posted
Is there any significant research being done in applying thermodynamics, particularly entropy, to macroeconomics? Quick google searches find a bit of work in the 70s, but nothing that my untrained eye would see as substantive.
Posts: 2926 | Registered: Sep 2005
| IP: Logged |
posted
After the Long Term Capital Management meltdown, I don't know that people are interested in applying advanced mathematical models to finance again for a long time. A Nova program about itPosts: 11017 | Registered: Apr 2003
| IP: Logged |
posted
You'd have to be specific about what you mean by 'applying thermodynamics'. There are a number of people who look to learn from physics, but as pooka indicates, people have backed off of just taking physics models and assuming they'll work with economics because their results looks similar.
Instead, economics is focusing more and more on microfoundations; building up larger models based on basic assumptions about individual and small group behavior.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
quote:Originally posted by fugu13: You'd have to be specific about what you mean by 'applying thermodynamics'. There are a number of people who look to learn from physics, but as pooka indicates, people have backed off of just taking physics models and assuming they'll work with economics because their results looks similar.
I should preface all my comments by saying I have only incidental exposure to economics and thermodynamics. They just have both had relevance to my research in engineering systems for estimation and control.
The thought was spurred while I was thinking about global economics and wondering if wealth diffusion obeyed scientific entropic principles, and we're headed toward some maximum entropy, evenly distributed fixed point.
I was also wondering whether populations with increasing wealth disparity (and hence decreasing entropy) are stable.
Further free-thought: is all "wealth" created through natural processes? If so, is there some transfer function from, say, the sun's energy to "wealth?" I guess in a far out way I could see maybe money as a potential energy that gets transferred around, or gets created through some transfer from some other energy expenditure (solar emissions, or oil reserves, or human labor). I don't know, none of this is very well formed. I've just seen some success applying principles of entropy, diffusion, and thermodynamics to information systems and I wondered if the same could hold true in economics.
quote:Instead, economics is focusing more and more on microfoundations; building up larger models based on basic assumptions about individual and small group behavior.
To go with another physical analogy, that seems like trying to model the universe based on the properties of sub-atomic particles. I would think there would be a problem, however, in the model's robustness to errors in the properties of the basic building block. For instance, if you assume absolute rationality in every economic interaction, when in fact you only have rationality in 99.9% of interactions, I imagine the models would end up diverging significantly by the time you reach a macroscopic level. Or am I misunderstanding the principle? Or perhaps the scale factor isn't sufficiently large to make the sorts of parameter deviations significant.
Posts: 2926 | Registered: Sep 2005
| IP: Logged |
posted
Wealth is not analogous to entropy in that way, from what we can tell. Population stability (or at least, measures one might consider indicative of that) seem at best loosely correlated with wealth disparity, and almost entirely not correlated when controlling for population homogeneity.
'Wealth' is created by someone valuing something they have. If there are more of things that people value, there tends to be more wealth, but that's the least of the mechanisms for creating wealth. This means wealth can change even if the total amount of stuff in the world remains unchanged. In particular, whenever mutually beneficial exchange happens (presumably every truly voluntary exchange is beneficial to at least one party and harmful to neither, absent unforeseen circumstances), there is more wealth despite the amount of stuff in the world not having changed at all.
While 'stuff' is no doubt an important part of wealth, and one could argue that even personal preferences come from people who are ultimately made out of stuff and powered by the sun, trying to reason economically from the latter is facile, and from the former is incomplete.
Economics models are not nearly as exact as physics models. Economists construct models they know to be wrong in attempts to find things they can say that are mostly right. They are also not so much attempts to model what will happen (though sometimes that can be done), but attempts to model why something has happened so that we can learn from it for the future.
Economists predict outcomes by drawing on models, but except in relatively constrained circumstances (usually including that predictions are very short term) rarely predict outcomes by reporting the output of a model.
Be warned about the papers you read in information theory, btw. Much as it might sound like some of them are using physics, most that sound like it probably aren't, beyond some minor equation-borrowing and talking about analogous concepts that have been given the same name (information theoretic entropy measures are not actually much like physics entropy measures beyond the superficial, for instance).
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
Ball, P., November 2002. The physical modelling of society: a historical perspective. Physica A: Statistical Mechanics and its Applications 314 (1-4), 1-14. doi: 10.1016/S0378-4371(02)01042-7
Abstract:
quote:By seeking to uncover the rules of collective human activities, today's statistical physicists are aiming to return to their roots. Statistics originated in the study of social numbers in the 17th century, and the discovery of statistical invariants in data on births and deaths, crimes and marriages led some scientists and philosophers to conclude that society was governed by immutable “natural” laws beyond the reach of governments, of which the Gaussian “error curve” became regarded as the leitmotif. While statistics flourished as a mathematical tool of all the sciences in the 19th century, it provoked passionate responses from philosophers, novelists and social commentators. Social statistics also guided Maxwell and Boltzmann towards the utilization of probability distributions in the development of the kinetic theory of gases, the foundation of statistical mechanics.
posted
It is an interesting paper, but it doesn't really touch on anything there's been a mentioned interest in.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
quote:To go with another physical analogy, that seems like trying to model the universe based on the properties of sub-atomic particles.
I actually see similarities between the boiling map of quantum events and the rippling of space time by gravity. But perhaps that's intentional.
Posts: 11017 | Registered: Apr 2003
| IP: Logged |
posted
Thanks, fugu. I'm a little disappointed that there isn't more there than I'd thought. Maybe I'll give it some time and reading and come back with a more reasoned relation.
quote:Originally posted by fugu13: Be warned about the papers you read in information theory, btw. Much as it might sound like some of them are using physics, most that sound like it probably aren't, beyond some minor equation-borrowing and talking about analogous concepts that have been given the same name (information theoretic entropy measures are not actually much like physics entropy measures beyond the superficial, for instance).
I'll let you know how much I agree with this statement in a few months, when I'm deeper into my research. On the face, I take a somewhat oppositional view, that I think there's a lot more there than immediately meets the eye, and that there's a lot to be gained from seeking to find and exploit the common structure of both. But that's a fairly uninformed stance at the moment, based solely on quite limited understanding of both fields.
Posts: 2926 | Registered: Sep 2005
| IP: Logged |
posted
Oh, I think there's something to be gained, just don't imagine that when you're reading an information theoretic paper using a lot of physics terms, what you're reading about is anything beyond mildly analogous to the physics in question.
Also, many of the fancy information theoretic things people come up with based on physics theories are outperformed by naive bayesian learning, regression analysis, or other everyday statistical techniques. (edit: this is particularly a common problem in the information retrieval field, which likes such problematic measures as TF-IDF).
Part of this is that there are a lot of physicists finding it hard to get on high profile physics projects, and are instead publishing in other fields. While many of them are making solid and important contributions, many of them are rehashing tired ground and making elementary mistakes, too.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |
quote:Originally posted by fugu13: Also, many of the fancy information theoretic things people come up with based on physics theories are outperformed by naive bayesian learning, regression analysis, or other everyday statistical techniques. (edit: this is particularly a common problem in the information retrieval field, which likes such problematic measures as TF-IDF).
Sure. I'm certainly a Bayesian at heart, and all the methods for inference and prediction that I've seen/used are motivated from a strict statistical basis. That said, there are (IMO) a lot of interesting commonalities being discovered between statistical mechanics and, for instance, optimal state estimation. From what I can tell, it's all about exploiting the underlying Markov structure. Whether the states are physical or information, the same sort of probabilistic transitions are happening, and can be measured (and directed) using the same sort of entropic metrics.
Posts: 2926 | Registered: Sep 2005
| IP: Logged |
Printer-friendly view of this topic
