1: \documentclass[reqno,dvips,12pt]{article}
2: \usepackage[dvips]{graphicx}
3: \usepackage{amsmath,amssymb,amsfonts}
4: \hoffset=-2.0 true cm
5: \voffset=-2.5 true cm
6: \textheight=25.0 true cm
7: \textwidth=16.0 true cm
8:
9: %\setlength{\textwidth}{16cm} \setlength{\textheight}{25cm}
10: %\addtolength{\topmargin}{-25mm} \tolerance=1000
11: \emergencystretch=10pt
12: \newtheorem{theorem}{\qquad Theorem}
13: \newtheorem{proposition}{\qquad Proposition}
14: \newtheorem{corollary}{\qquad Corollary}
15: \newtheorem{lemma}{\qquad Lemma}
16: \def\K{{\rm K}}
17: \def\P{{\rm P}}
18: \def\Rand{{\rm Rand}}
19: \def\Max{{\rm Max}}
20: \newcommand{\al}{\alpha}
21: \newcommand{\om}{\omega}
22: \newcommand{\wh}{\widehat}
23: \newcommand{\wt}{\widetilde}
24: \newcommand{\cE}{{\cal{E}}}
25: \newcommand{\bfE}{{\bf E}}
26: \newcommand{\bR}{{\bf R}}
27: \newcommand{\pa}{\partial}
28: \newcommand{\ov}{\overline}
29: \newcommand{\vp}{\varphi}
30: \newcommand{\ve}{\varepsilon}
31: \newcommand{\be}{\beta}
32: \newcommand{\la}{\lambda}
33: \newcommand{\ka}{\kappa}
34:
35: \newcommand{\nn}{\nonumber}
36:
37:
38:
39: \begin{document}
40: \title{The law of large numbers for completely random behavior
41: of market participants. Quantum economics}
42: \author {V.~P.~Maslov}
43: \date{}
44: \maketitle
45:
46: \begin{abstract}
47: In this paper, we briefly discuss a mathematical concept
48: that can be used in economics.
49: \end{abstract}
50:
51:
52: %\tableofcontents
53:
54: It has been known since Jacob Bernoulli that averaging in
55: economics is nonlinear and does not obey the rules of ordinary
56: arithmetics. For example, the average of winning or losing,
57: say, \$100,000 is not equivalent to customary life for a person
58: who is not wealthy, since losing could mean their becoming
59: homeless and even going to jail, whereas winning would not raise
60: their living standards high enough to justify the mere risk of
61: losing.
62:
63: Accumulating 51% of stock in a business results in a controlling
64: interest. By pooling their capital, owners can become
65: monopolists and raise prices. This also corroborates
66: the nonarithmetic nature of addition. Further, one observes
67: nonlinear addition when buying goods wholesale: the more you
68: buy, the less you pay per item.
69:
70: The mathematical problem of nonlinear averaging was attacked by
71: the outstanding Russian mathematician Kolmogorov, who obtained a
72: general formula for the average. I supplement his axioms by an
73: additional axiom saying that if a (small) number is added to
74: each term in the sum, then the average increases by same number.
75:
76: This unambiguously results in the following rule for the average
77: of two numbers $a$ and $b$:
78: $(a\text{ ``+'' }b)/2 = 1/\beta\log[(2^{\beta a}a + 2^{\beta b}b)/2]$,
79: where $\beta$ is an unknown parameter and base~$2$ logarithm
80: is used. For $\beta=0$, one arrives at the conventional linear
81: average.
82: It is easily seen that $\beta$ is negative for purchase averaging
83: and positive for sale averaging.
84:
85: If the operation
86: $a\text{ `+' }b = 1/\beta\log [ 2^{\beta a} + 2^{\beta b}]$ is
87: taken as addition and the conventional addition $a+b$ as
88: multiplication,
89: then the commutativity, associativity, and distributivity laws
90: hold. This arithmetics is possibly used somewhere in the
91: ``Kingdom of Distorting Mirrors.''
92:
93: In my opinion, this arithmetics is more adequate to a market
94: economy than the conventional, classical arithmetics and permits
95: one, using model examples, to explain stock price breakout,
96: default, and a number of laws observed in statistical data
97: processing.
98:
99: In this note, I wish to present an economic effect that proved
100: quite unexpected to myself. I speak of a statistical sensation
101: that would be nicknamed ``fool's luck'' by ordinary people.
102: Since 2003, studies related to the analysis of operations at
103: London stock exchange have been evoking quite a response. About
104: five hundred papers that can be found in the Internet under the
105: keywords zero intelligence discuss the problem of completely
106: random behavior of market participants. The point is that
107: traders buying and selling stock have to take into account so
108: many various factors that they cannot make the right decision
109: unambiguously, and so the behavior of market participants does
110: not differ from a random behavior very much. And if it does, the
111: trader usually loses. Here we assume that the sales volume, the
112: number of participants, the number of nomenclatures of financial
113: tools for each price, and the number of tools are sufficiently
114: large.
115:
116: Thus what is a random choice?
117:
118: To make a random choice, one should first calculate all possible
119: choices of purchases or sales and average them. The
120: parameter~$\beta$ can relatively easily be expressed via the
121: budget restraint (BR) in the first case and via the required
122: rate of return (RRR) in the second case.
123:
124: Note that possible choices of purchases and sales obey the
125: so-called Bose statistics. For example, the number of ways to
126: buy nails and bread for 2 cents is as follows:
127: 1) 2 cents for nails; 2) 2 cents for bread;
128: 3) 1 cent for nails and 1 cent for bread.
129: All in all, there are three ways, since bank notes or
130: coins of the same value are indistinguishable.
131: (The law that can be stated as ``money does not smell": nothing
132: changes if one note is exchanged for another.)
133:
134: How to make a random choice? It suffices to number all possible
135: purchases whose price does not exceed BR and then choose a
136: number randomly, as if throwing dice. Computer generates random
137: (more precisely, ``pseudorandom") numbers. It turns out that the
138: law of large numbers holds: given a BR, a vast majority of such
139: random choices gives the same value for the number of purchases
140: at a given price. This value can be determined in a very simple
141: manner via the nonlinear average of all possible purchases by a
142: small variation of the given price. (The number of tools
143: purchased at a given price is equal to the partial derivative
144: with respect to this price of the nonlinear average over all
145: possible purchases.) A similar law holds for sellers.
146:
147: Note that the parameter $\beta<0$ for purchasers can also be
148: determined by equating the derivative of the nonlinear average
149: with respect to $1/\beta$ with the logarithm of the number of
150: possible purchases whose price does not exceed BR. Accordingly,
151: for sellers ($\beta>0$) the derivative should be equated
152: with the logarithm of the number of possible sales that give a
153: profit not less than RRR.
154:
155: Thus if the behavior of market participants is completely
156: random, then we can find the number of goods purchased
157: by a majority for a given price and even estimate the fraction
158: constituted by this majority. Conversely, if we cannot do that,
159: then the behavior of market participants cannot be viewed as
160: random.
161:
162: If the prices can be varied, then, by equating the amount of
163: goods sold as a function of RRR with the amount of goods
164: purchased as a function of BR, we can find equilibrium prices.
165:
166: It follows from an analysis of papers in the ``zero intelligence"
167: series that traders who do not subtilize, i.e., who act at
168: random rather than deliberately, mainly do not lose. It is not
169: without reason that one speaks of fool's luck or beginner's
170: luck: he who gambles for the first time (and hence has not yet
171: been spoiled by calculations) does not lose as a rule. There is
172: a large body of corroborating statistical evidence. Hence if we
173: assume that a vast majority of traders have zero intelligence,
174: then the law established by the author gives a right forecast of
175: free market prices.
176:
177: Let us discuss a similar situation for consumer goods. Suppose
178: that a customer who has a certain amount BR of money for buying
179: gifts in advance for a large number of people enters a gift shop
180: (say, in Mexico) with $k$ shop floors each of which offers a large
181: variety of souvenirs for the same price. The consumer does not
182: know the tastes of all acquaintances whom the gifts are intended
183: for and buys at random. Then the above-mentioned law applies,
184: and one can rather accurately predict how much money the
185: consumer will spend at each shop floor. More precisely, if
186: there are many similar consumers, then a vast majority of them
187: will spend exactly the amount predicted by this law at each shop
188: floor.
189:
190: This is a very simple model, which can be generalized. For
191: example, consider the case in which the goods are divided
192: into~$i$ clusters with close prices within each cluster.
193: Then one determines the (nonlinear) average profit over all
194: goods if $N_i$ goods are sold for these prices and then the
195: average price for the $i$th cluster is found.
196: The above-mentioned law remains valid in this case and
197: determines the amount spent by a customer for a given cluster of
198: goods.
199:
200: In this paper, we briefly discuss a mathematical concept
201: that can be used in economics.
202:
203: \section{Nonlinear averaging in the sense of Kolmogorov}
204:
205: A sequence of functions $M_n$ determines the {\it{regular\/}}
206: type of the average if the following (Kolmogorov) conditions are
207: satisfied:
208:
209: I. $M(x_1, x_2, \ldots, x_n)$ is a continuous and
210: monotone function in each variable.
211: To be definite, we assume that $M$ increases in each variable.
212:
213: II. $M(x_1, x_2, \ldots, x_n)$
214: is a symmetric function\footnote{In our case,
215: the symmetry follows from the Bose statistics
216: for bank notes.}.
217:
218: III. The average of identical numbers is equal to their common
219: value: $M(x, x, \ldots, x) =x$.
220:
221: IV. A group of values can be replaced by their average
222: so that the common average does not change:
223: $$
224: M(x_1,\ldots, x_m, y_1, \ldots, y_n)
225: = M_{n+m}(x \ldots, x, y, \ldots, y_n),
226: $$
227: where $x = M(x_1, \ldots, x_n)$.
228:
229: \begin{theorem}[Kolmogorov]
230: Under conditions {\rm I--IY},
231: the average $M(x_1, x_2, \ldots, x_n)$ takes the form
232: \begin{equation}
233: M(x_1, x_2, \ldots, x_n)= \psi \frac{\varphi(x_1)+\varphi(x_2)+
234: \ldots + \varphi(x_n)}{n}, \label{Kol}
235: \end{equation}
236: where $\varphi$ is a continuous strictly monotone function
237: and $\psi$ is its inverse.
238: \end{theorem}
239:
240: For the proof of the theorem, see~\cite{Kolmog1}.
241:
242: \section{The main averaging axiom}
243: It is rather obvious for a stable system
244: that the following axiom must hold.
245:
246: If the same value $\omega$ is added to $x_k$,
247: then their average increases by this value~$\omega$.
248:
249: Obviously, the nonlinear averaging of $x_i$
250: under normal conditions must also increase by this value.
251: We take this fact as {\bf Axiom~5}.
252:
253: This axiom leads to a unique solution in the nonlinear case,
254: i.e.,
255: the linear case (the arithmetic mean) naturally
256: satisfies this axiom, as well as a unique
257: (up to the same constant by which we can multiply
258: all the incomes~$x_i$) nonlinear function.
259:
260: In fact, the incomes~$x_i$ are calculated in some currency
261: and, in general, must be multiplied by a quantity $\be$,
262: which is responsible for the purchasing power
263: of this currency, so that this constant (the parameter~$\be$)
264: must {\it a priori\/} be contained in the definition
265: of the income.
266: Hence we can state that there exists a unique nonlinear function
267: that satisfies Axiom~5.
268:
269: The function $f(x)$ has the form
270: \begin{equation}
271: f(x)=C\exp(Dx)+B, \label{1kv}
272: \end{equation}
273: where $C,D\ne0$ and $B$ are numbers independent of~$x$.
274:
275: \section{Semiring, an example of self-adjoint linear operators}
276: We consider the semiring generated by nonlinear averaging
277: and the space $L_2$ ranging in this semiring.
278:
279: First, we consider a heat equation of the form
280: \begin{equation}
281: \frac{\pa u}{\pa t} = \frac h2 \frac{\pa^2 u}{\pa x^2}.
282: \label{2kv}
283: \end{equation}
284:
285: Here $h$ is a small parameter,
286: but we do not use its smallness now.
287:
288: Equation (\ref{2kv}) is a linear equation.
289: As is known, this means that
290: if $u_1$ and $u_2$ are its solutions,
291: then the linear combination
292: \begin{equation}
293: u= \la_1 u_1 + \la_2 u_2 \label{3kv}
294: \end{equation}
295: is also its solution.
296: Here $\la_1$ and $\la_2$ are constants.
297:
298: Now we perform the following change. We set
299: \begin{equation}
300: u=e^{-\frac wh}.\label{4kv}
301: \end{equation}
302: Then we obtain the following nonlinear equation
303: for the unknown function $w(x,t)$:
304: \begin{equation}
305: \frac{\pa w}{\pa t} + \frac 12 \left (\frac{\pa w}{\pa x}
306: \right)^2 - \frac h2 \frac{\pa^2w}{\pa x^2} = 0. \label{5kv}
307: \end{equation}
308:
309: This well-known equation is sometimes called
310: the B\"urgers equation\footnote{The usual B\"urgers equation is
311: obtained from this equation by differentiating with respect
312: to~$x$ and applying the change $v=\frac{\pa w}{\pa x}$.}.
313:
314: The solution $u_1$ of Eq.~(\ref{2kv})
315: is associated with the solution
316: $w_1 =-h \ln u_1$ of Eq.~(\ref{5kv}),
317: and the solution $u_2$ of Eq.~(\ref{2kv})
318: is associated with the solution
319: $w_2 =-h \ln u_2$ of Eq.~(\ref{5kv}).
320: The solution (\ref{3kv}) of Eq.~(\ref{2kv})
321: is associated with the solution
322: $w= -h\ln(e^{-\frac{w_1+\mu_1}{h}} +e^{-\frac{w_2+\mu_2}{h}})$,
323: where $\mu_i = -h \ln \la_i$, $i=1,2$.
324:
325: This implies that Eq.~(\ref{5kv}) is a linear equation,
326: but it is linear in a function space,
327: where the following operations were introduced:
328:
329: the operation of taking the sum \ \ $a\oplus b = -h \ln(e^{\frac{-a}{h}} +
330: e^{\frac{-b}{h}})$;
331:
332: and the operation of multiplication \ \ $a\odot \la= a+\la$.
333:
334: In this case, the change $w=-h\ln u$ takes zero to infinity
335: and the unity to zero.
336: Thus, $\infty$ is a generalized zero in this new space:
337: $\O= \infty$,
338: and the usual zero is a generalized unity:
339: ${\bf{1}}=0$.
340: The function space,
341: where the operations $\oplus$ and $\odot$ are introduced,
342: with the associated zero $\O$ and the unity ${\bf{1}}$
343: is isomorphic to the usual function space
344: with the usual multiplication and addition.
345:
346: This can be interpreted in the following way:
347: somewhere on another planet,
348: the people are used to deal with precisely these operations
349: $\oplus$ and $\odot$,
350: and then Eq.~(\ref{5kv}) is a linear equation from their
351: viewpoint.
352:
353: Everything written here is, of course, trivial,
354: and the people on our planet need not study
355: new arithmetic operations,
356: because, using a change of the function,
357: one can pass from Eq.~(\ref{5kv}) to Eq.~(\ref{2kv}),
358: which is linear in the usual sense.
359: But it turns out that
360: the ``Kingdom of distorting mirrors'' given by this semiring
361: is related to the ``capitalistic" economics.
362:
363: In the function space ranging in the ring
364: $a\oplus b=-h\ln(e^{-\frac ah} + e^{-\frac bh})$,
365: $\la \odot b = \la + b$,
366: we introduce the inner product
367: $$
368: (w_1, w_2) = -h \ln \int e^{\frac{w_1+w_2}{h}} dx.
369: $$
370: We show that the product in this space has
371: the following bilinear properties:
372: $(a\oplus b, c) = (a,c)\oplus(b.c)$ and
373: $(\la \odot a,c)= \la \odot(a,c)$.
374: Indeed,
375: \begin{eqnarray}
376: &&(a\oplus b,c)= -h \ln \left ( \int \exp
377: \left(\frac{-(-h\ln(e^{\frac{-a}{h}} + e^{\frac{-b}{h}})+c)}{h}
378: \right) dx\right) = \nn \\
379: &&=-h \ln \left( \int (e^{\frac{-a}{h}} + e^{\frac{-b}{h}})
380: e^{\frac{-c}{h}} dx\right)= -h \ln\left( \int e^{-\frac{a+c}{h}}
381: dx + \int e^{-\frac{b+c}{h}} dx \right)=(a,c)\oplus(b,c), \\ \nn
382: &&(\la\odot a,c) = -h\ln\int e^{-\frac{a+\la}{h}} e^{-\frac ch}dx
383: =\\ \nn
384: && = -h \ln\left(e^{-\frac \la h} \int e^{-\frac{a+c}{h}}
385: dx\right)=\la+\ln\int e^{-\frac{a+c}{h}} dx= \la\odot(a,c).
386: %\label{}
387: \end{eqnarray}
388: We consider an example of self-adjoint operators
389: in this space,
390: namely, the operator
391: $$
392: L:W \longrightarrow W\odot(-h\ln\left(\frac{(W')^2}{h^2} -
393: \frac{W''}{h}\right).
394: $$
395: And now we verify whether it is self-adjoint:
396: \begin{eqnarray}
397: &&(W_1,LW_2) =-h\ln \int e^{-\frac{W_1+LW_2}{h}} dx= \\ \nn
398: &&=-h\ln\int \exp \left[-\left(W_1+W_2-
399: h\ln\left(\frac{W'_2)^2}{h^2} - \frac{W''_2}{h}\right)\right)
400: /h\right] dx= \\ \nn &&= -h \ln\int e^{\frac{-W_1}{h}}
401: e^{\frac{-W_2}{h}}\left( \frac{(W_2')^2}{h^2} -
402: \frac{W''_2}{h}\right) dx = -h \ln\int e^{\frac{-W_1}{h}}
403: \frac{d^2}{dx^2} e^{\frac{-W_2}{h}} dx=
404: \\ \nn
405: &&=-h\ln\int\frac{d^2}{dx^2}
406: e^{\frac{-W_1}{h}}e^{\frac{-W_2}{h}}dx= -h\ln\int
407: e^{\frac{-W_1}{h}}\left(\frac{(W'_1)^2}{h^2}
408: -\frac{W''_1}{h}\right) e^{\frac{-W_2}{h}}dx= \\ \nn
409: &&=-h\ln\int\exp\left[-\left(W_1-h\ln\left(\frac{(W'_1)^2}{h^2}
410: -\frac{W'_2}{h}\right)\right) /h\right] dx= \\ \nn
411: &&= -h\ln\int
412: e^{-\frac{LW_1+W_2}{h}} dx =(LW_1,W_2).
413: %\label{}
414: \end{eqnarray}
415: Its linearity can also be verified easily.
416:
417: We construct the resolvent operator of the B\"urgers equation:
418: $L:W_0\rightarrow W$, where $W$ is a solution of Eq.~(\ref{5kv})
419: satisfying the initial condition $W|_{t=0}=0$.
420:
421: The solution of Eq.~(\ref{2kv})
422: satisfying the condition $u|_{t=0}=u_0$
423: has the form
424: $$
425: u=\frac{1}{\sqrt{2\pi h}} \int e^{-\frac{(x-\xi)^2}{2th}} u_0(\xi)
426: d\xi.
427: $$
428: Taking into account that $u=e^{-\frac Wh}$ and $W=-h\ln u$,
429: we obtain the resolvent $L_t$ of the B\"urgers equation
430: \begin{equation}
431: L_tW_0 =-\frac{h}{\sqrt{2\pi h}} \ln \int
432: e^{-\left(\frac{(x-\xi)^2}{2th} + \frac{w(\xi)}{h}\right)} d\xi.
433: \end{equation}
434: The operator $L_t$ is self-adjoint in the new inner product.
435:
436: \section{Entropy for the producer and the consumer.
437: Condition for the producer income and the consumer expenditure.
438: Production and consumption.
439: Equilibrium prices}
440:
441: We consider a large group of producers
442: manufacturing goods of $M$ types.
443: The corresponding {\it production structure\/}
444: is characterized by the vector
445: $\omega^N=\langle K_1,\dots, K_M\rangle$,
446: where $K_i$ is the number of goods of the $i$th type,
447: $i=1,\dots, M$, and $\sum\limits_{i=1}^M K_i=N$.
448: The {\it consumption structure\/} is treated
449: similarly.
450: Suppose that $\epsilon_i$ is the price of goods of the $i$th
451: type.
452: The income obtained by selling $N$ units of goods
453: is equal to $E=\sum\limits_{i=1}^M \epsilon_i K_i$.
454:
455: The concepts introduced below
456: are based on the notion of nonlinear averaging
457: of the incomes obtained by realizing $N$ units of goods
458: over all possible versions $\omega^N$
459: of the production structure
460: for a given positive value of the parameter~$\beta$:
461: \begin{eqnarray}
462: M_N= \frac{1}{\beta}\log\left(\frac{1}{L}
463: \sum\limits_{K_1+\dots+K_M=N}2^{\beta\sum\limits_{i=1}^n\epsilon_i
464: K_i}\right). \label{average-1}
465: \end{eqnarray}
466: The nonlinear averaging is discussed in detail in~\cite{Mas2003}.
467:
468:
469: \subsection{Psychology of an ordinary depositor.
470: Psychological law of status quo preservation}
471:
472:
473: We note that losing \$100,000
474: is much heavier in its psychological ``cost"
475: than winning the same sum.
476: This means that an ordinary person prefers
477: to preserve status quo,
478: i.e., not to take risk of losing \$100,000.
479: Therefore,
480: if a person deposits a certain sum $N$ in several banks
481: at high interest, he must calculate what sum is sufficient
482: for him to live like a rentier, to preserve status quo,
483: and, accordingly, how to spread the money over several banks
484: so as not to take any risk and not to lose his status.
485:
486: This purely psychological fact
487: is the base of our mathematical calculations.
488: We note that an ordinary depositor can rather easily
489: calculate the sum of income, but it is difficult for him
490: to evaluate the reliability of a bank proposing
491: high interest rates.
492: Therefore, in the formulas given below,
493: the free parameter~$\be$ can be determined
494: in terms of the sum of income introduced above,
495: and then we obtain graphs showing
496: how the bank deposits $n_i$
497: depend on the {\it a priori\/} given income.
498:
499: The problem of calculating the ``surviving probability,"
500: which is close to the ``status quo preservation law,''
501: has been discussed by specialists
502: in mathematical economics.
503:
504: If the required rate of return (RRR)
505: $\sum n_i\la_i= E$ is assumed to be an independent variable,
506: then the nonlinear averaging based on the Kolmogorov axioms
507: and the additional axiom proposed by the author
508: is unique.
509:
510: We assume that production runs normally
511: if the producer's income $\sum_1^n \ve_i N_i$
512: is larger than or at least equal to a quantity~$E_1$.
513: The consumer cannot spend more that
514: $E_2 =\sum_1^n \ve_i N_2$ (the budget restraint -- BR).
515: We consider the ``entropies" of the producer
516: and the consumer as follows:
517: $H_1(\ve_1,\cdots, \ve_n, E_1)$ is the base~$2$ logarithm
518: of the number of sale versions
519: for a sum no less than $E_1$; \
520: $H_2(\ve'_1, \cdots, \ve'_n, E_2)$ is the base~$2$ logarithm
521: of the number of purchase versions
522: for a sum not exceeding~$E_2$.
523:
524: Let $\tilde{M}$ be the sum (\ref{average-1})
525: for $\be <0$, and let $\theta=\frac1\be$.
526:
527: The equation
528: $$
529: \frac{\pa M}{\pa \theta} =H_1
530: $$
531: allows us to obtain $\theta =\theta_1(H_1)$,
532: and the equation
533: $$
534: \frac{\pa \tilde{M}}{\pa \theta} =H_2
535: $$
536: allows us to obtain $\theta =\theta_2(H_1)$.
537:
538: We choose $k$ possible purchase versions at random.
539:
540: It turns our that, for the majority of these versions,
541: the sum of money spent for buying goods at price~$\ve_i$,
542: is ``almost" equal to
543: $$
544: \frac{\pa M}{\pa \ve_i}|_{\theta=\theta_1}
545: $$
546: (this is an analog of the law of large numbers;
547: the exact estimates are the same as in the usual law
548: of large numbers).
549:
550: A similar statement holds for the seller.
551: The equilibrium prices follows from the relation
552: $$
553: \frac{\pa M}{\pa \ve_i}|_{\theta=\theta_1(\ve)}
554: = \frac{\pa \tilde{M}}{\pa \ve_i}|_{\theta=\theta_2(\ve)}, \ \
555: i=1,\cdots, n.
556: $$
557: Thus, the ``resources--producer--consumer--etc.''
558: vertical line is divided into pairs each of which
559: contains two new numbers, RRR and BR.
560: The entropy (the Kolmogorov complexity) and
561: the degree of risk are ``hidden" in the intermediate calculations.
562: This is how the usual model of economical equilibrium
563: (the general equilibrium) varies.
564:
565: The models of dynamical equilibrium
566: (the intertemporal general equilibrium)
567: vary similarly according to formula (41)
568: given in \cite{TeorVer2004}, p.~276.
569:
570: The price equilibrium condition is determined by the relations
571: \begin{equation}
572: \frac{\pa M}{\pa \ve_i}|_{\theta=\theta_1(\ve,E_1)}
573: = \frac{\pa tilde{M}}{\pa \ve_i}|_{\theta=\theta_2(\ve,E_2)}.
574: \label{6kv}
575: \end{equation}.
576:
577: It follows from the ``pair'' law that
578: the derivative of the average
579: with respect to the ``temperature" $\theta$
580: is the entropy
581: and the derivative with respect to the price
582: is the quantity of goods.
583: Relation (\ref{6kv}) is well known in the linear case.
584: Here we generalize it to the nonlinear case.
585: This generalization allows us to take
586: the Kolmogorov complexity into account,
587: and hence the entropy,
588: which is one of the most important notions in economics.
589: Its conjugate, the temperature, determines the degree of risk,
590: and sometimes, the volatility.
591: But the variables in the final formulas
592: contain only the incomes $RRR$ and the expenditures $BR$,
593: which can be calculated easily.
594: Thus,~$H$ and~$\theta$ are ``hidden parameters" here.
595:
596: \section{Tunnel canonical operator in economics}
597:
598: The equilibrium prices are determined by the condition
599: that the demand and offer must be the same
600: for each item of goods and resources.
601: The following pairs can be determined similarly:
602: flows of goods and services -- prices;
603: flows of trade of different kinds -- salary rates;
604: flows of raw material resources -- rents;
605: interest -- loan volume.
606:
607: The asymptotics of $M$ and $\tilde{M}$
608: is given by the tunnel canonical operator
609: in the phase space of pairs.
610:
611: We consider the phase space $R^{2n}$, where
612: the intensive variables play the role of coordinates
613: and the extensive variables play the role of momenta.
614: In economics, the role of values of a random variable~$\la_i$
615: can be played by the prices of the corresponding goods,
616: and~$N_i$ can, for example, be the number of sold goods,
617: i.e., the number of people who bought goods of this particular
618: type, or the interest paid by the $i$th bank, etc.
619: Obviously,
620: the price depends on the demand, i.e.,
621: $\la_i(N_i)$ is a curve in the two-dimensional phase space.
622: In the two-dimensional phase space,
623: to each point (vector) $\la_i$, $i=1, \ldots, n$,
624: there corresponds a vector $N_i (\la_1, \ldots, \la_n)$,
625: $i=1, \ldots, n$.
626: In a more general case, this is an $N$-dimensional manifold
627: (surface),
628: where the ``coordinates" and ``momenta" locally depend on~$n$
629: parameters,
630: and the following condition is satisfied:
631: the Lagrange brackets of the ``coordinates" and ``momenta"
632: are zero with respect to these parameters.
633: Hence such a manifold was called a Lagrangian manifold
634: by the author.
635: In other words, the form $\sum N_i d \la_i$ is closed
636: (see the Afterword in~\cite{Heding} and \cite{TeorVozm65}).
637: This means that $\int N_i d \la_i$ is independent of the path
638: and is called an action like $\int pdq$
639: ($p$ is the momentum and $q$ is the coordinate)
640: in mechanics.
641:
642: The producer acquires resources and transforms
643: the resources expenditure vector
644: into the vector of material wealth production.
645: Then the consumer acquires this material wealth.
646: Thus, the equilibrium prices of resources and
647: of the consumer material wealth
648: are determined according to the above relations.
649:
650: In addition to such equilibrium prices,
651: there can also be vertical pairs for some types of material wealth
652: and seller--buyer pairs
653: (i.e., permanent seller -- permanent buyer pairs),
654: and the prices related to these pairs are formed.
655: This is an analog of the Cooper pairs in quantum statistics.
656:
657: This construction requires the use of the ultra-secondary
658: quantization method in abstract algebraic form,
659: which could be applied in economics.
660: The ``vertical'' clusters are also formed in this theory.
661:
662: Thus, we have determined the parameter $\beta$,
663: and the problem is solved completely.
664:
665: {\bf{Example.}}
666: In what follows, a person who buys stocks will be called a
667: player.
668: Suppose that there are only two types of stocks,
669: the first are conditionally called ``cheap'' stocks
670: and the stocks of the second type
671: are said to be ``expensive.''
672: We assume that a player buys a packet of $N$ stocks
673: in which the number of cheap stocks is~$N_1$
674: and the number of expensive stocks is $N_2=N-N_1$,
675: respectively.
676: A player spends money to buy stocks,
677: and the number of purchased stocks affects
678: the price of stocks of both types.
679: In particular, the larger is the number of expensive stocks
680: bought by a player, the less their price will be.
681: Consequently, in what follows, we assume that
682: the player's expenditures for a packet of stocks
683: depend nonlinearly
684: on the number of purchased cheap and expensive stocks.
685: For example, it depends quadratically as follows:
686: \begin{equation}
687: \cE(N_1)=\lambda_1N_1+\lambda_2N_2-\frac{\gamma N_1^2}{2N}-
688: \frac{\gamma N_2^2}{2N}=\lambda_2N-\frac{\gamma N}{2}
689: +(\lambda_1-\lambda_2+\gamma)N_1-\frac{\gamma N_1^2}{N},
690: \label{1d}
691: \end{equation}
692: where the numbers $\lambda_1$, $\lambda_2$, and $\gamma$
693: satisfy the conditions
694: \begin{equation}
695: \frac{\gamma}{2}<\lambda_1<\lambda_2,\qquad
696: \lambda_2-\lambda_1<\gamma <2(\la_2 - \la_1). \label{2d}
697: \end{equation}
698: It follows from conditions~(\ref{2d}) that the function~(\ref{1d})
699: with $N_1=0,1,\dots,N$ has the global minimum for $N_1=N$
700: and a local minimum for $N_1=0$.
701:
702: \begin{figure}[htbp]
703: \includegraphics[]{RISUNOK2.EPS}
704: \vskip-0.5cm
705: \caption{Graph of $E(N_1)$ for $T=0$}
706: \end{figure}
707:
708: If, at the initial moment,
709: the player buys~$N_1$ cheap stocks
710: and $N_2$ expensive stocks so that
711: $N_1 < \frac{\la_1-\la_2+\gamma}{\gamma} N$,
712: then selling one cheap and one expensive stock
713: so as to decrease $\cE(N_2)$,
714: he will come to a local minimum at $N_1=0$,
715: i.e., he will buy all expensive stocks.
716: But if $N_1 > \frac{\la_1-\la_2+\gamma}{\gamma}N$,
717: then, as a result of a monotone process,
718: the player will buy all cheap stocks.
719:
720: Now we consider local financial averagings
721: of the player income.
722: We assume that $G_1$ dealers sell cheap stocks
723: and $G_2$ dealers sell expensive stocks.
724: In this case, the number of different ways
725: in which the player can buy a packet of stocks is
726: \begin{equation}
727: \Gamma(N_1)=\frac{(N_1+G_1-1)!}{(G_1-1)!N_1!}
728: \frac{(N-N_1+G_2-1)!}{(G_2-1)!(N-N_1)!}. \label{3d}
729: \end{equation}
730:
731: {\bf Remark.} Instead of introducing different dealers,
732: we can assume that
733: the cheapest and the most expensive stocks
734: are, respectively, of $G_1$ and $G_2$ different types,
735: but of the same price.
736:
737: We assume that, for $\be= \infty$,
738: the player is at the point of local minimum for $N_1=0$.
739: Since he tries to change the stocks pairwise and gradually
740: (monotonically) so that not to increase his expenditures
741: \footnote{The least risk principle in economics.},
742: we can consider the averaging
743: only in a neighborhood of the point of local minimum
744: (the local financial averaging).
745: If $\be$ varies slowly and $N\to \infty$,
746: then the asymptotics of $M_\be$ as $N\to \infty$
747: again corresponds to the local minimum
748: \begin{eqnarray}
749: &&\cE(N_1) = \be(\lambda_1N_1+\lambda_2N_2
750: -\frac{\gamma N_1^2}{2N}-\frac{\gamma N_2^2}{2N}) + \nn \\
751: &&+\ln \frac{(N_1+G_1-1)!}{(G_1-1)!N_1!}
752: \frac{(N-N_1+G_2-1)!}{(G_2-1)!(N-N_1)!}. \label{3dd}
753: \end{eqnarray}
754:
755: \begin{figure}[htbp]
756: \includegraphics[width=10cm, height=18cm]{RISUNOK3.EPS}
757: \vskip-0.5cm
758: \caption{Graph of $E(N_1)$ for $T=5$, $G_2=30$, and $\gamma=1.5$}
759: \end{figure}
760:
761:
762: Figure~3 shows how the entropy depends on the temperature and
763: the local and global minima.
764:
765: \begin{figure}[htbp]
766: \includegraphics[width=10cm, height=18cm]{RISUNOK4.EPS}
767: \vskip-0.5cm
768: \caption{The heavy line corresponds to the local minimum,
769: the global minimum lies below.
770: At the point $T=40$, the local minimum and the global maximum
771: coincide and do not exist for $T>40$.}
772: \end{figure}
773:
774: The curve breaks at the point $T\approx 40$.
775: At this point,
776: the derivative $\frac{\pa S}{\pa T}$ becomes infinite,
777: and the modified Laplace method,
778: which could be used for the asymptotics
779: of the local $M_\be$ at the other points,
780: cannot be used near this point.
781: The local minimum obtained by computer
782: shows unstable ``spreading."
783: It turns out that the asymptotics near this point
784: can be expressed in terms of the Airy function
785: of an imaginary argument,
786: and this fact removes all the problems listed above.
787:
788: For $T>40$, no equilibrium can exist
789: when small changes in purchasing and selling occur.
790: Therefore, to return to the equilibrium point
791: formed as a result of changes in the local averaging
792: near another local minimum,
793: the player must change a large amount of stocks at once
794: (see Fig.~3).
795:
796: A similar situation occurs when $\ka=1$
797: and the player wants to win as much as possible.
798: Then the points of minima are replaced by the points of maxima
799: between which,
800: in the case of a quadratic dependence on $N_1$ and $N_2$,
801: there is a minimum.
802: If~$T$ varies from zero to some~$T_0$
803: at which the local maximum disappears,
804: then a jump occurs, which can be treated as
805: a stock price break-down.
806: In our case,
807: if the player simultaneously and very fast
808: sells a large amount of stocks of one type
809: and buys a large amount of stocks of a different type,
810: then he can again get to another equilibrium point.
811:
812: We have considered only the simplest model.
813: In a more complicated economical situation
814: concerning the interests of a great mass of the population,
815: the people cannot change their behavior very fast,
816: passing, for example,
817: from the usual consumer basket to a different basket
818: and thus changing their mode of life.
819: Then there is no equilibrium (balance) point in general,
820: and a sharp disbalance leads to general default.
821:
822: {\bf{Generalization.}}
823: In our example, we obtained a one-dimensional curve
824: corresponding to a local minimum of the entropy dependence
825: on the ``temperature" in the two-dimensional $S,T$-space
826: and considered its projection on the $T$-axis at the
827: point~$T_0$.
828: This projection is not ``good,"
829: and we said that the asymptotics of~$M_\be$
830: in a neighborhood of this point must be replaced
831: by the Airy function.
832:
833: What picture appears in the general case
834: where we have two ``conjugate" pairs:
835: the entropy--temperature pair and, for example,
836: the number of people~$N$ corresponding to
837: some average salary~$\ve_k$.
838:
839: In this case, we consider a four-dimensional (phase) space,
840: where $T,\ve$ are the ``coordinates''
841: and $N,S$ are the momenta.
842: The surface corresponding to our curve
843: is two-dimensional and can (locally) be written
844: in parametric form as
845: \begin{equation}
846: T=T(\al_1, \al_2), \quad \ve = \ve(\al_1, \al_2),\quad
847: N= N(\al_1, \al_2), \quad S=S(\al_1, \al_2), \nn
848: \end{equation}
849: where $\al_1, \al_2$ are parameters.
850:
851: Because this surface must be obtained
852: (at least, at the simple points of projection
853: on the ``coordinate" plane)
854: from the asymptotics of sums of the form
855: $$
856: M_\be = \frac {1}{\ka\be} \ln(\frac{e^{\ka\be a}+ e^{\ka\be
857: b}}{2}),\quad \ka =\pm 1, \quad \be>0,
858: $$
859: it must be a Lagrangian manifold
860: (this notion was introduced by the author
861: in \cite{TeorVozm65}).
862: At the points of ``bad projection" on the $T,\ve$-plane,
863: the asymptotics is given by the tunnel canonical operator
864: \cite{TeorVozm65}, and its simplification,
865: depending on the form of the surface near
866: the point of ``bad projection,"
867: can be obtained using~\cite{ArnoldVar}.
868:
869: These general considerations can help constructing
870: the corresponding model of a given economical situation
871: if the appropriate statistical data are available.
872:
873: \section{The law of large numbers}
874:
875: We consider the sellers who sell goods of $M$ types.
876: The corresponding {\it structure of sold goods}
877: is characterized by the vector
878: $\omega^N=\langle K_1,\dots, K_M\rangle$,
879: where $K_i$ is the number of goods of the $i$th type,
880: $i=1,\dots, M$, and
881: $\sum\limits_{i=1}^M K_i\le N$.
882: The {\it structure of purchased goods}
883: is considered similarly.
884: Suppose that $\epsilon_i$ is the price of goods of the $i$th
885: type.
886: The income obtained by realizing $K_1+\dots +K_n$ units of goods
887: is $E(\omega^N)=\sum\limits_{i=1}^M\epsilon_i K_i$.
888:
889: The concepts introduced below
890: are based on the notion of nonlinear averaging
891: of the incomes obtained by realizing $\le N$ units of goods
892: over all possible versions $\omega^N$
893: of the structure of sold goods
894: for a given positive value of the parameter~$\beta$:
895: \begin{eqnarray}
896: M_N(\beta)=\frac{1}{\beta}\log\left(\frac{1}{L}
897: \sum\limits_{\omega^N} 2^{\beta E(\omega^N)}\right)=
898: \frac{1}{\beta}\log\left(\frac{1}{L} \sum\limits_{K_1+\dots+K_M\le
899: N} 2^{\beta\sum\limits_{i=1}^n\epsilon_i K_i}\right),
900: \label{average-1a}
901: \end{eqnarray}
902: where $L$ is the number of terms in the sum.
903: The nonlinear averaging is discussed in detail
904: in~\cite{Mas2003}.
905: In the present paper, we show that,
906: under certain natural assumptions,
907: the average can be calculated only in two ways:
908: as the usual linear average
909: and as the nonlinear average of the form (\ref{average-1a}).
910:
911: We divide the goods of all types into groups
912: with different prices.
913: Suppose that the total number of groups of such goods is~$n$,
914: $n\le M$,
915: the $i$th group contains $G_i$ types of goods
916: with the common price $\lambda_i$
917: and $\lambda_1<\lambda_2<\dots <\lambda_n$.
918:
919: In our statement, each $\lambda_i$ is equal to one of~$\epsilon_j$.
920: We denote
921: $\overline{\lambda}=(\lambda_1,\lambda_2, \dots,\lambda_n)$.
922: By definition, we have $\sum\limits_{i=1}^n G_i=M$.
923: Let $K_{i,j}$ be the total number of units of goods of the $j$th
924: type in the $i$th group of goods with the common price~$\lambda_i$.
925: Thus, the structure of sold goods is given
926: by the vector
927: \begin{eqnarray}\label{stru-1}
928: \omega^N=\langle K_{1,1},\dots, K_{1,G_1},\dots, K_{i,1},\dots,
929: K_{i,G_i},\dots, K_{n,1},\dots, K_{n,G_n}\rangle.
930: \end{eqnarray}
931: Let $N_i=\sum\limits_{j=1}^{G_i}K_{i,j}$
932: be the number of all units of goods with the price $\lambda_i$,
933: and let $\nu_i=N_i/N$ be the part of all such goods
934: in their total amount,
935: where $i=1,\dots, n$.
936: The {\it price structure} of sold goods
937: is characterized by the vector
938: $\langle N_1,\dots, N_n \rangle$
939: or, in fractions, by the vector
940: \begin{eqnarray}\label{stru-2}
941: \langle\nu_1,\dots, \nu_n\rangle.
942: \end{eqnarray}
943: Then the total cost of all $N$ units of goods is
944: \begin{eqnarray}\label{income-1}
945: E(\omega^N)= \sum\limits_{i=1}^{n}\lambda_i
946: N_i=N\sum\limits_{i=1}^{n}\lambda_i\nu_i.
947: \end{eqnarray}
948: By $\Xi_N^{N_1,\dots, N_n}$ we denote the set of all versions
949: of the structure of sold goods
950: with a prescribed price structure
951: $\langle N_1,\dots, N_n\rangle$.
952: The number of elements in this set is\footnote{$|A|$ denotes
953: the number of elements in the set~$A$. }
954: \begin{eqnarray}
955: |\Xi_N^{N_1,\dots, N_n}|= \binom{N_1+G_1-1}{G_1-1}
956: \cdot\dots\cdot \binom{N_n+G_n-1}{G_n-1}. \label{xi-1}
957: \end{eqnarray}
958: If the goods are separated in groups,
959: then the sum (\ref{average-1a}) becomes
960: \begin{eqnarray}
961: M_N(\beta,\overline{\lambda})=\frac{1}{\beta}\log\left(\frac{1}{L}
962: \sum\limits_{N_1+\dots+N_n\le N} |\Xi_N^{N_1,\dots, N_n}|
963: 2^{\beta\sum\limits_{i=1}^n\lambda_i N_i}\right),
964: \label{average-1s}
965: \end{eqnarray}
966: where $L$ is the total number of all possible versions
967: of the structure of sold goods,
968: and the second sum in (\ref{average-1a})
969: is taken over all the price structures
970: $\langle N_1,\dots, N_n\rangle$.
971:
972: We assume that the sellers have sold at most $N$
973: units of goods of different types.
974: We consider the problem of predicting
975: the ``most expected'' structure of goods realized
976: in the free market,
977: which is necessary for the average income
978: to be at least~$E_1$.
979: Thus, the following inequality must hold:
980: \begin{eqnarray}\label{income-2}
981: \sum\limits_{i=1}^{n}\lambda_i N_i\ge E_1.
982: \end{eqnarray}
983: In addition,
984: the following natural inequality must hold:
985: \begin{eqnarray}\label{income-2g}
986: \sum\limits_{i=1}^{n}N_i\le N.
987: \end{eqnarray}
988: We introduce the main definitions.
989: The base~$2$ logarithm of the number of possible
990: versions of sale of~$N$ units of goods
991: for which the income is no less than~$E_1$,
992: i.e., (\ref{income-2}) is satisfied,
993: will be called the entropy $H(E_1,\overline{\lambda},N)$.
994:
995: By $\theta(E_1,\overline{\lambda},N)$
996: we denote the solution (with respect to~$\theta$)
997: of the equation
998: \begin{eqnarray}
999: \frac{\partial
1000: M_N(\theta^{-1},\overline{\lambda}))}{\partial\theta}=
1001: H(E_1,\overline{\lambda},N). \label{def-te}
1002: \end{eqnarray}
1003:
1004: The solution of this equation exists and is unique.
1005: This follows from the monotonicity.
1006:
1007: We set
1008: \begin{eqnarray}
1009: \tilde\nu_i(E_1,\overline{\lambda},N))
1010: =\frac{1}{N} \frac{\partial M_N} {\partial\lambda_i}
1011: |_{\theta=\theta(E_1,\lambda)}
1012: \label{nu-i-1}
1013: \end{eqnarray}
1014: for $i=1,\dots, n$.
1015:
1016: The notions introduced above are of general character and can be
1017: used for all possible versions of the goods distribution.
1018: In what follows, we consider a specific case
1019: for which we find the asymptotics of the introduced variables
1020: and prove a certain simplest version of the
1021: ``law of large numbers'' under the following conditions.
1022:
1023: We introduce the notation $\nu_i=N_i/N$ and $p_i(N)=G_i/N$
1024: for $i=1,\dots,n$.
1025: We assume that the variables $p_i(N)$ have the limit $p_i$
1026: as $N\to\infty$, and this limit is greater than zero
1027: for all~$i$. We also assume that $E_1=e_1 N$.
1028: Let $\rho=N/M$ be a constant
1029: (one can also assume that $N/M\to\rho$ as $N\to\infty$).
1030:
1031: In what follows, we also need the restriction
1032: $e_1\le\rho\sum\limits_{i=1}^n\lambda_i p_i$
1033: on the value of the average income per unit sold goods.
1034:
1035: \begin{theorem}%\label{cons-1}
1036: Let $n\ge 3$.
1037: Then, for an arbitrary $\epsilon>0$,
1038: the part of all versions
1039: of the structure of sold goods {\rm(\ref{stru-2})}
1040: for which the average income per unit goods
1041: is no less than $e_1$
1042: \footnote{i.e., under restrictions {\rm(\ref{income-2})}
1043: and (\ref{income-2g}). }
1044: and $|\nu_i-\tilde \nu_i|\ge\epsilon$
1045: at least for a single~$i$, $1\le i\le n$,
1046: does not exceed $2^{-c\epsilon^2N}$,
1047: where $c$ is a constant.
1048: \end{theorem}
1049:
1050: The conditions of the theorem contain
1051: the assumption that the number of nomenclatures of goods
1052: with the same price is sufficiently large.
1053: But one can unite a group of goods with different prices
1054: and, calculating the nonlinear average of these prices,
1055: replace this group of prices in the statement of the theorem
1056: by this nonlinear average price.
1057:
1058: We shall write the corresponding formula.
1059:
1060: We consider a set of prices $\lambda_i$,
1061: where $i=1,\dots,n$,
1062: and a set of numbers $g_i$ equal, for example,
1063: to the number of goods of different types but with the same
1064: price $\lambda_i$.
1065: The numbers $g_i$ are quantities of the order of~$1$.
1066: By~$N_i$ we denote the number of goods purchased at the price
1067: $\lambda_i$.
1068: Taking into account the fact that
1069: $g_i$ different goods can be purchased
1070: at the price $\lambda_i$,
1071: the number of different way for buying $N_i$ goods
1072: at the price $\lambda_i$ is given by the formula
1073: \begin{equation}
1074: \gamma_i(N_i)=\frac{(N_i+g_i-1)!}{N_i!(g_i-1)!}, \label{1ge}
1075: \end{equation}
1076: and the number of different ways for buying
1077: the set of goods
1078: $\{N\}=N_1,\dots,N_n$ is, respectively, equal to
1079: \begin{equation}
1080: \Gamma(\{N\})=\prod_{i=1}^n\gamma_i(N_i)=
1081: \prod_{i=1}^n\frac{(N_i+g_i-1)!}{N_i!(g_i-1)!}. \label{2ge}
1082: \end{equation}
1083:
1084: We assume that the goods are divided in $m\le n$
1085: groups as follows.
1086: Suppose that there are two sequences~$i_\alpha$
1087: and~$j_\alpha$, where $\alpha=1,\dots,m$,
1088: such that
1089: \begin{eqnarray}
1090: &&i_\alpha\le j_\alpha,\qquad
1091: i_{\alpha+1}=j_\alpha+1,\qquad\alpha=1,\dots,m,
1092: \nn\\
1093: &&i_1=1,\qquad j_m=n. \label{3ge}
1094: \end{eqnarray}
1095: In this case,
1096: we say that the goods belong to the group of goods
1097: with number~$\alpha$ if their price $\lambda_i$
1098: satisfies the condition $i_\alpha\le i\le j_\alpha$.
1099: By $\wt{N}_\alpha$ we denote the number of purchased goods
1100: from the group with number~$\alpha$.
1101: This number is given by the formula
1102: \begin{equation}
1103: \wt{N}_\alpha=\sum_{i=i_\alpha}^{j_\alpha}N_i. \label{4ge}
1104: \end{equation}
1105: We introduce $w_\alpha$, which is the nonlinearly averaged price
1106: of goods in group~$\alpha$,
1107: \begin{equation}
1108: w_\alpha=\frac1{\beta\wt{N}_\alpha}\left(
1109: \log\left(\sum_{\{N\}}^\alpha\prod_{i=i_\alpha}^{j_\alpha}\gamma_i(N_i)
1110: 2^{-\beta\lambda_iN_i}\right)
1111: -\log\left(\wt{\gamma}_\alpha(\wt{N}_\alpha)\right)\right),
1112: \label{5ge}
1113: \end{equation}
1114: where $\beta$ is an economic parameter,
1115: for example, the volatility.
1116: We also use the notation
1117: \begin{equation}
1118: \sum_{\{N\}}^\alpha, \label{6ge}
1119: \end{equation}
1120: which means that the summation is performed
1121: over all the sets of nonnegative integers
1122: $N_{i_\alpha},\dots, N_{j_\alpha}$
1123: such that
1124: \begin{equation}
1125: \sum_{i=i_\alpha}^{j_\alpha}N_i=\wt{N}_\alpha.
1126: \label{7ge}
1127: \end{equation}
1128: Moreover,
1129: $\wt{\gamma}_\alpha(\wt{N}_\alpha)$ is given by the formula
1130: \begin{equation}
1131: \wt{\gamma}_\alpha(\wt{N}_\alpha)=\sum_{\{N\}}^\alpha
1132: \prod_{i=i_\alpha}^{j_\alpha}\gamma_i(N_i).
1133: \label{8ge}
1134: \end{equation}
1135: We can show that~(\ref{8ge}) satisfies the formula
1136: \begin{equation}
1137: \wt{\gamma}_\alpha(\wt{N}_\alpha)
1138: =\frac{(\wt{N}_\alpha+\wt{g}_\alpha-1)!}
1139: {\wt{N}_\alpha!(\wt{g}_\alpha-1)!}, \label{9ge}
1140: \end{equation}
1141: where
1142: \begin{equation}
1143: \wt{g}_\alpha=\sum_{i=i_\alpha}^{j_\alpha}g_i. \label{10ge}
1144: \end{equation}
1145:
1146: We note that, in general,
1147: the nonlinearly averaged price of goods in group~$\alpha$
1148: depends on~$\beta$ and~$\wt{N}_\alpha$.
1149:
1150: Now we consider the nonlinear expectation of the income
1151: \begin{equation}
1152: M(\beta,\lambda,N)=-\frac1\beta\log\left(\sum_{\{N\}}'\prod_{i=1}^n\gamma_i(N_i)
1153: 2^{-\beta\lambda_iN_i}\right), \label{11ge}
1154: \end{equation}
1155: where
1156: \begin{equation}
1157: \sum_{\{N\}}' \label{12ge}
1158: \end{equation}
1159: denotes the sum over all sets of nonnegative integers
1160: $\{N\}=N_1,\dots,N_n$
1161: such that
1162: \begin{equation}
1163: \sum_{i=1}^nN_i=N. \label{13ge}
1164: \end{equation}
1165:
1166: \begin{proposition}
1167: The nonlinear expectation~{\rm(\ref{11ge})}
1168: satisfies the relation
1169: \begin{equation}
1170: M(\beta,\lambda,N)=-\frac1\beta\log
1171: \left(\sum_{\{\wt{N}\}}'\prod_{\alpha=1}^m
1172: \wt{\gamma}_\alpha(\wt{N}_\alpha)2^{-\beta
1173: w_\alpha\wt{N}_\alpha}\right), \label{14ge}
1174: \end{equation}
1175: where
1176: \begin{equation}
1177: \sum_{\{\wt{N}\}}' \label{15ge}
1178: \end{equation}
1179: denotes the sum over all sets of nonnegative integers
1180: $\{\wt{N}\}=\wt{N}_1,\dots,\wt{N}_m$
1181: such that
1182: \begin{equation}
1183: \sum_{\alpha=1}^m\wt{N}_\alpha=N. \label{16ge}
1184: \end{equation}
1185: \end{proposition}
1186:
1187: This assertion readily follows from the definitions
1188: of nonlinearly averaged prices~(\ref{9ge})
1189: and nonlinear expectation of income~(\ref{11ge}).
1190:
1191: Now we have
1192: $$
1193: \tilde \nu_\al = \frac{\pa M}{\pa \omega_\la},
1194: $$
1195: and the law of large numbers theorem remains the same,
1196: where $\tilde \nu_\al$ is the number of goods
1197: purchased at the prices $\la_i$
1198: for $i_\al \leq i \leq j_\al$.
1199:
1200:
1201: \begin{thebibliography}{99}
1202:
1203: \bibitem{Kolmog1}
1204: A.~N.~Kolmogorov,
1205: {\it Selected Works in Mathematics and Mechanics},
1206: 1985, 136--137.
1207:
1208: \bibitem{Mas2003}
1209: V.~P.~Maslov,
1210: {\it Axioms of nonlinear averaging in financial mathematics
1211: and the stock price dynamics},
1212: Teor. Veroyatnost. i Primenen.,
1213: 2003, vol.~48, no.~4, 800--810.
1214:
1215: \bibitem{TeorVer2004}
1216: V.~P.~Maslov,
1217: Teor. Veroyatnost. i Primenen.,
1218: 2004, vol.~49, no.~2, 269--296.
1219:
1220: \bibitem{Heding}
1221: J.~Heading,
1222: {\it An Introduction to Phase-Integral Methods} (in English),
1223: Russian transl., Moscow, Mir, 1965.
1224: Afterword by V.~P.~Maslov ``WKB in the multidimensional case."
1225:
1226: \bibitem{TeorVozm65}
1227: V.~P.~Maslov,
1228: {\it Perturbation Theory and Asymptotic Methods},
1229: Moscow, Izd-vo MGU, 1965.
1230:
1231: \bibitem{ArnoldVar}
1232: V.~I.~Arnold, A.~N.~Verchenko, S.~M.~Gusein-Zade,
1233: {\it Singularities of Differentiable Mappings},
1234: I, II, Moscow, Nauka, 1982, 1984.
1235:
1236: \end{thebibliography}
1237: \end{document}
1238: