arXiv:cond-mat/0012479 v2 19 Jan 2001

Stability of Pareto-Zipf Law in Non-Stationary Economies

Sorin Solomon, Racah Institute of Physics, Hebrew Uni ersity of !erusalem an" Peter Richmon", #epartment of Physics, $rinity %olle&e #ublin ', Irelan"

Abstract (enerali)e" Lot*a-+olterra ,(L+- mo"els e.ten"in& the ,/0 year ol"- lo&istic e1uation to stochastic systems consistin& of a multitu"e of competin& autocatalytic components lea" to power "istribution laws of the ,200 year ol"- ParetoZipf type3 In particular, when applie" to economic systems, (L+ lea"s to power laws in the relati e in"i i"ual wealth "istribution an" in the mar*et returns3 $hese power laws an" their e.ponent are in ariant to arbitrary ariations in the total wealth of the system an" to other en"o&enous an" e.o&enous factors3 $he measure" alue of the e.ponent 4 235 is relate" to built-in human social an" biolo&ical constraints3

%6N$EN$S 23 7ac*&roun" on Lo&istic E1uations an" on Power Laws 8 '3 $he (enerali)e" Lot*a -+olterra 9o"el 8 :3 $he Pareto ;ealth "istribution in (L+ 8 53 $he social an" biolo&ical constraints< stability 8 =3 Hea y $ails of 9ar*et Returns in (L+ 8 >3 %onclusions 8 ?PPEN#I@ Econo"ynamics s $hermo"ynamicsA 9ar*et Efficiency s3 $hermal E1uilibriumA Pareto s 7olt)mann laws 8 References 8

1. Background on Logistic Equations and on Power Laws

Aoki [1998] and Aoki and Yoshikawa [1999] have emphasized the importance of the logistic equation 1.1 dw dt ! A w " # w $ in generic economic s%stems. &he% interpret w as the total product demand in a market. &he linear term Aw represents the fact that the emergence of new products is proportional to the present size of the market. &he nonlinear term " # w $ e'presses the fact that the products have to compete with one another for a finite total potential market. (quation 1.1. was used for a ver% long time under the name of #ass) formula *for a review see +aha,an et al 199-. to parametrize the spread of new methods products concepts in a market *see /olomon et al $--- for a multi"agent spatial generalization.. 0owever1 Aoki and Yoshikawa emphasize that the logistic equation 1.1. admits a much wider range of interpretations. /olomon and 2ev% [1993] have suggested that w can represent the total capital within a financial s%stem. 4n this interpretation1 the first term represents the average returns that the s%stem offers1 while the term "# w $ represents the effects of competition and other growth limiting factors. 0of5auer and /igmund [1998] have studied similar s%stems in the conte't of evolutionar% games and genetic d%namics. 4n fact1 equations of the t%pe 1.1 were introduced long ago in population 5iolog% 5% 2otka [19$6] and 7olterra [19$3]. 4n this conte't1 w represents the population size1 Aw the aggregated effects of 5irth and natural death1 while " # w $ represents the effects of the competition for limited resources. 4n its discrete form1 the logistic equation had a crucial role in the stud% of chaos *+a% 19891 :eigen5aum 1981.. Aoki and Yoshikawa quote +ontroll [1988] to the effect that ;almost all the social phenomena1 e'cept in their relativel% 5rief a5normal times o5e% the logistic growth;. A non less universal *and until recentl% un"related. propert%1 spanning a wide range of disciplines from linguistics to economics and to 5iolog%1 is the presence of scale" invariant pro5a5ilit% distri5utions [see /tanle% et al. 1998 for a review]. &his propert% was initiall% o5served 5% <areto [1898 sic= ] in the conte't of individual wealth distri5ution> in each econom%1 the fraction <*w. of people owning a wealth w is proportional to a power of w> 1.$ <*w. ? w " 1"

:or the last hundred %ears the value of ? @ $ changed little in time and across the various capitalist economies. &he emergence of <areto power laws 1.$ in d%namical s%stems with random multiplicative d%namics was known since a long lime in a variet% of fields> the frequenc% of words in te'ts [Yule 19$9]1 economic growth [Ai5rat 19@11 Bhampernowne 196@]1

cities populations [Cipf 1999]1 wealth distri5ution [4,iri and /imon 1988]1 renewal stochastic processes [Desten 198@] etc. <ower laws are a crucial phenomenon in the emergence of macroscopic features in s%stems consisting of man% su5"components. 4ndeed1 the power laws impl% that the sizes of the su5"components span man% orders of magnitude. &herefore1 the presence of the power laws constitutes a 5ridge 5etween the microscopic structure of a s%stem and its macroscopic emergent features *see 5ook 5% 2ev%1 2ev% and /olomon $---.. &he fact that the sizes of the su5"components are so different invalidates the usual ;mean field; approach that aggregates all the su5"components into a ;representative; component governed 5% a single aggregate equation. 4t was shown [/olomon and 2ev% 19931 2ev% and /olomon 19931 +alcai et. al 1999] that in fact1 s%stems of the t%pe 1.11 when studied at the microscopic agents level rather than in the aggregate form 1.11 lead to power law distri5utions of the form 1.$. &hese Aeneralized 2otka"7olterra *A27. s%stems [/olomon 19981 $---] treat each component of the s%stem individuall% while taking into account their non"linear interactions. A27 e'plains not onl% the u5iquitous emergence of the power laws in man% fields 5ut also their sta5ilit% in generic s%stems with non"stationar% d%namics and ar5itraril% var%ing total size [#iham et al 19981 #lank and /olomon $---]. 4n particular1 A27 e'plains the measured values of the e'ponent of the <areto wealth distri5ution in terms of the social and 5iological constraints on the econom% [/olomon and 2ev% $---]. Ene can therefore sa% that the careful reconsideration of the s%stem 1.1.1 led to the solution of a 1-- %ear old puzzle 5% a 86 %ear old equation. 4n the ne't section we introduce the A27 model1 its various interpretations and show how it reduces to a set of decoupled stationar% linear stochastic equations. 4n section @ we derive anal%ticall% the <areto law for the relative wealth distri5ution in the A27 model. 4n section 9 we discuss the sta5ilit% of the <areto e'ponent 5ased on 5iological and social constraints and on the anal%tic pro5a5ilit% distri5ution found at section @. 4n section 6 we relate via the A27 d%namics the <areto individual wealth distri5ution to the market returns distri5ution. /ection 3 discusses the generic features that A27 implies for economic s%stems. 4n the Appendi' we relate the properties of the economic s%stems as descri5ed 5% A27 to the properties of statistical mechanics s%stems.

2. The Generalized Lotka -Volterra Model

&he d%namics of the A27 s%stem in the discrete formulation is defined 5% the recursive formula [/olomon and 2ev% 931 /olomon 981 #iham et al 981 /olomon $---]> $.1 w i *tF . ! r i *t. w i *t. F a w,t. "c*w1t. w i *t. where r i *t.)s are random num5ers *of order unit%. distri5uted with the same pro5a5ilit% distri5ution *independent on i. with a square standard deviation G of order . &he functions a and c*w1t. are of order too in order to insure a meaningful ;continuum limit; " -.

4f one considers w i *t. as the individual wealth of the agent i1 then the random multiplicative factor r i *t. represents the random part of the returns that its capital w i *t. produces during the time 5etween t and tF . &he coefficient a e'presses the auto"catal%tic propert% of wealth at the social level1 i.e. it represents the wealth the individuals receive as mem5ers of the societ% in su5sidies1 services and social 5enefits. &his is the reason it is proportional to the average wealth. &his term prevents1 as we shall show1 the individual wealth falling 5elow a certain minimum fraction of the average. &he e'act mechanism 5% which this happens *su5sidies1 minimal insurance or wage1 elimination of the weak and their su5stitution 5% the more fit. is not1 at this level of description1 important. &he coefficient c *w1t. controls the overall growth of the wealth in the s%stem. 4t represents e'ternal limiting factors> finite amount of resources and mone% in the econom%1 technological inventions1 wars1 disasters etc. 4t also includes internal market effects> competition 5etween investors1 adverse influence of 5ids on prices *such as when large investors sell assets to realize their profits and cause there5% prices profits to fall.. &his term has the effect of limiting the growth of w *t. to values sustaina5le for the current conditions and resources.

c *w1t. parametrizes the general state of the econom%. &ime periods during which "c*w1t. is large and positive correspond to 5oom periods during which the wealth is on average increasing. <eriods during which "c*w1t. is negative correspond to recessions1 when t%picall% the investments lead to negative or small returns. &he surprising fact *proven in section @. is that as long as the term c*w1t. and the distri5ution of the r i *t.)s are common for all the equations $.1 *for all i)s.1 the <areto power law 1.$ holds and its e'ponent is independent on c*w1t.. Ene can also look to the term c*w1t. as an e'pression of the inflation. 4f one thinks of w i as the real *as opposed to numerar%. wealth of each individual1 an increase of the total numerar% in the s%stem w*t. means that an agent with individual wealth w i will loose due to inflation an amount proportional to the increase in w and proportional to ones own wealth> "c w*t. w i*t.. A different interpretation of A27 ma% consider the market as a set of companies i!11.....1H whose shares are traded at varia5le prices wi*t.. &he price of each stock wi*t. is proportional to the capitalization of the corresponding compan% i* the total wealth of all the market shares of the compan%.. 4n this case1 r i *t. represents the fluctuations in the market worth of the compan%. :or a fi'ed total num5er of market shares1 r i *t. also measures the relative changes in the individual share prices. &hese changes are t%picall% fractions of the nominal share price *measured in percents or in points.. aw represents the correlation 5etween the worth of each compan% w i and the market inde' w*t.. "c w*t. w i*t. represents the competition 5etween the companies for the finite amount of mone% in the market *and limits their worth.. &ime variation in the glo5al resources ma% lead to lower or higher values of c that in turn lead to

increases or decreases in the total *or average. wealth w*t.. Yet another interpretation of the A27 equation $.1 is in the conte't of the investors herding 5ehavior> wi*t.. is the num5er of traders adopting a similar investment polic% position *the% constitute a ;herd; i.. one assumes that the sizes of these sets var% auto"catal%ticall% according to the random factor r i*t.. &his can 5e ,ustified 5% the fact that the visi5ilit% and social connections of a herd are proportional to its size. the aw term represents the diffusion of traders 5etween the herds. &he nonlinear term c *w1t. represents the general status of popularit% of the stock market as a whole. &his term also includes the competition 5etween various herds in attracting individual traders as mem5ers.

7er% une'pectedl%1 man% of the properties of the nonlinear s%stem of coupled differential equations with time"depenent *and varia5le"dependent. coefficients $.1 can 5e studied anal%ticall%. &o do this1 let us first take the average in 5oth mem5ers of $.1 and get> $.$ w *tF . ! * r *t. F a. w *t. " c *w1t. w ,t. where w and r are averages over i. &he equation $.$ reduces in the continuum limit to a differential equation of the form 1.1. &herefore1 at the aggregate level1 the s%stem $.1 represents the same s%stem as 1.1. 0owever1 the detailed representation $.1 allows one to uncover properties that would 5e impossi5le to guess from contemplating 1.1. &he equation $.1 can 5e written> $.@ w i *tF . " w i *t. ! [r i *t. "1] w i *t. F aw,t. "c*w1t. w i *t. 2et us assume further that the average $.9 s ! *r*t. "1. !I r i *t. "1 is of the same order of magnitude with $.6 G! I r i $ " r$ ? I *r i "r.$ ! I [i *t.] $ where $.3 i *t. ! r i *t. " r Hote that

$.8 I i *t. ! Jith these notations1 $.@ 5ecomes> $.8 w i *tF . " w i *t. ! [i *t. F s ] w i *t. F a w ,t. "c*w1t. w i *t. and consequentl% *assuming that in the H " infinit% limit the random fluctuations cancel according $.8 [see however 0uang and /olomon $---1 +alcai et al 1999 and #lank and /olomon $---].> $.9 w *tF . " w *t. ! [s F a] w *t. " c *w1t. w ,t. Hote that $.9 shows e'plicitl% that when aggregated1 the s%stem $.1 reduces in the " limit to the logistic form 1.1 *with the identifications A![sFa] and c*w1t. ! #w). #% introducing the new varia5le $.1- ' i *t. ! w i *t. w*t. and appl%ing the chain rule for the differentials d ' i ! ' i *tF . " ' i *t.1 d w i ! w i *tF . " w i *t. and dw ! w *tF . " w *t.1 the equation $.8 5ecomes *considering $.9.> $.11 d ' i*t. ! 1 w*t. dw i " w i*t. w*t. d w *t. ! [i *t. F s ] ' i *t. F a " c*w1t. ' i *t. " ' i *t. [sF a " c *w1t.] ! [i *t. " a] ' i F a Up to here there was no assumption that the system of w i is in a stea"y state, yet we were able to show that the stochastic "ynamics of the relati e in"i i"ual wealths . i re"uces to a set of i"entical "ecouple" linear e1uations '322 which are in"epen"ent on c,w,t-3 In fact the "ynamics of the relati e wealths "epen"s only on #Ba ,since the a era&e r,t- is substracte" in '322, '3>- an" not on the "etails of the interactions c,w,t- or on the a era&e &rowth r,t-3 $he combination #Ba representin& the ratio between the fluctuations of the speculati e income an" the a""iti e socially insure" income is the only parameter influencin& the relati e wealth "ynamics3 In particular, e en in the presence of lar&e arbitrary time ariations of c,w,t-- an" w,t-, if one *eeps aB# constant, the relati e wealth will e entually reach a time in"epen"ent "istribution that we compute analytically in the ne.t section3 $he approach of this asymptotic "istribution by the . iCs is &o erne" by the e1uations '322 an" therefore is itself in"epen"ent on the &lobal nonstationary "ynamics in"uce" by c ,w,t- ,an"Bor r,t-- on w,t-3 ?ctually the result '322 hol"s for yet a wi"er ran&e of mo"els< $.1$ w i *tF . " w i *t.! i *t. w i *t. F a ,5,w, *t. "c*w11w$1...1 wH1t. w i *t. where 5i are ar5itrar% positive coefficients *we e'tracted an overal factor in a such that

one can assume without loss in generalit% that i5i! 1.. #% multipl%ing each equation $.1$ *for each w i . 5% 5i and summing1 one gets *in the infinite H limit.> $.1@ u *tF . " u*t. ! a u *t. "c*w11w$1...1 wH1t. u*t. where we used the notation $.19 u*t. ! ,5,w, *t. 2et us now perform the change of varia5les> #% denoting $.16 'i *t. ! wi*t. u*t. and using the chain differential rule1 one o5tains given $.1@ and $.1$> $.13 d ' i ! d wi u " wi u$ d u ! = i *t. ' i *t. F a "c*w11w$1...1 wH1t. ' i *t. " ' i *t.[ a "c*w11w$1...1 wH1t.] ! *i *t. " a. ' i *t. F a which is identical to the equation $.11 and therefore shares the same properties highlighted a5ove. Bonsequentl%1 the <areto"like formulae @.18"$.18 hold for the s%stem $.1$. &he proviso to the a5ove results is that the coefficients 5 i are small enough to insure that the random terms in $.1@ genuinel% cancell in the large H limit.

. The Pareto !ealth distri"ution in GLV

2et us write the generic equation> @.1 '*tF . " '*t. ! *t. g *'*t..F f *'*t.. Jithout loss of generalit%1 we can assume @.$ I *t. ! since the non"random part of *t.can 5e a5sor5ed in a redefinition of f " f F I *t. g. 4n order for the noise *t.to 5e relevant as one takes ;the continuum limit; " - we assume the square standard deviation> @.@ G ! I *t. 2 to 5e of order . As a consequence1 we will have to keep in the computations 5elow terms of order *t.2 and therefore will have to keep occasionall% terms of the second order in the

differential d'! '*tF . " '*t.. :or a meaningful ;continuum limit;1 the function f *'. is taken of order while g*'. is of order 1. 4n order to find the as%mptotic pro5a5ilit% distri5ution corresponding to the d%namics @.1 one will tr% to reduce it 5% performing an appropriate change of varia5les> @.9 %*t. ! % *'*t.. to a 2angevin process [Kichmond $--1] with constant *unit. coefficient for the random term> @.6 %*tF . " %*t. ! *t. F , *%*t.. /uch an equation is known to lead to the *+a'well"#oltzmann. stationar% distri5ution [+cLuarrie $---] which is the e'ponential of the integral of the ;drift force; , normalized to G $> @.3 <*%. d% ! e'p [ $ G S , *%. d% ] d% &he time evolution equation for the new varia5le %*t. is related to the one of '*t. @.1 through the chain differential rule *in order to keep the terms of order G we e'pand up to second order in d'.> @.8 %*'*tF.." %*'*t.. ! d% ! !*d% d '. d' F 1 $ d2% d'2 *d'. 2 F etc. ! d% d' ['*tF . " '*t.] F1 $ d2% d'2 ['*tF . "'*t.] 2F etc. ! d% d' [ *t. g *'*t..F f *'*t.] F Gg2 $ d2% d'2F etc. Jhere we denoted 5% etc. the terms in the r.h.s. that vanish faster than in the continuum limit " -. 4n the classical particular case g*'. ! ' and f*'. ! - one has> @.8 '*tF . " '*t. ! *t. '*t. which transforms through %*'. ! ln ' into> @.9 %*tF ." %*t. ! *t. " G $ rather than ,ust naivel% @.1- %*tF ." %*t. ! *t. *see also +aslov1 +arsili and Chang 981 /ornette and Bont 981 #ouchaud and +ezard $--- which parametrize the stochastic term using an e'ponential form> '*tF . " '*t. ! [e'p *t. " 1] '*t.

? [ *t. F G $] '*t. which therefore transforms into %*tF ." %*t. ! *t. Je choose here to parameterize the stochastic terms 5% the simple form @.81 @.11 $.11 which is more directl% related to the parameters used in the discrete numerical simulations of A27 [#iham et al 981 0uang and /olomon $---].. E5viousl%1 in order to 5ring @.8 to the form @.6 one needs to make the change of varia5les> @.11 d% ! 1 g d' Jith this change1 the equation @.8 5ecomes> @.1$ %*tF ." %*t. ! *t. F f *'*%.. g*'*%.. " G $ dg d' According to @.3 [see also Kichmond $--1]1 this means> @.1@ <*%. d% ! e'p [ $ G S *f *'*%.. g*'*%.. " G $ dg d' .d% ] d% Ene can use @.11 to change the varia5les in the integrals and o5tain @.19 <*'. d' ! <*%. d% ! e'p [ $ G S *f *'. g2 *'. " G $*dg d' . g . d' ] 1 g*'. d' ! e'p [ $ G S f *'. g2 *'. d ' " ln g ] 1 g*'. d' ! e'p [ $ G S f *'. g2 *'. d ' ] 1 [g*'.]2 d' 4n order to find the stationar% distri5ution of ' i *t. ! w i *t. w*t. corresponding to the d%namics $.111 all one has to do is to appl% @.19 to the particular case> @.16 f! a*1" '. and @.13 g! ' and o5tain therefore1 according @.19> @.18 <*'. d' ! e'p [ $ G S f *'. g2 *'. d' ] 1 g*'.2 d' ! e'p [ $ G S *a"'. '2 d'] 1 '2 d' ! e'p [ $ G S *a"'. '2 d'] 1 '2 d' ! '1 e'p ["$a *'G.]

with @.18 ! 1 F $a G &he distri5ution <*'. has a peak at ' - ! 1 *1F G a.. A5ove ' -1 the relative wealth distri5ution <*'. 5ehaves like a power law while 5elow it <*'. vanishes ver% fast. Ene can show that for finite H1 the main corrections are a factor which vanishes at '!H *which is consistent with the fact that there cannot 5e an agent with wealth w i *t. larger than the total wealth H w*t..> @.19 <*'. ! ! '1 e'p ["$a *'G.] e'p ["$a *G*1"' H..] and a correction to @.$- ! 1F $[a G " D] [1F D] where @.$1 D ! H "$ F $ ? H "9a G
*1F$a G.

&his implies I 1 if H II e'p *G a. i.e. the wealth gets concentrated in ,ust a few hands. 4n the general case $.1$> @.$$ D !I * i i *t. ' i *t. 5 i *t. .$ where i are random num5ers of standard deviation 1 and average -. /ince the 5)s are normalized1 i5i! 11 for a ver% wide range of conditions *e.g. that the values of 5 are not too inequall% distri5uted.1 D vanishes in the infinite H limit.

#. The social and "iological constraints$ sta"ilit%

Mntil now1 we have e'plained the survival of the power law @.18"@.18 in the presence of large e'ogenous and endogenous changes in the total wealth. Je now relate the constant value of ? @ $ measured over the last 1-- %ears *and for all the ma,or capitalist economies. to the social and 5iological constraints that an% societ% is su5mitted to. &he main idea is to e'ploit the particularities of the wealth distri5ution shape @.18 in order to relate the power deca% of the pro5a5ilit% distri5ution at large wealth to the wealth distri5ution of the poorest. &his is o5viousl% possi5le since 5oth the e'ponent ! 1 F $a G of the large wealth power law and the coefficient $a G in the e'ponential of "1 ' which dominates the low wealth 5ehavior1 depend on the single parameter a G. Bonsequentl%1 the constraints on how poor the poor are allowed *or can afford. to 5e1 determine the power low distri5ution of the upper societ% wealth. &his relation is not limited to the A27 d%namics and takes place for an% d%namics that leads to an as%mptotic power law distri5ution for large relative wealth and to a ver%

sharp deca% at low values. 4n our case note that the deca% of the pro5a5ilit% densit% @.18 as ' " - is e'tremel% fast. 4n fact1 since at all the derivatives are - at '!- the lowest relative wealth ' m is estimated roughl% 5% assuming that there are no individuals 5elow it and that a5ove it1 the power law is fulfilled. &hen one gets ' m from the identit% I'! I ' i *t. ! I wi *t. w*t. ! w w ! 1 which implies> 9.1 I' ! 1 ! [ S' m' d' ] [ S' m'1 d'] ! ["1 *1") ' m 1] ["1 *") ' m ] ! ' m *"1) or> 9.$ = 1 *1" ' m. *+alcai et al 99.. And according @.18> 9.@ ' m ! 1 " 1 ! 1 *1 F 1 $ G a. &his is a reasona5le value for ' m considering that the peak of <*'. @.18 is at ' - ! 1 *1F G a. and that the deca% 5elow this value as '"- is e'tremel% sharp. #ased on 9.$"9.@ one can now give a general scenario of how the internal interests and constraints within societ% lead to the actual value of ? @ $ measured repeatedl% in various economies in the last 1-- %ears. /uppose that in a given econom% the wealth necessar% to keep a person alive is D. Bertainl%1 an%5od% having less than that will have a ver% desta5ilizing effect on the societ%1 so the num5er of people with wealth less then D should 5e negligi5le if that societ% is to survive. 2et us now suppose that the average famil% supported 5% an average wealth1 has in average 2 mem5ers . Blearl% the% will need a wealth of order D21 otherwise the wage earners will tr% to correct the situation 5% strikes1 negotiations1 elections or revolts. Hote that in a sense1 the wealth of the average famil% is the de&inition of the minimal amount for supporting 2 dependents1 since the prices of the prime necessities will alwa%s ad,ust to it> if the average wealth increases so will do the prices of housing1 services1 etc. 4n short1 while the poorest people *who cannot even afford a famil%. will ensure the% do not get less than 1 2 of the average1 the average will almost 5% definition take care that their income is at least 2 times the minimal wealth necessar% for supporting one person. All in all1 we are lead to the prediction that ' m ?1 2 and thus *according 9.$. ! 1 *1" ' m. ? 2 *2"1.. &hese relations fit well the known num5ers for t%pical capitalist economies in the last centur%> famil% size 2 ? @"91 povert% line *5elow which people get su5sidized. ' m ? 1 9 " 1 @ and ? 1.@@" 1.6. &he ke% result we o5tain is1 therefore1 that the relative povert% lower 5ound totall% governs the overall relative wealth distri5ution. &he d%namical details 5% which this distri5ution arises are of course comple' and depend on the interactions in the s%stem.

&he low 5irth rate in some of toda%)s societies might suggest higher values for 'mwith associated higher values for leading to greater equalit% and sta5ilit%. En the other hand1 if the speculative fluctuations G are large1 the social su5sidies1 as measured 5% the coefficient a need to 5e increased in order to ensure that 'm ! 1 *1 F 1 $ G a. *and ! 1F $a G. remain constant. :or e'ample1 energetic stock markets com5ined with stagnant social securit% or pensions ma% lead to a decrease in . &his is a well known effect> a period of large financial fluctuations leads to a significant num5er of ;nouveau riches; which ma% leave man% others far 5ehind financiall%.

'. (ea)% Tails o& Market *eturns in GLV

Je will discuss here some of the financial market implications of the formal results o5tained in the previous sections. 2et us first discuss the fluctuations induced in w*t. 5% the d%namics $.1. &he fluctuations in w*t. are important 5ecause one can think of w*t. as a measure of the total worth *capitalization. of the stock market. &herefore w*t. is proportional to the market inde' and its time variation is related to the market returns> 6.1 K*t. ! ln[w*tF1. w*t.] Er1 assuming 6.$ w*t. ! w*tF1. " w*t. II w*t.1 and e'panding the logarithm in 6.1> 6.@ K*t. ? ln[*w*t. F w*t. . w*t.] ? w*t. w*t. &his quantit% measures the wealth at time tF1 of an agent that invested 1 Gollar in the stock at time t. 4n order to estimate the pro5a5ilit% distri5ution of the returns K*t. as resulting from the A27 model $.1 let us consider here the discrete A27 d%namics in which the individual wealths w i are updated sequentiall%. +ore precisel%1 at each time t1 a random integer i 5etween 1 and H is e'tracted and the corresponding wealth w i *t. is updated according to $.1. *Mpdating one agent at a time means effectivel% that the time is rescaled " H and therefore1 in order to descri5e the same continuum process1 one has to rescale also a1 G 1c in $.1 5% a factor H.. &he change in w*t. effectuated 5% the updating of a single w i at time t will 5e> 6.9 w*t. ! [w i *tF1. " w i *t.] H since the changes due to the contri5utions to w*t. due to all the other agents are null *w k *tF1. ! w k *t. if k is different from i.. Msing $.1 in 6.9 one gets 6.6 w*t. ! [* r i*t. "1. w i *t. F a w ,t. "c*w1t. w i *t.] H

/u5stituting 6.6 in 6.@ and using $.1- ' i *t. ! w i *t. w*t. one gets> 6.3 K*t. ? w*t. w*t. ? [* r i *t. "1. ' i *t. F a "c*w1t.' i *t.] H Ene sees that the returns consist of $ components> one deterministic [a "c*w1t.] ' i *t. H depending on the social securit% polic% a and the state of the econom% c*w1t. and one stochastic which dominates the short time fluctuations>

6.8 K*t. ? * r i *t. "1. ' i *t. H &he stochastic part 6.8 is seen to 5e proportional to the ' i)s and therefore it inherits the stochastic properties of the pro5a5ilit% distri5ution <*'. @.18. 4n particular1 in a wide range of parameters1 the variations K*t. have a power law distri5ution> 6.8 <* K . ? K "1" A random walk with steps of sizes distri5uted 5% the power law pro5a5ilit% distri5ution 6.8 is called a 2ev% walk of inde' . &he sum of man% such steps does not converge to a Aaussian distri5ution as e'pected *5% the central limit theorem. from a random walk with steps of fi'ed scale. Kather1 the sum converges to a universal shape called a 2ev% distri5ution of inde' denoted 5% the s%m5ol 2*K.. 4n a certain range of w the function 2*K. itself 5ehaves as a power law 6.8. Accordingl%1 A27 predicts that the market returns will 5e distri5uted *in a certain K range. 5% a *truncated. 2ev% distri5ution 2*K. of inde' given 5% (q. @.18 */olomon 1998.. &his une'pected relation 5etween the wealth distri5ution and the market returns [2ev% and /olomon 1998] turns out to 5e in accordance with the actual e'perimental data [+antegna and /tanle% 1993]. 0owever1 for larger values of K1 the e'ponent is much larger due to finite size effects [0uang and /olomon $---].

+. ,onclusions
4t is well known and sometimes over emphasized that ill"willed or incapa5le politicians ma% influence economics in the negative wa% 5% preventing people from working and trading or simpl% 5% stealing. A less clear issue is whether good"willed capa5le politicians can do an%thing positive to improve the economic and social welfare of the

citizens. #% anal%zing the economic d%namics from a ver% general point of view we e'tracted in this paper1 features which are common to most economies and which put generic limits on how much *and at which price. one can improve the financial and social realities. (ven from weak generic assumptions on the capital d%namics1 one was a5le o5tain ver% specific predictions on the wa% the social wealth is distri5uted. A crucial assumption was that the capital market is fair1 i.e. equal capitals have equal opportunities. (.g. 5% investing twice 1-- M/G in the same asset one is likel% to o5tain the same output as from investing once $-- M/G *independentl% on the investor)s identit%.. +athematicall%1 this was e'pressed 5% our assumption that there is a unique pro5a5ilit% distri5ution1 independent on i for all the random factors r i *t. and that the same function c*w1t. appears in all the equations $.1 *for all i)s.. Je showed that in such a market1 the wealth distri5ution among the individual investors fulfills a power law @.18. &he e'ponent @.18 has 5een measured repeatedl% in the last hundred %ears and found to 5e a constant of order @ $. &his means that in a s%stem with sa% $6- million people1 the poorest one will have appro'imatel% 9----- times less than the richest one. &he average individual will have roughl% 1----- less than the wealthiest. &hese num5ers are in agreement with the actual ones in the M/ econom%. /ocial securit% initiatives cannot change the <areto ;power law;1 the% can onl% seek to change the value of the e'ponent . :or instance1 if one su5sidizes the poorest citizens in order to prevent the last one to fall 5elow a certain ;povert% line; *sa% a fraction ' m of the average wealth. one is lead to a value of the power law e'ponent 9.$. !1 *1"' m.. &his connection 5etween the relative wealth of the poorest and the wealth hierarch% among of the richest [Anderson 1996] emphasizes the su5tle connections that make financial management of the social ecolog% [2ev% et al 1993] ver% difficult to control and predict. &he value !@ $ a5ove is common to most capitalist economies over most of their histor%. As discussed in section 91 this indicates that having a ratio ' m ! 1 9"1 @ is not the result of the policies actions of the various governments 5ut rather a result of more 5asic 5iological constraints. &he 5alance 5etween ;fair pla%; for the capital and minimal socio"5iological needs of the humans seems to trap the world econom% into a power law wealth distri5ution which determines much of its d%namical and equili5rium properties. Ene sees now that without underestimating the responsi5ilit% of the governments to pursue fair1 humane and efficient policies1 one cannot e'pect them to change in a ver% dramatic wa% the a5ove economic financial realities. 2et us remark that low ' m values that lead to ? 1 have cf. 9.$ a dramatic influence on the stock markets sta5ilit%> ? 1 means all of the wealth 5elongs to ,ust a few individuals. &his in turn leads cf. 6.8 [#iham et al 981 0uang and /olomon $---] to macroscopic fluctuations in the financial indices. 0aving all the wealth concentrated in ,ust a few hands1 implies chaotic insta5ilit% in the markets * in contrast to the case in which the wealth is distri5uted among man% individuals and their various fluctuations average smoothl%.. Ene sees that 5e%ond the humanistic arguments1 a ,udicious social securit% polic% is a requirement of the capital markets sta5ilit% as well. +echanisms similar to the descri5ed a5ove appl% in appropriatel% modified wa%s to

companies and countries [/olomon $---1 /olomon $--1] and esta5lish severe limits to how equalitarian *or how unequal. one can e'pect afford the world econom% to 5e.

-PPE./01 Econod%na2ics )s Ther2od%na2ics3 Market E&&icienc% )s. Ther2al Equili"riu23 Pareto )s Boltz2ann laws
Je have used intensivel% in this paper the formal equivalence 5etween the non" stationar% s%stems $.1 of interacting wi)s and the equili5rium statistical mechanics s%stems governed 5% the universal #oltzmann distri5ution @.3. Ene can take seriousl% literall% this formal equivalence and construct a series of analogies 5etween the two s%stems. &his leads to new connections 5etween known economic and financial facts. (.g one can relate the <areto distri5ution to the efficient market h%pothesis> Je have seen that in order to o5tain a <areto power law wealth distri5ution it is sufficient that the relative returns of the agents are stochasticall% equivalent1 i.e. there are no investors or strategies that can o5tain ;a5normal; returns. &his is usuall% the claim of the 5elievers in the e&&icient 2arket h%4othesis. #% definition an efficient market is a market in which the market pricing mechanism is so efficient that it reaches the ;right price; 5efore an% of the agents can take s%stematic advantage *ar5itrage. of the mis"pricing of one item vs. another. &herefore1 the presence of a <areto wealth distri5ution is a sign of ;market efficienc%; in analog% to the #oltzmann distri5ution in statistical mechanics s%stems whose presence is a sign of thermal equili5rium. 4ndeed ph%sical s%stems which are not in thermal equili5rium *e.g. are forced 5% some e'ternal field " sa% 5% laser pumping. do not fulfill the #oltzmann law. /imilarl%1 markets that are not efficient *e.g. when some groups of investors make s%stematicall% more profit than others. do not %ield power laws [/olomon and 2ev% $---]. +arket efficienc% and power laws can then 5e thought as the short time and long time faces of the same medal phenomenon. &his analog% is consistent with the interpretation of market efficienc% as an analog to the /econd law of &hermod%namics> one can e'tract energ% *onl%. from s%stems that are not in thermal equili5rium one can e'tract wealth *onl%. from markets that are not efficient. 5% e'tracting energ% from a non"equili5rium thermal s%stem one gets it closer to an equili5rium one. 5% e'tracting wealth from a non"efficient market one 5rings it closer to an efficient one in the process of approaching thermal equili5rium1 one also approaches the #oltzmann energ% distri5ution in the process of approaching the efficient market one also approaches the

<areto wealth distri5ution. 5% having microscopic information on the state of the s%stem *5e%ond the knowledge of the macroscopic thermod%namic measura5les.1 one can e'tract additional energ% from a s%stems in thermal equili5rium *e.g.+a'well demons ;gedanken e'periment; [2eff and Ke' 199-].. 5% having detailed private information on a financial market1 *5e%ond the pu5licl% availa5le data.1 one can e'tract e'cess profits if the market pricing is efficient.

References
<. J. Anderson in &he (conom% as an (volving Bomple' /%stem 44 *Kedwood Bit%1 Balif.> Addison"Jesle%1 1996.1 eds. J. #. Arthur1 /. H. Gurlauf1 and G. A. 2an. +. Aoki1 Hew Approaches to +acroeconomic +odeling > (volutionar% /tochastic G%namics1 +ultiple (quili5ria1 and ('ternalities As :ield (ffects1 Bam5ridge Mniv <r 1998 +. Aoki and 0. Yoshikawa1 Gemand creation and economic growth1 M. of &okio 1 Btr. for 4nt)l. Kesearch on the Napanese (con. 1999. B.0. #ennett1 ;Gemons1 (ngines1 and the /econd 2aw.; /cientific American Hov. 1988> 1-8"113. E. #iham1 E. +alcai1 +. 2ev%1 /. /olomon1 <h%s. Kev. ( 681 1@6$ *1998. A. #lank and /. /olomon ;<ower laws in cities population1 financial markets and internet sites *scaling in s%stems with a varia5le num5er of components.; <h%sica A $88 *1"$. *$---. pp. $89"$88. N. <. #ouchaud and +. +O)ezard1 <h%sica A $8$1 6@3 *$---. G.A. Bhampenowne1 (conomic Nournal 3@ *196@. @18. +. N. :eigen5aum1 Mniversal 5ehavior in nonlinear s%stems1 2os Alamos /cience1 11 9" $8 *1981.. K. Ai5rat1 2es in)egalite)s )economiques *19@1 <aris1 /ire%.. N. 0of5auer1 D. /igmund1 (volutionar% Aames and <opulation G%namics1 Bam5ridge Mniv <r 1998. C. :. 0uang and /. /olomon1 e"print1 cond"mat ---8-$31 and to appear in (ur. <h%s. N. #. Y. 4,iri and 0. A. /imon1 /kew Gistri5utions and the /izes of #usiness :irms *Horth" 0olland1 Amsterdam1 1988.. 0. Desten1 Acta +ath. 1@1 *198@. $-8. 0./. 2eff and A. :. Ke'. +a'well)s Gemon> (ntrop%1 4nformation1Bomputing. <rinceton Mniversit% <ress 199-. +. 2ev%1 /. /olomon <ower 2aws are 2ogarithmic #oltzmann 2aws

4nternational Nournal of +odern <h%sics B 1 7ol. 81 Ho. 9 *1993. 696P adap"org 93-8--1 +. 2ev% +1 and /. /olomon *1998.1 <h%sica A $9$1 9-. +. 2ev%1 0. 2ev% and /. /olomon1 ;+icroscopic /imulation of :inancial +arketsP from 4nvestor #ehavior to +arket <henomena; Academic <ress1 Hew York1 $---. +. 2ev%1 H. <ersk%1 and /. /olomon1 ;&he Bomple' G%namics of a /imple /tock +arket +odel; 4nternational Nournal of 0igh /peed Bomputing1 81 1993 http> www.ge.infm.it econoph%sics papers solomonpapers stockOQe'OQmodel.ps.gz . A.N. 2otka1 *editor. (lements of <h%sical #iolog%1 Jilliams and Jilkins1 #altimore1 19$6P 7. +aha,an1 (. +uller1 and :.+. #ass1 Hew product diffusion models in marketing> A review and directions for research. Nournal of +arketing 691 1 *Nanuar% 199-.1 1"$3. G. A. +cLuarrie1 /tatistical +echanics1 Mniversit% /cience #ooks $---. E. +alcai1 E. #iham and /. /olomon1 <h%s. Kev. (1 3-1 1$991 *1999.. K. +antegna and 0. (. /tanle%1 An 4ntroduction to (conoph%sics> Borrelations and Bomple'it% in :inance Bam5ridge Mniversit% <ress1 Bam5ridge1 1999. +. +arsili1 /. +aslov and Y"B. Chang1 <h%sica A $6@1*1998. 9-@ (. J. +ontroll and +. :. /hlesinger1 <roc. Hat. Acad. /ci. M/A 891 @@8- *198$.. K. +a%1 #iological populations with nonoverlapping generations> sta5le points1 sta5le c%cles1 and chaos. /cience1 1831 396"981 *1989.. . 7. <areto1 Bours d))economie politique. Keprinted as a volume of Eeuvres BomplRetes *Groz1 Aeneva1 1893Ssh%P1936.. 7. <areto1 Bours d)(conomique <olitique *+acmillan1 <aris1 1898.1 7ol. $. 7. <areto1 2e Bours d) ) (conomie <olitique *+acmillan1 2ondon1 1898.. /. Kedner1 Am. N. <h%s. 681 $38 *199-. P (ur. <h%s. N. #91 1@1 *1998.. <. Kichmond1 <ower 2aw Gistri5utions and G%namic 5ehaviour of /tock +arkets1 to appear in (ur. N. <h%s $--1. +. :. /hlesinger and (. J. +ontroll1 <roc. Hat. Acad. /ci. M/A *Appl. +ath. /ci.. 891 @@8- *198$. 0. A. /imon and B. <. #onini1 Amer. (con. Kev. 981 3-8 *1968. /. /olomon and +. 2ev%1 adap"org 93-9--$ 1 4nt. N. +od. <h%s. B81 *1993. 896 /. /olomon1 in Gecision &echnologies for Bomputational :inance1 edited 5% A."<. Kefenes1 A. H. #urgess1 and N. (. +ood% *Dluwer Academic <u5lishers1 1998.. /. /olomon1 Aeneralized 2otka"7olterra *A27. +odels and Aeneric (mergence of /caling 2aws in /tock +arkets1 in ;Applications of /imulation to /ocial /ciences; 1(ds> A #allot and A. Jeis5uchP 0ermes /cience <u5lications $---. /. /olomon1 Jh% Go 4mpossi5le &hings Alwa%s 0appenT1 to appear <rinceton M. <ress $--1.

/. /olomon and +. 2ev%1 +arket (colog%1 <areto Jealth Gistri5ution and 2eptokurtic Keturns in +icroscopic /imulation of the 22/ /tock +arket +odel http> arUiv.org a5s cond"mat ---6913 P &o appear in the <roceedings of ;Bomple' 5ehavior in economics> Ai' en <rovence *+arseille.1 :rance1 $---;. /. /olomon1 A. Jeis5uch1 2. de Arcangelis1 H. Nan1 G. /tauffer1 /ocial <ercolation +odels1 <h%sica A $881 $@9 *$---. G. /ornette and K. Bont 1in N. <h%s. 4 :rance 8 *1998. 9@1 0.(. /tanle%1 2.A.H. Amaral1 N./. Andrade1 /.7. #uld%rev1 /. 0avlin1 0.A. +akse1 B.D. <eng1 #. /uki and A. 7iswanathan1 /cale"4nvariant Borrelations in the #iological and /ocial /ciences. <hil. +ag. #1 vol. 881 19981 p. 1@8@. 7. 7olterra [19$3]1 Hature1 1181 668. /. +oss de Eliveira 1 0. de Eliveira and G./tauffer1 (volution1 +one%1 Jar and Bomputers1 #.A. &eu5ner /tuttgart"2eipzig 1999. M. A. Yule1 <hil. &rans. #. $1@1 $1 *19$9. A. D. Cipf1 0uman #ehavior and the <rinciple of 2east (ffort *Addison"Jesle% <ress1 Bam5ridge1 +A1 1999..