Very topical, Pedro. A few remarks, interleaved:
At 01:58 PM 10/29/2008, Pedro C. Marijuan wrote:
>Dear FIS colleagues,
>
>Some aspects of the current financial crisis might be related to
>discussions we had in this list on information and the nature of
>economic flows years ago ("economic networks", and also, central aspects
>of ecological "ascendancy").
>
>The amazing growth of financial assets of many kinds during last decade
>may have conduced finally to a brutal crisis like the current one, not
>just for "greed" or political lack of control, but also for dearth of
>scientific understanding. I would argue that:
>
>1. Financial flows are "anticipatory" information flows that preclude
>the structural changes and the evolution to follow by real economic
>structures.
This seems correct in general, but in the current situation a lot of
the response is reactive.
>2. Without financial anticipation, economic changes could not keep pace
>with technology & science progress due to the "viscosity" of social and
>legal webs of relationships.
I think there is also a regulative role in distributing risk according to
risk tolerance and greed. Futures markets (and other derivatives
like hedges) are a bit like predators, with prey the primary market.
In predator-prey relations, predators can even out booms and busts
in the prey. This is well documented in ecological work (e.g., the
introduction of wolves to islands like Grand Mannan in New Brunswick
and Isle Royale in Lake Superior). However things are not quite so
simple, since predators can increase to much, leading to a drop in
prey and a lagging drop in predators (lynx and rabbits in the Canadian
subarctic are a classic example). Now imagine that the secondary
market becomes much larger than the primary market. Perilous,
I would say, especially if it is unregulated.
>3. The creation of successive informational (financial) layers becomes
>an exercise in complexity accumulation, that almost inexorably leads to
>cross instability thresholds and a general loss resilience.
Right, as above. There is an additional problem with complex derivatives,
though: the information about them is obscure. Neo-classical economics
relies on perfect information. My friend Don Ross has invetigated
neo-classical economics in some depth (some relevant books at
http://mitpress.mit.edu/catalog/author/default.asp?aid=238) across
a number of applications. We discussed applying information theory
with an eye to understanding the role of imperfect information
in evolutionary game theory, but I found that nobody really knows
what economic equilibrium is (Nash equilibria and Pareto optimality
are only a small part of the story). However, if the market is not
ideal, then its evolution is highly path dependent.
This is from a letter I wrote to Don some time ago:
> Dear Don,
>
> I hope this gets to you in time. This is off the top of my head. It isn't
> as organized as I would like it to be.
>
> State Functions and Non-Equlibrium Systems
>
> In traditional physics we deal only with conserved quantities like energy,
> mass, spin, charge and the like. In this case the conservation laws serve
> as the reference for calculations of change. The conservation laws permit
> the use of the Hamiltonian formulation of physics. The fundamental
> quantities are also guaranteed to be state functions. In fact their sum
> over all component systems is an invariant. This also preserves
> reversibility. The basic idea can be extended to other types of systems in
> which all of the fundamental properties are conserved. Although the
> equations of motion are not linear, conservation of fundamental quantities
> permits linear additivity of these quantities, making analysis of state
> changes relatively easy through approximative methods. This is fairly
> straight-forward. You will recall my claim in my Laplace paper that this is
> the model for not only physics, but for most of modern science.
>
> The extension to equilibrium systems is also straight-forward. In this case
> the usual assumption is that the system moves from one equilibrium state to
> another. Examples are classical thermodynamics, classical economics and
> classical population genetics. In this case, however, there are
> nonconserved quantities, such as entropy and work capacity, money supply,
> and adaptation, respectively (the last is explained in my papers on
> increases in fitness). Note that these are all information based functions,
> even if they are not directly related to energy in an obvious way. Since
> these quantities are not conserved, change can induce sources and sinks,
> and the quantities are not linearly additive. The solution to this problem
> is to look at changes as if they occur arbitrarily close to equilibrium,
> making the changes reversible. This fiction requires that the changes occur
> arbitrarily slowly. In real cases this does not occur, of course. The
> fiction, though preserves linearity and reversibility, and allows us to
> define the nonconserved quantities as state functions. This permits the use
> of Hamiltonian methods for the state functions, even though they do not
> specifically apply to the whole process of change in any realistic
> situation (changes that take finite time). Instead, we look only at the
> endpoints. This is justifiable since the equilibrium state wipes out
> historical information, so the path does not matter. In other words, the
> state change is path independent. In general, state variables are path
> independent.
>
> When we turn to non-equilibrium systems, this fiction is undermined. It is
> well-known that non-conservation undermines the Hamiltonian formulation and
> Hamiltonian techniques (conservation is a central assumption of the
> derivation of the Hamiltonian formulation from Newtonina dynamics -- which
> apply to all successive physical theories, including relativity and Quantum
> Mechanics). However, we can still refer the state variables to the
> equilibrium states. This gives a well-defined reference point for the
> fundamental variables.
>
> The essence of an equilibrium state is equiprobability of the complexions
> of the system (possible microstates) for any given macrostate. This
> minimizes information (maximizes microstate information carrying capacity,
> so most information capacity is hidden from the macrostate). Typically, we
> define the nonconserved state variables as if the nonequilibrium states
> were equally distributed, since the microinformation is not accessible to
> the macrostate (principle of indifference). So we can define the
> information in the macrostate at a nonequilibrium point as the difference
> between the actual entropy and the entropy of a state in which all internal
> constraints are relaxed. Since an equilibrium state contains no information
> about history beyond the macrostate information, the order of relaxation
> does not matter (path independence). This result is due, I think, to
> Kestin, but it can be traced back to Caratheodorie (perhaps to Maxwell in
> the statistical mechanical version). So we have a well-defined actual
> entropy and a well defined maximum entropy given the actual macrostate. We
> can calculate these if we know the complexions of the system, but otherwise
> they have to estimate them by indirect means. The information of the
> macrostate, which is a measure of its work capacity or causal power, is
> given by the difference. In the physical case, this difference relates to
> work capacity when divided by absolute temperature. For pure information
> systems, where the work to be done is sorting, or making of distinctions,
> it is demonstrable that the temperature equivalent is a rate that is
> characteristic of the state of the system. For other cases, I don't know
> yet. This approach, anyway, meets an widespread argument that there is no
> unique maximal entropy or definable entropy for the system -- basically we
> use indifference to specify a pseudo-equilibrium for the actual
> nonequilibrium state, and use Kestin plus the actual state to determine the
> maximal entropy.
>
> Obviously, we must lose something, though, with these fictions. What is it?
> Number one, we lose path independence of the state change. In particular,
> state changes are noncommutative (Robert Rosen). Second, since the entropy
> production in the change of state is both non-zero and, furthermore, the
> end state will have historical information about the path (information that
> is not contained in the macrostate alone), there will be a self-interaction
> of the information in the system with itself, which can alter the dynamics.
> In other words, the changes are almost certainly highly nonlinear. In some
> cases, there is no linear approximation possible. However, we still can
> talk about the entropy and information changes between the initial and
> final states, using the psuedo-equilibrium and fictional maximal entropies.
> This places some useful constraints on possible dynamics.
>
> I hope this is enough for now, and clear enough for your use.
>
> Regards,
>
> John
Neo-classical economics assumes some sort of equilibrium (supply and
demand is perhaps the most basic) throughout the system, so developments
are largely linear, as above. However, if the market is not ideal, then it
is, by definition, not in equilibrium, and the above reasoning holds. I
think this is especially true in cases of imperfect information, though
I would be pressed to fully justify that right now. Unpredictability and
rapid change also produces panic, which leads to irrational choices,
something else that undermines the ideal market. Behavioural
economics can help a bit here, but I think that we need also to have
a better understanding of market dynamics when equilibrium does
not hold. (Saying "Don't panic" is apparently not enough.)
>4. Though the financial info is a sort of virtual builder, a potential
>energy of sorts, it has to suffer "closure" upon the real economy; then
>its excessive flows in out from some sector (eg, housing in some
>strategic countries), amplified in the global complexity, have now
>potential to destabilize the whole financial layers and bring the real
>economy to havoc.
Again, the most likely reason is that the assumptions of neoclassical
economics do not hold due to non-ideal market conditions. The housing
bubble in the US caused trouble mostly because the risks were not
properly represented in the financial markets for derivatives. This
allowed the amplification you mention. This process was aided by
'quant' programs that were set up on the basis of normal markets, and
were not suited to the abnormal conditions (this has happened in
other crashes like the one in 1987 -- bigger percentagewise over one
day -- and the Asian and dot.com crashes), but the obscurity of
market information was not as extreme in these cases, so there
was not as much chance for the problem to spread. I suspect that
panic was the bigger factor. In the 2007-08 panic the information
was simply not available, and rating houses like Moodey's had
helped obscure the real situation by rating collective low rating based
derivatives as high rated because they were distributed by banks
with high ratings (that presumably would not fail), and had real estate
to back them up, so the bottom was far from $0. This did not take
into consideration that the financial companies might have a failure
of cash flow, and have to sell off blue chip stocks, where the value
can drop to 0 for at least some companies (including major banks
and insurance companies). it wasn't so much a scam as self-delusion
made possible by the obscuring of information (and deregulation,
such as the 2007 abolishment of the "upward tick" rule on short
selling that was put in back in 1933 or so -- I suspect the last, by
the way, invalidated the quants in one act, since it changed
market conditions so much).
>5. Economy is an informational systems, in crucial aspects, not well
>explained yet... advancing an "info economics" would be quite timely.
Quite. However, having tried it with a very good economist, it is not
very easy. Definitely worth trying again now that neoclassical economics
does not hold quite so much sway, though.
>Would it be interesting to argue on some of these very roughly penned
>aspects (while our pockets get emptier and emptier)?
It is definitely going to be a while before I can retire. Anybody know
of a nice job where the retirement age is over 60, as it is here? It's
coming up a bit too soon for the markets to recover, I fear.
Cheers,
John
----------
Professor John Collier [EMAIL PROTECTED]
Philosophy and Ethics, University of KwaZulu-Natal, Durban 4041 South Africa
T: +27 (31) 260 3248 / 260 2292 F: +27 (31) 260 3031
http://www.ukzn.ac.za/undphil/collier/index.html
_______________________________________________
fis mailing list
fis@listas.unizar.es
https://webmail.unizar.es/cgi-bin/mailman/listinfo/fis
