physicists think gauge invariance is so elegant isn't going to work. Best,

B.

Michael F. Martin said...

What it's good for is showing that breaking financial statements into quarterly statements and annual reports makes it nearly impossible to see when the firm is not in a steady state. Financial statements now have a balance sheet, which represents a snapshot in time of assets and liabilities, and income/cash-flow statements, which represent a time-integrated view of the revenue and expenses.

Why don't financial statements include, in addition, turnover rates that (by implication of gauge invariance) represent observables?

If one believes this theory of gauge invariance, then one should also want financial statements to report invariant quantities if possible, I should think.

This is really important because most people think the problem with how companies are run now is one of corporate governance. That's some of it, but the bigger problem is with the financial statement, the theory behind which is neoclassical.

Bee said...

If one believes this theory of gauge invariance, then

This is science, not religion. I don't believe anything without evidence. And this symmetry is evidently broken in reality, so why should I care?

anonymous snowboarder said...

Bee -

Re health care mandates in the US - it is not entirely true to say there are no mandated services. In fact, a large reason why the cost of coverage varies so much from state to state is that each state is able to say what is required (ie, maternity, psychiatric, dependancy)

Re CDOs, CDO^2,ABS, CPDOs I might be able to put you in touch with one of the capital market specialists for the NY State Insurance Board (my exboss). He could probably shed some light on financial theory, or more likely, the pitfalls of recent financial theories.

Michael F. Martin said...

Galilean invariance could not be directly observed either. But Newtonian mechanics turn out to be a pretty useful approximation for projectiles -- one we still teach undergraduates in physics!

Jean-Philippe said...

I would like to react to the following passage from Lee Smolin's paper:

"Thus, we are inclined to side with the neoclassical economic theorists against critics
who suggest that because human behavior is involved there are no useful regularities
that can be captured by mathematical models of a market."

There is at least one alternative to an agent-based models approach that see some worth in mathematical models, and it is the one provided by Fractal Geometry.

Contrary to agent-based model who assumes various things about human behaviors, a fractal approach will first look at the raw data, and try to characterize the patterns in it, but without automatically linking them causally to human behaviors. A causal link may appear locally and provides an understanding of what's at play in the economic systems, but there is no need for an exhaustive system.

I think that we can actually learned more about economics from such an approach than from one that makes the kind of assumptions we find in agent-based models.

As for adding some gauge symmetries to salvage these models, it seems a bit like adding epicycles to the Ptolemaic cosmology, a change of perspective may be more constructive.

Let me precise that, locally, agent-based models can be very useful. In terms of healthcare for instance, a simple analysis of the relationship between the supply and the demand of a life-important service clearly shows that an unregulated market will drive prices so high that many individuals will be excluded from getting it.

Michael F. Martin said...

This is science, not religion. I don't believe anything without evidence. And this symmetry is evidently broken in reality, so why should I care?

Actually, more to the point in response to this is that if you care, then you should get behind Lee because neoclassical theory represents the theory that cannot be falsified with evidence more than a nonequilibrium statistical model.

Plato said...
This comment has been removed by the author.
smekhovo said...

One small point; South Africa is not considered a developed nation.

Phil Warnell said...

Hi Bee,

“Gauge principle: Nothing in the dynamics of an economic system can depend on the units in which the inventories of the different agents are valued. Moreover the different agents are free to use different units or measures to value their own inventories.
Hence the observables and dynamics of an economy should be invariant under
Transformations.”

‘-from the here cited current Smolin paper

Regardless if gauge theory turns out as being relevant to economics or not, Emmy Noether’s initial insight has certainly been a valuable one in terms of applicability to science in general.. Albert Einstein in a piece published in the New York Times upon Noether’s death expressed this as follows:

“Pure mathematics is, in its way, the poetry of logical ideas. One seeks the most general ideas of operation which will bring together in simple, logical and unified form the largest possible circle of formal relationships. In this effort towards logical beauty spiritual formulas are discovered necessary for the deeper penetration into the laws of nature.”

“[20] S. Hossenfelder, research in progress.”

-also from Smolin’s cited paper

This has it appear that you also will be coming forth with an related upcoming paper and as such perhaps you might have the opportunity to give Dr. Noether a little credit when it comes to from where such ideas sprang.

Best,

Phil

Bee said...

Hi Smekhovo,

Not sure if that makes it better or worse. I got the info from this map. Best,

B.

Bee said...

Hi Snowboarder,

I am somewhat confused. What sense does it make to mandate services if it isn't mandatory to have coverage at all? Is it a mandate of the sort 'somebody has to offer'? Best,

B.

Bee said...

Hi Martin,

Galilean invariance could not be directly observed either. But Newtonian mechanics turn out to be a pretty useful approximation for projectiles

Indeed, thus there is evidence it is a good approximation. I was asking for evidence, esp in the face of evidence for the contrary. Best,

B.

Bee said...

Hi Phil,

Well. "Progress" might have been somewhat too optimistic, but if you follow this blog you'll be up to date. Best,

B.

Aaron said...

The problem with gauge invariance of monetary units is the American Tourist Conundrum:

Consider an American Tourist on the streets of Kota Kinabalu on Borneo. He would get charged five American dollars for a five Malaysian Ringgit taxi ride.

The point is the cost of the product or service does not transform locally linearly with the relative valuation of the currencies. This non-linearity occurs when the agents are not familiar with the relative valuations of the currencies.

Plato said...

The Financial Crisis and the Systemic Failure of Academic Economics*


In fact, if one browses through the academic macroeconomics and finance literature, “systemic crisis” appears like an otherworldly event that is absent from economic models. Most models, by design, offer no immediate handle on how to think about or deal with this recurring phenomenon.2 In our hour of greatest need, societies around the world are left to grope in the dark without a theory. That, to us, is a systemic failure of the economics profession.

Michael F. Martin said...

There are analogies to Noether's Theorems in neoclassical economic theory, including the Coase Theorem and the Miller-Modigliani theorem. Making that explicit ought to help make the case to economists with no physics training.

anonymous snowboarder said...

Hi Bee -

Yes I was speaking of coverage that you would have to buy for yourself though the govt doesn't make you buy. But many states have rules which say certain things must be included in the coverage and this is what pushes prices up, at least in part. While you may think I'm quite off my rocker, I really don't feel the need to be paying extra for psychiatric care I wont use, but my state forces it to be included on the mininum HMO.

Anyways, CDOs are far more interesting than HMOs.

Plato said...

The root of the problem had to be identified in terms of an "E8 object" and it's constituents, and thusly, the complexity of that economic situation understood in a global perspective in relation to that root problem. That "is" the "perfect symmetry" in this case?

Gauge theory

Many powerful theories in physics are described by Lagrangians which are invariant under certain symmetry transformation groups.

All transactions.

This then is a illuminative vision according to a complex relation of data which is related to the symmetry of money(coin.)

TO THE JOKER:

HERE I HOLD A GOLD COIN

What a false illusion thou art to human mind ! How cruelly thou deceivest thy possessor and those who covet thee ! Thou buyest for me by thy betrayal of mankind. Thou didst tax my energy to gain thee, and thy discount has lost to me and my fellow-men the greatest blessings of a continent, as well as the principal products of our toil. Few indeed are they who know and understand thy seductive power. We shall expose thy falseness so that our children shalt not be deceived by thee.

General Observations-Charles Lindbergh,
Banking and Currency and the Money Trust

How is this money conserved in the system and travelled through all phases? Maybe the snow boarder might know?;)

Best,

Bee said...

Hi Plato,

Funny you are mentioning the paper on the systemic failure. I was just reading it yesterday. It's a funny mixture between too many and too few details. I have no clue which audience it's addressed to. Best,

B.

Aaron said...

If you have ever bought a house with a spouse, particularly when you are six months pregnant, you would know that the hypothesis of rational agents is completely wrong. Valuation of commodities is an intrinsically emotional process.

Because valuation is based on human perception, doesn't that make single agent based economic theories useless? That is, human perception is a source of statistical noise in the records of transactions of commodities.

After all when viewed as a purely physical system the trade of commodities should reach a the matter equilibrium given by entropic equilibrium, but we know that the trade of commodities is far outside of thermal equilibrium.

So instead of looking for deterministic models for the specific values of commodities, we should look for statistical models. And instead of looking for causal models of physical effects on valuation, we should look for models of human perception of physical effects.

For example, it is not real scarcity or excess of a commodity that drives valuation, but rather the human perception of scarcity or excess.

If I were to look for symmetries in any economic models I would look for demographic symmetries, such as gender, age group, economic status, and cultural self identification.

Bee said...

Hi Snowboarder,

Thanks for the clarification. I can't say it makes much sense to me, except possibly that the health insurances make money.

However, as far as psychiatric care is concerned, you'd be surprised how fast completely normal people can be in need of psychotherapy. The father or a friend of mine was a subway driver, and somebody jumped in front of his train. He couldn't have done anything to avoid running over the guy, but had a hard time coping anyway. More generally, consider you cause an accident and inadvertently kill somebody. Consider the death of a close family member (especially bad: own children). Consider your house burns down and you get out, but your whole family dies. Consider you learn you have a fatal disease, or have an accident and will be disabled for the rest of your life. Consider victims of rape or other violent crimes. And so on. Best,

B.

Michael F. Martin said...

Because valuation is based on human perception, doesn't that make single agent based economic theories useless?

The answer to that is basically same as the answer given by rational choice theory: yes, there are deviations from rationality, but these deviations tend to cancel out in the aggregate. For every pregnant mother who is irrationally desperate to buy there is some gambler who is irrationaly desperate to sell. In the aggregate, rationality is a fair approximation.

Behavioral law & econ is basically taken up with systematic classification of the cases where the deviations don't cancel out.

Aaron said...

The example you give is pertinent, but given the prevalence of influences like marketing, warfare, crime, educational disparity, corporate governance differences, natural disasters, misinformation, and deceit the cases where the factors don't cancel are much broader then the cases where the factors do cancel.

Hasn't the current economic crisis been instigated by an asymmetry in information about the risk of asset backed papers?

I'm going to stick to my guns on this one and insist that talking about the individual valuations made by individual agents is nonsense, and that the more fruitful approach is to view commodity transactions has samples from a time dependent distribution of psychologically driven valuations.

Economics will only make sense if you understand the demographic structure of your population, whether it is individual people or multinational corporations, and you model the perceptive valuation of each demographic group.

Michael F. Martin said...

The example you give is pertinent, but given the prevalence of influences like marketing, warfare, crime, educational disparity, corporate governance differences, natural disasters, misinformation, and deceit the cases where the factors don't cancel are much broader then the cases where the factors do cancel.

Sorry, but these are the domain of macroeconomics, which somewhat confusingly has nothing much to do with neoclassical theory, which is a theory of microeconomics.

Part of the benefit of working out the dynamics at the micro level is that it might help reveal the connections better.

So I agree with you that more attention must be paid to psychology, but I disagree with the conclusion that there cannot (therefore) be hypotheses that provide good approximations to human behavior without resort to a model of individual perceptions.

For example, why not simply count up how often people engage in particular activities associated wtih consumption or production? That would give you a frequency distribution within a window of time, which would correspond roughly to the supply or demand (as applicable) associated with an activity.

Fourier transforming back into the time domain from these frequency distributions you get the time-dependent dynamics -- minus the phase information.

Aaron said...

I agree, I over stated the impossibility of utilitarian hypotheses in micro-economics.

But the micro economic models appear to have two inhibiting implicit assumptions (that can be addressed through demographic and psychological experiments):

First, that all the agents in a system have access to the same information.

Second, that all the agents will act in the same way.

Current micro-economic dynamical theories are implicitly prefaced by being only hypothesized for a group of exchangeable agents (yet another statistical concept poking its head up). So a micro-economic theory is really an avenue of attack on a psychological theory for a specific demographic of agents. Unfortunately rarely is the demographic group of agents for which the theory applies ever explicitly defined.

Thus, to evaluate the transactions of even a single commodity one has to combine micro-economic models for each demographic of agents acting on the commodity, which minimally requires knowing the demographic structure of the agents, and of course their micro-economic behavior.

And even still this does not present a clear method to handle agents that are not exchangeable. For example in the Southern States Caucasian-American agents are definitely not exchangeable with African-American agents because of sales staff preferences (racism), yet they act on the same set of commodities.

Plato said...

HI Bee,

Maybe it is better to deal with it in context of a situation in recorded history. Not to replay itself, as if in a recession initiated in response to today's economy. But to see what a "new proposal" might reveal under a system that is conserved.

Who is breaking these laws? I know who is.:)

Best,

Aaron said...

Just to show that I'm not blowing criticism of economics out of my ass, here is how to formulate a statistical theory:

Begin with the utility function but turn it into a statistical sample from all possible perceptions of utility conditional on each transaction between agents, call this the preceptive valuation distribution. The way this would read in lay terms would be something like: 90% of the time Bob and Sue are happy with the exchange of 2 of Sue's oranges for 1 of Bob's apples. To determine the general rules governing perceptive valuation one needs to use the symmetries of exchange within demographics.

To find the distribution of prices, or relative values etc...Begin by inverting the conditional probability of perceptive valuation using Bayes' Theorem. This requires postulating a prior distribution on the perceptive valuations, essential an estimate of how likely it is you understand how the agent perceives the value of the transaction. Next marginalize the distribution over all agents, given empirical demographic data (the perceptive valuation of a transaction includes the demographic interaction cross sections, e.g. how likely is a teenager will find value in a transaction of commodities with a with a polar explorer). Using this marginal distribution you then have the probabilities of transactions which you can use to calculate average quantities of the market. Any MCMC implementation in R would do the trick.

The nice part is we didn't have to use dynamical systems specified by locally linear actions, which usually imply finiteness of polynomial moments of statistical samples of the phase space. When for example, the distribution of stock market price changes and volumes do not have finite polynomial moments, but rather have finite logarithmic moments.

Jean-Philippe said...

Aaron,

I agree with your point against the hypothesis of rational behavior, though I think I will even go further than you do.
I would say that it is not that this hypothesis is sometimes not respected, the problem is that it is never valid, I don't think there is one single human that ever behave rationally in relation to buying or selling a commodity.

I back my point on the observation that to engage in any relation (and that includes economic relations) is not an outside activity that one can simply look at, on an utilitarian ground, it always is at some level (sometimes minimal) a definition of ourselves, of our identity, there is an existential dimension in consumption, and this one can never be reduced to rationality.

Interestingly, as you allude to, if the rational hypothesis had any relevance, marketing and advertising would be relatively inefficient, and confined really at a purely factual information. I think everybody can agree that factual information is not really an important content of advertising.

I also agree with your recommendation of looking at economics phenomenon from a statistical point of view (which is what I said in my earlier comment) and possibly used these observations to propose some low-level causal processes. The agent-based models do the opposite, but at the cost of reducing human beings to consuming robots.

Furthermore, it is futile to separate the economic, the political and the psychological, all these areas are intimately interconnected, they actually define each others in a very fundamental way and ought to be studied with an holistic approach, rather than a reductionist one, at least in a first time.

Bee said...

Hi Aaron,

Yes, your suggestion sounds like the right thing to do. It would be interesting to know how much into that direction has already been done. Do you happen to know some references, maybe a review? Though I know that there have been models allowing for heterogeneous agents, the process of learning seems to me entirely incompatible with the neoclassical approach in which agents just do have preferences and they do know everything for all times. They don't update from past experience, they don't learn from other's experience, they just know. This is clearly in stark contrast with reality. Thus, it raises the question how much this deviation of assumptions from reality impacts the conclusions. Best,

B.

Bee said...

Hi Martin,

The answer to that is basically same as the answer given by rational choice theory: yes, there are deviations from rationality, but these deviations tend to cancel out in the aggregate.

Could you be more specific here? They do cancel out for which variables? Though I could imagine they cancel out when it comes to, say, total demand, I would think they have an impact on the process through which equilibrium is reached and how well it converges, if it does at all. Consider a situation in which people are irrational to two extremes while the average itself is totally unpopulated. Does that converge? And if it does, is it Pareto efficient?

Something different: I have a very naive question about the idea of revealed preferences. I could live with the idea that people prefer what they buy (even though I don't believe this in general to be a good or useful assumption) but it seems to me essential for the whole framework they have preferences for things they do not actually buy. After all, they have to 'optimize' over all possibilities and their preferences for those.

Best,

B.

Michael F. Martin said...

Could you be more specific here? They do cancel out for which variables? Though I could imagine they cancel out when it comes to, say, total demand, I would think they have an impact on the process through which equilibrium is reached and how well it converges, if it does at all.

Good questions. Your instincts seem right to me, but remember that neoclassical theory doesn't have much to say about the process by which equilibrium is reached.

A point about Lee's paper that I'm not quite sure I agree with is his disavowal of the relationship between thermodynamics and neoclassical theory. Of course an economy is an open system. But neoclassical theory assumes that all of the unknowns are specified. It's only macroeconomic theory that gets away from "thermodynamics." At least that's how I understand neocloassical theory -- not a steady state; a general equilibrium.

I could live with the idea that people prefer what they buy (even though I don't believe this in general to be a good or useful assumption) but it seems to me essential for the whole framework they have preferences for things they do not actually buy. After all, they have to 'optimize' over all possibilities and their preferences for those.

That's correct. A key assumption is that every economic actor have a ranked list of preferences for economic "goods," a list which is static, complete, and transitive. Utility is maximized by matching the scarce goods up with actors so as to achieve the highest aggregate utility overall.

This is a little off the subject, but another major line of attack on neoclassical theory is based on the fact that the utilities of the actors are not, in fact, commensurable. Often economists acknowledge that limitation on their theory, but then ignore it, treating income as a proxy for utility in analyzing a policy.

Let me recommend Thomas Schelling's *Micromotives and Macrobehavior* as an excellent introduction to rational choice theory. I think he's the best around.

Michael F. Martin said...

Bee,

It occurs to me that the another piece of neoclassical doctrine that your second question is getting at is the concept of shadow prices:

http://en.wikipedia.org/wiki/Shadow_price

These are particularly important in the law & economics literature, where few, if any, preferences are revealed through explicit transactions. Gary Becker has made a career out of analyzing things like the family through shadow prices.

Jean-Philippe said...

Aaron,

I disagree with your methodology, at least to some extent.

You wrote:
"To find the distribution of prices, or relative values etc..."

But we don't need to find that distribution, we already have it, we have the data that tell us that this distribution (for prices) is Levy-stable, or a derivation of it, so why would we want to ignore this fact and replace it with an artificial construction.

Your approach has some merit, but I would apply it in reverse. Instead of starting from a utility function, be it a probabilistic one, I would rather start from the hard data we have, and endeavour to develop a model that satisfies their actuality.

Economics academia needs to acknowledge the idea that the value of a model does not lie with the reasonableness of the assumptions or their mathematical convenience, but with the fit to known data.
And if these data are indicative of a dynamic structure, any static model can only fail.

Kaleberg said...

One big difference between economics and physics is that in physics every theory has to pass a common sense test. You can propose any wild cosmological principle you want, add hundreds of dimensions and symmetric particles, postulate bizarre forces and strange new types of matter and energy, but whatever you do has to leave Newtonian mechanics as a useful approximation at familiar scales. Economics does not have this constraint, so you often find mathematically sophisticated economic theories that assume people will work longer hours at the 7-11 after they win big in the lottery or that people will build factories to sell products to people who cannot afford to buy them.

Even the most outlandish string theorists know that their theory has to have water running downhill. Even the most conventional economist is willing to abandon that sort of reality check in pursuit of an elegant theory.

To be honest, I think there is a great deal that economists could learn from physicists, but it isn't what a lot of people think.

plato said...

The Way We Were and What We Are Becoming

Georg said...

"And when the Pharisees saw it, they said unto his disciples, Why eateth your Master with publicans and sinners? "
A modern version would read:
"Why eatheth Bee with bankers and
their jesters?"
:=)
Georg

Bee said...

Coz the publicans and sinners are more fun ;-p

Bee said...

Hi Martin,

remember that neoclassical theory doesn't have much to say about the process by which equilibrium is reached

Yes, that's what I was aiming at.

I think I read half of Schelling's book. It is very nicely written, but I was missing equations. Maybe I should give the second half a try.

treating income as a proxy for utility in analyzing a policy.

In macroeconomics? I hadn't quite realized that. Another question: I am still somewhat confused about the actual relation between micro- and macro-economics. I have been reading some chapters of Wicken's book on macroeconomic theory, but he doesn't even address the question in how far it's derived from the micro-level. Otoh the ingredients look strikingly similar anyway, and I wonder how hard can it be to go from 1 to N households if they are all the same anyway (representative agent!) or trade with themselves on and island (and similar jokes, sorry, I think this is really funny). The relevant difference that I can find is one of policy and financial instruments, ie external ingredients added. Best,

B.

Bee said...

Hi Jean-Philippe,

And if these data are indicative of a dynamic structure, any static model can only fail.

I might be totally wrong here, but to my understanding the neo-classical model is not static. It does have a time-dependence. It's just that this time-dependence isn't dynamical in the sense that it's not generated through a differential equation. Instead, this time dependence is decided upon once and for always at one moment in time because households and firms have perfect information and preferences about all and everything at all times and for all states the world can possibly be in. That having been said, this time-dependence however doesn't seem to be sufficient to describe the dynamics that is actually observed. Best,

B.

Michael F. Martin said...

n macroeconomics? I hadn't quite realized that.

Here's where I need to reveal my own limitations: I know very little about macro. I learned most of what I know about economics at the University of Chicago Law School (not as improbable as it might sound). So-called "law & economics" is all about applying rational choice and neoclassical theory to implicit markets, with a few recognized caveats (like the Coase Theorem, which is really more about transactions costs than equilibrium) thrown in. My frustration with what I was taught began when I realized how poor a model neoclassical theory is when applied to innovation -- my field of law. About a year ago at a conference I heard an antitrust expert talk about how antitrust law needed to become more "dynamic," in the tradition of Schumpeter. Since then I've been very interested in how time is treated in micro theories of innovation. I should be more careful about how I describe how time is treated in micro theory. As you point out, it's not strictly speaking static. Rather it's time invariant in the absence of new information. Some folks (like the efficient market hypothesis guys Fama and French) think this caveat is minor, but it seems rather like the exception that swallows the rule in a field like innovation!

With that caveat, I will say that I understand enough macro to see that many macro theories take a more explicit and realistic approach to time. To that extent, having a better dynamic theory of micro equilibrium would be very useful at identifying the relationship between successful micro and macro.

The macro theorist that I seem to find the most useful is John Taylor at Stanford. There appears to be some political valence at least to how his ideas are received (a complicating factor in sorting the wheat from the chaff in macro theory), but I haven't let it bother me since when I read his work the political valence doesn't seem to originate there.

The other macro theorists who I love aren't recognized as macro theorists at all. They're Warren Buffett and Charlie Munger. If you want empirical evidence that they have a better sense of macro than most, check out their track record. Buffett has beat the market for so many years running that the efficient market theorists have had to accept that he actually does know or understand more than the market!

Buffett in turn based his ideas on Benjamin Graham.

In a nutshell, the difference between their approach and the approach of most micro economists is that they focus on the internal dynamics of the firms that they invest in. They look at market prices only to see whether the prices match up with their own discounted cash-flow analysis of how the company will grow.

Munger's contribution was to point out that if you like holding companies for a long time (like Buffett does) then it's better to pay a little too much in the short-run for the companies that will grow at the same rate at every scale -- i.e., that will grow exponentially. Figuring out exactly how and when a company will continue to grow at the same rate at successively larger scales over time is the "secret" to their success. Buffett in particular showboats a bit by pretending like it's fun and easy. Obviously, it's not easy...

Aaron said...

Jean-Phillipe,

You're right that I haven't fully explained the process of model fitting and predictive testing. I raised the example of looking for the distributions of relative values as an example of a testable prediction one could make, in the parlance of Bayesian Theory it is the predictive posterior.

To get to that point one would need to propose statistical models for the perceptive valuation random variable (i.e. the utility function as a random variable, that in part represents two sources of uncertainty, first within an individual agent taste is capricious, second the statistical uncertainty in the demographic variables that identify agents). The dependence of the distribution of perceptive valuation on transactions would be determined by parameters which one can determine through MCMC priored on empirically observed purchases.

Bee

You are one step ahead of me, I'm doing a lit search today. I'm looking for linked terms like economics, Bayesian analysis, psychology, and sociology. It should turn up a lot of hits. I'm sure marketing agencies have worked out models similar to this, and protect their proprietary models vigilantly.

Aaron said...

Bayesian Analysis in Econometrics - I just started reading. It sounds like an adequate starting point at least for setting up empirically derived posteriors on the natural parameters of the preceptive valuation random variables.

I think reading in Health Economics might be fruitful too.

The hard part will be proposing parameterized families of preceptive valuation distributions. I think that will require reading a lot of psychology literature.

I apologize for my heavy use of statistics jargon.

Exciting!

Plato said...

Dear Bee,

I had to explain myself so I hope you don't mind.

Secondly I hope your readers have had a chance to listen to that radio program that I drop into your discussion as anonymous plato.

It had to ring true to some of you while you "access the American dream." It did become a useful tool to take advantage of a population. If you had never sat in front of a banker for pre-approval how would you know?

That the okay given was a permission to go ahead while recognizing that the financial picture had to be in order, in order to negotiate from a most advantageous position. So couples dream, and are taken advantage of.

How many of you recognize a feudal system in the denigration of a economic system, that has been reduced too, is beginning in the European states? To see that these countries have the families whose farm lands will allow them to successfully barter for a better lifetime.

How many of them recognize the black market that can overtake a country and allow gangs and mafia to become the rule of the land, while it infiltrates governments and feeds the upper 1%, has indeed been reduced to a feudal system?

So I would say while there is this acknowledgement of the scepticism and cynicism that can become part of the intellectual elite, it is by remembering the "roots of labour" that they will become part of that society "who need to be in place" while it produces new models, as well as, leaders, to ensure this justice takes place in application.

So the work then is to prevent such feudalisms in becoming part of the fabric of culture. Not by the products(law enforcement and armies) of NeoLiberalism to maintain it's agenda.

Best,

watzabatza said...

trouble with economics? big losses? is it because of corruption, i think...

Jean-Philippe said...

Hi Bee,

Yes, you are right, there is indeed a dynamic dimension to neoclassical economics via its time-dependence, and I also agree with you to say that this feature is however falling short of accounting to what is really at play in actual economic processes.

To clarify a bit my position, I think the assumption of rationality is fundamentally wrong, and it is this hypothesis that jeopardizes the ability of the model to be truly dynamic (and with a better fit with what is observed), at a conceptual level.
I believe the dynamical nature of economic processes goes beyond the determination (or even the approximation) of a rationally conditioned utility function, it involves culture, psychology, demographics, social and political considerations, and none of this dimension is negligible, not even in the sense of an offsetting at an aggregate level.

Cheers

Jean-Philippe said...

Hi Aaron,

You may be interested in the following article who actually applies something akin to the methodology you recommend, and I believe it has some relevance to some cases:
http://www.e-m-h.org/Mand66.pdf

However, if you intend to refine this approach with the ambition to account satisfactorily for economic processes at a macro-economic level, I think you may soon find yourself facing a very unpleasant choice:
In order to ensure the realism of your model, your utility function will have to be a random process, or even a compound of several random processes that may not be independent, you will also face some challenge in defining workable measures for rather complex parameters that are usually understood in pure qualitative ways (such as the culture of your agents), all this will be necessary in order to reach a reasonable procedure to assess your distribution.
Furthermore, MCMC already assumes that the dependence is Markovian, however, that may not be the case, it may well be scaling, this is actually what is observed in financial data (for a detailed criticism of the systematic usage of Markov Chains in science, see Mandelbrot: Fractals and Scaling in Finance,1997).
Eventually, the difficult choice will be between simplifying your hypothesis so much that they will lead you to a classical Brownian Motion type of conclusion, which is basically where neo-classical theory have led to, followed by financial institutions (and Black, Merton Scholes type of arguments), or you will have to change the approach, and opt for an analysis of prices in terms of fractals, without any apriori assumption of causality, though some conclusions in causal terms may be obtained, but only a posteriori, after the completion of a purely mathematical analysis.

Cheers

jal said...

WOW!!
I find this discussion out of touch with reality.
----
Did you listen to:
http://kpfa.org/archive/id/48892
Re: Guns and Butter

Now go and read :
http://en.wikipedia.org/wiki/Neo-liberal
neoliberalism

Question:
Why does reality spoil a good theory and a good model?

Tkk said...

They tried to outsmart Wall Street:

http://www.nytimes.com/2009/03/10/science/10quant.html?_r=1&em