1: \documentstyle[aps,epsf]{revtex}
2:
3: \begin{document} \renewcommand{\thefootnote}{\fnsymbol{footnote}}
4:
5: \draft \twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname]
6: \title{Model of macroeconomic evolution in stable regionally dependent economic fields}
7:
8: \author{M. Ausloos$^1$, P. Clippe$^2$ and A. Pekalski$^3$ }
9:
10: \address{ $^1$ GRASP and SUPRATECS, B5, Sart Tilman, B-4000 Li\`ege, Belgium \\$^2$ GRASP, B5, Universit\'{e} de Li\`{e}ge, B-4000 Li\`{e}ge, Belgium, \\$^3$ Institute of Theoretical Physics, University of Wroclaw, pl. Maxa Borna 9,
11: PL-50-204 Wroclaw, Poland }
12:
13: \date{\today}
14: \maketitle
15:
16:
17:
18: \begin{abstract}
19:
20: We develop a model for the evolution of economic entities within a geographical
21: type of framework. On a square symmetry lattice made of three (economic)
22: regions, firms, described by a scalar fitness, are allowed to move, adapt,
23: merge or create spin-offs under predetermined rules, in a space and time
24: dependent economic environment. We only consider here one timely variation
25: of the ``external economic field condition''. For the firm fitness evolution
26: we take into account a constraint such that the disappearance of a firm
27: modifies the fitness of nearest neighboring ones, as in Bak-Sneppen population
28: fitness evolution model. The concentration of firms, the averaged fitness, the
29: regional distribution of firms, and fitness for different time moments, the
30: number of collapsed, merged and new firms as a function of time have been
31: recorded and are discussed. Also the asymptotic values of the number of
32: firms present in the three regions together with their average fitness, as
33: well as the number of respective births and collapses in the three regions
34: are examined. It appears that a sort of $critical$ selection pressure exists.
35: A power law dependence, signature of self-critical organization is seen in
36: the birth and collapse asymptotic values for a high selection pressure only.
37: A lack of self-organization is also seen at region borders.
38: \end{abstract}
39:
40: \pacs{PACS numbers: 789.65.Gh, 05.10.Ln, 89.75.-k, 07.05.Tp, 05.65.+b }
41:
42: \footnotetext[1]{electronic address: marcel.ausloos@ulg.ac.be}
43: \footnotetext[2]{electronic address: P.Clippe\@ulg.ac.be}
44: \footnotetext[3]{electronic address: apekal\@ift.uni.wroc.pl}
45:
46:
47: \section{INTRODUCTION}
48:
49: The present contribution extends our earlier paper,\cite{ACP,ACP04}. World
50: economic conditions evolved and are quite varied on different time and space
51: scales. Basic questions are whether the consequences of political conditions
52: have predictable effects, or not and whether annoying situations can be avoided.
53: The questions pertain to macroeconomic themes, not so often touched upon in
54: econophysics (see exceptions in\cite{Richards,Gligor,Aoki2}).
55:
56: The case of the Berlin wall fall followed by Eastern Europe and Central Asia
57: market openings to a so called liberal economy is a intriguing event. From an
58: econophysicist point of view, the event(s) can be considered as an increase in
59: "physical volume, or available space" as well as in a modification of the {\it
60: external fields}\cite{ACP}. In the latter reference we have discussed whether one
61: can describe, within a simple model, the concentration of enterprises and their
62: so called "fitness" as a function of time and space, under varying in time and
63: space economic field conditions. We have observed a non-trivial behavior with
64: cycle features; in our opinion more pronounced that those discussed by Aoki,
65: Kalecki, Freeman and others\cite{Aoki2,AokiShirai,Kalecki,Freeman}. The model
66: however seemed to be too optimistic in not allowing enough bankruptcies,
67: collapses or mere disappearance of firms in the process, in particular with
68: respect to the so called economic selection pressure\cite{ACP,ACP04}. Moreover we
69: neglected the fact that a firm disappearance could locally modify the probability
70: of survival of neighboring ones. Indeed an immense political worry concerns the
71: effect of degradation of an economy through avalanche-like processes.
72:
73: In practice, adaptation to a dynamic environment contains a trade-off: it is
74: more difficult to adapt in the next (time) step as shown e.g. also on financial
75: time series by Yamasaki et al.\cite{Kazuko}. It is also expected that the
76: adaptive behaviors of ''populations'' differ depending whether a Darwinian
77: or Lamarckian scheme is implied. A large discussion exists whether such schemes
78: hold or not, and how in economy, e.g. see
79: \cite{Alchian,Hirshleifer,Boulding,Eldredge3,Kelm,Hogdson,BerghGowdy01,NelsonWinter82}
80: for references and a recent review. Alas, the definition of Lamarckism (or
81: Darwinism) is not totally agreed upon in economy circles.
82:
83: Therefore we present and discuss here below a more elaborate model than in
84: Ausloos {\it et al.}\cite{ACP,ACP04}, i.e. in order to describe changing basic
85: economic conditions and implement more realistic Darwinian-like adaptation rules,
86: as discussed in economy
87: circles\cite{Alchian,Hirshleifer,Boulding,Eldredge3,Kelm,Hogdson,BerghGowdy01,NelsonWinter82}
88: - in space and time. In essence we introduce a Darwin-Bak-Sneppen-like
89: constraint\cite{BakSneppen} in our previous macro-economy\cite{ACP,ACP04}
90: microscopic physics-like model, i.e. the choice of the initial firm or company
91: depends on its relative fitness with respect to the external field as
92: in\cite{ACP04}; the chosen firm is the less adapted one, instead of being chosen
93: at random\cite{ACP}. The fitness of neighboring ones evolve according to some
94: {\it a priori} defined dynamics.
95:
96: The notion of economic external field is also important. Nelson and
97: Winter\cite{NelsonWinter82} write that ''at any time, firms in an industry can be
98: viewed as operating with a set of techniques and decision rules (routines) keyed
99: to conditions external to the firm'' ... ''and to various internal state
100: conditions''. Surely self-organization and external (field) conditions exist at
101: the same time, and are hard to separate\cite{Foster97,Silverberg}. However there
102: are also many different conditions affecting many characteristic variables of a
103: company or industry state. Thus searching for as simple as possible description
104: and consequences of such external though often qualitatively only apparent
105: constraint, to be contrasted to self-organization is of interest. As usual in
106: physics when intending extreme simplification this field should be applied
107: toward or coupled to some variable such that some function be
108: optimized\cite{Radner}, like the free energy in thermodynamics. In macro and
109: micro-economy the field can be understood as of political origin, but also of
110: more physical origin like the weather. In the latter case, see the weather
111: influence on the energy prices\cite{Weron}. As example of the former one can
112: think about deregulation, or wage and benefit taxation influences upon
113: localization and production or market approach by a company.
114:
115:
116:
117:
118: Along this line, the most extreme simplification leads to represent a company by
119: a variable having the same number of characteristic symmetry group elements as
120: the field. Thus a scalar, called the fitness of a company is
121: introduced\cite{ACP,ACP04}. One can consider that the fitness is the price of
122: stocks and the external field can be in some sense the value of the S\&P500 to
123: which the price is compared. A $buy/sell$ strategy will depend on the difference
124: between the two values. A too large difference can suggest bankruptcy and
125: disappearance of a company, or the 'need'' for a modification of the business
126: plan. The notion of fitness of a company is related to such a book value. The
127: fitness might also represent the number found on gross sales or benefit lines....
128: The field notion has to be subsequently adapted.
129:
130:
131: Any such system will be evolutionary if it has a fourth ingredients: beside the
132: set of degrees of freedom, each ranked by a quality fitness criterion, and a
133: mechanism for introducing mutations to interject diversity into the productive
134: process, the system must be governed by a selection process based on the fitness
135: ranking\cite{AokiShirai,Aoki1,JMMALadek} as in biology\cite{APKSV}.
136:
137:
138:
139: Thus we consider a system composed of firms located at the nodes of a locally
140: square symmetry lattice of $L_x, L_y$ size dimensions. As in Ausloos {\it et
141: al.}\cite{ACP} the system is divided into three parts, $k$= I, II and III, by
142: vertical boundaries at $L_{xk}$, i.e. lines along the $y$ axis at
143: $L_{x1},L_{x2}$. There is no periodic boundary conditions. At the beginning there
144: are $N(t=0)$ firms, located only in the first (I) region. One lattice site may be
145: occupied by only one firm. Apart from its location on the lattice a firm $i$ is
146: characterized by its scalar fitness, $f_i$, taken from a uniform distribution,
147: $\in $ (0,1). There is no other factor distinguishing the firms, hence we may
148: consider them to belong to only one group of industrial companies, i.e.
149: representing a particular industrial branch. Initially the fitness values are
150: attributed to firms in a random way.
151:
152: Any firm may change its position\cite{Silverberg} to a vacant nearest
153: neighboring site chosen at random and may either merge with another firm (with a
154: given probability, - according to its business plan) or create a new one. In a
155: region, a firm is under an {\it external field} $F$ $ \in$ (0,1), which
156: characterizes the local economic conditions. The local political system may be
157: more or less demanding,- more or less high taxes, salary rules, worker age (and
158: other) conditions, ...; this is taken into account through the parameter, $sel$,
159: - corresponding to selection in biological systems \cite{APKSV}. In our model it
160: is also a scaling parameter for the field and the fitness, favoring or not the
161: possible disappearance of a company unfit to its environment, the
162: field\cite{Kauffman2000,Auerswald2000,Frenken}.
163:
164: The field $F$ may (and will) differ from region to region and may also change in
165: time, measured in Monte Carlo Steps (MCS). In the following we will only consider
166: one drastic field change and keep stable (or static) all environmental (economic
167: field) conditions for evaluating its effect on the population of entities
168: evolution.
169:
170: As in Ausloos {\it et al.}\cite{ACP} after some time the barrier separating the
171: region I from regions II and III is open and the diffusing firms may enter the
172: latter regions. At that time the external field also changes in the first region
173: and some values are given to the fields in the regions II and III.
174:
175: Before any activity is taken for the firm (moving, merging...) its condition is
176: checked by comparing its fitness with the field $F$. If the two values differ
177: strongly, the firm may collapse and be removed from the system (see the algorithm
178: below).
179:
180: The Bak-Sneppen condition, i.e. a modification of the fitness of the
181: $eliminated$ firm nearest neighbors, is implemented through distributing new
182: (uncorrelated in space and time) random fitness to these (at most 4) neighbors.
183: This corresponds to e.g. simulating the effect a bankruptcy may have on other
184: firms connected to the one which went down. Obviously the fitness neighbors can
185: increase or decrease. Some firms may profit or regret the disappearance of a
186: competitor in the local market.
187:
188:
189: Firms will also be systematically moving on the lattice, looking for partners in
190: view of merging or producing spin-offs. We will take into account the number of
191: contacts in the company network (neighbors on a lattice in our case) for
192: describing the firm evolution. It is empirically known that the replacement of
193: some contacts is needed for
194: innovation\cite{Kauffman2000,Auerswald2000,Frenken,Aoki3}.
195:
196:
197:
198: The (new) firm(s) fitness evolution rule is slightly different from that in
199: Ausloos {\it et al.}\cite{ACP}, but still takes into account the fitness of
200: parents.
201:
202:
203: The following quantities are recorded: concentration of the firms in the three
204: regions and the averaged fitness in the three regions both as a function of time,
205: the spatial distribution of firm concentration and fitness for different time
206: moments, the number of collapsed, merged and new firms as a function of time,
207: finally the asymptotic, i.e. when a stationary state is reached, values of the
208: number of firms present in the three regions together with their average fitness,
209: as well as the number of respective birth and collapses.
210:
211: In section 2, we outline the simulation technique used for implementing the
212: model. We present a few results in Sect. 3, and end with a conclusion in Sect.4.
213:
214:
215:
216:
217: \section{MODEL}
218:
219:
220: The algorithm we implemented is the following one:
221:
222: \begin{enumerate}
223:
224: \item from a number of firms $N(t)$ at a given time $t$ a firm $j$ is randomly
225: picked
226:
227: \item its survival probability is calculated from the condition
228:
229: \begin{equation} p_j = \exp\left(- sel |F - f_j|\right) \end{equation}
230:
231: and compared to a random number $r$ taken from a uniform distribution $\in $
232: (0,1). If $r > p_j$, the firm collapses, i.e. it is removed from the system; the
233: nearest neighbors of the eliminated firm receive new $random$ numbers (taken from
234: a uniform distribution) for their fitness. The algorithm then goes back to 1,
235:
236: \item if $r < p_j$, we try to move the firm to a nearest neighborhood (NN). A
237: random number, $r_1$, is generated from a uniform distribution and if it is
238: smaller than 0.25 then we check whether the ''Northern NN'' is an empty site,
239: and we move the firm there, if possible; if $r_1$ is between 0.25 and 0.50 then
240: we look into the Western NN, etc. If the displacement trial is not successful the
241: algorithm goes back to 1 and search for a new firm;
242:
243: \item next we look for a partner in the nearest neighborhood of the displaced
244: firm new position. If there is a firm in the NN,
245:
246: \item with a probability ($b$ =) 0.01 the neighbor, say $i$, merges with the old
247: firm $j$, which changes its fitness to
248:
249: \begin{equation} f_j = 0.5 (f_i + f_j + (0.5 - r_2) |f_i - f_j|),\end{equation}
250: where $r_2$ is a random number in the range (0,1). The firm $i$ then disappears
251: from the system;
252:
253: \item otherwise with a probability ($1-b$ =) 0.99 the firms $i$ and $j$ produce
254: a new firm $k$, a $spin-off$. The $k$ firm is randomly positioned in the Moore
255: neighborhood of the $j$ firm on an empty site if any exists. The procedure for
256: finding an empty place is similar to the one used when looking for an empty site
257: to move the firm, except that on a square lattice the Moore neighborhood consists
258: of 8 sites - NW, N, NE, W, E, SW, S and SE; whence if the random number $r_3$ is
259: smaller than 0.125 we check the site NW, if it is larger than 0.125 but smaller
260: than 0.250 we check the N site, etc. The new firm receives its fitness depending
261: on that of both parents (which remain in the system) as
262:
263: \begin{equation} f_k = 0.5 (f_i + f_j + (0.5 - r_4) |f_i - f_j|),\end{equation}
264: where $r_3$ and $r_4$ are random numbers in the range (0,1).
265:
266: \item When $N(t)$ firms were picked through (1), one MCS is said completed.
267:
268: \end{enumerate}
269:
270: Typical values taken for the discussed simulations below were : $L_{x1}$ = 50,
271: $L_{x2}$ = 100, $L_y$ = 201, $c(0)$ $\simeq$ 0.8, corresponding to a number of
272: firms $N(0)$ = 8040.
273:
274: The parameters of the model are: $sel$, selection pressure; $t_{change}$, time
275: after which the barrier at $L_{x1}$ is open and the field changed in I; the
276: values of the fields. At the beginning the field in region I has the value $F_I$
277: = 0.5, which is ''optimal''; the values in the regions II and III are irrelevant
278: before the opening of the $L_{x1}$ barrier, since there are no firms there. After
279: the change, at time $t_{change} $ = 100 MCS, the values were arbitrarily taken to
280: be: $F_I$ = 0.3, $F_{II}$ = 0.5, $F_{III}$ = 0.6, which means that the conditions
281: in the region I deteriorated and the best situation is in region II. The
282: conditions in region III are also imposed to be better than those in region I.
283: These values remain unchanged till the end of the simulations.
284:
285: Although the presented curves were obtained each from a single simulation, we
286: have checked that a different initial distribution of the firms (for the same
287: initial concentration, external field and selection) leads to a very similar,
288: even quantitatively, set of situations. We are not interested here in the
289: transient regime before the first 100 MCS. The simulations were carried out till
290: a stationary state was reached, when each investigated quantity mildly oscillated
291: around a rather stable average value. This happened quite soon (several hundreds
292: of MCS) for low selection, but we had to run till 20000 MCS for the highest
293: selection $sel$ = 1.7. For a $sel$ as ''high '' as 1.8, we found that no firm
294: survived after the initial 800 MCS in the whole system.
295:
296:
297:
298:
299: \section{RESULTS and COMMENTS}
300:
301: There are several ways of presenting pertinent results. In the following we
302: stress cases demonstrating the pertinence of the model in view of economic
303: qualitative observations. The scales are chosen in order to allow for better
304: visually comparing cases and emphasizing differences in behavior.
305:
306: Recall that from time $t_{change} $ = 100 MCS, the field values are kept to be
307: $F_I$ = 0.3, $F_{II}$ = 0.5, and $F_{III}$ = 0.6 respectively.
308:
309: \subsection { Time dependent and regional values}In Fig. 1 (a-b) the number of
310: new firms appearing in each region is shown as a function of (MC) time, (a) in
311: the case of a rather low $sel$, i.e. = 0.7 and in the case (b) of a rather high
312: $sel$ = 1.7. In Fig. 2 (a-b) the number of eliminated firms versus time, for the
313: above two values of $sel$ are shown. It is worth pointing here, the difference
314: with respect to the case examined in ref. \cite{ACP}. In the latter paper the
315: long time stability (asymptotic value) was obtained through a sharp increase
316: followed by an exponential decay toward a $zero$ death (and also birth)
317: evolution. In the present case, the number of born or eliminated firms increases
318: smoothly and remains stable at a finite value. Recall that there are 8000 firms
319: (''possible parents'') at $t$=0. Observe the $x$-axis scale indicating the effect
320: of selection pressure on first time invasion of regions II and III.
321:
322:
323: A minimum in the birth and death number is always observed at short time (after
324: 100 MCS) in region I. Notice from Fig. 1(b) and Fig. 2(b) that a strong $sel$
325: might quickly kill any evolution process : zero-birth and zero death. It is also
326: clear from comparing Fig. 1 and Fig. 2 that there is a rapid increase in
327: concentration in the newly opened regions as soon as the barrier at $L_{x1}$ is
328: removed. The death and birth processes are always quantitatively equivalent,
329: indicating that a detailed balance equilibrium is intrinsically met in the
330: model.
331:
332:
333: The number of firms present in the three regions vs. time (in MCS) is shown for
334: several $sel$ values, from 0.4 till 1.7 in Fig.3 (a-d). At very low $sel$ the
335: number of firms rather quickly reaches a rather high saturation value in each
336: region, in fact approximately equal to the initial value in region I. At high
337: $sel$, Fig. 3 (c-d), there is a marked drop in region I, while regions II and
338: III are invaded much later than in Fig.3 (a); the saturation level has a lower
339: value than the initial starting one in region I. The asymptotic number of firm
340: values is seen to decrease with $sel$. The strongly marked oscillations might be
341: considered as a signature of cycles \cite{Aoki2,Freeman,JMMALadek}. At high $sel$
342: values the effects are much enhanced. Also, regions II and III are invaded very
343: slowly and late in time. Yet the fitness optimal value in each region is easily
344: reached and remain stable in time (see Fig.4 below).
345:
346:
347: In Fig. 4(a-d) we show the average fitness $f$ in the three regions vs. MCS time
348: for several values of the selection pressure, ranging from 0.4 to 1.7. Recall
349: that the values of the external field are respectively 0.3, 0.5 and 0.6, and the
350: wall at $L_{x1}$ falls at $t$=100. For a low $sel$ (=0.4) small fluctuations
351: are seen around the average (asymptotic) values which are close to 0.5 in all
352: regions. Since the selection pressure is low, the firms are not strongly forced
353: to being in perfect matching to the field value, whence large deviations from the
354: regional average fitness can be enforced.
355:
356: For higher $sel$, data is quite scattered, and indicate a strongly hard learning
357: (adaptation) process, but the average fitness in region I is near 0.3, and that
358: in region III close to 0.6 as expected. For high $sel$ large fluctuations in the
359: average $f$ value are visible as in Fig. 4(b), at rather short time. At very
360: large $sel$ the beginning of the process is rather noisy, but the average fitness
361: stabilizes at a value expected from the imposed external field. We adopted scales
362: for the figures in order to show qualitatively the decrease in the variance of
363: the fitness distribution with time, for various selection pressures. The
364: adaptation is clearly faster in region I and II, and the variance much reduced
365: as soon as the expected average values are obtained.
366:
367:
368: Considering the physical diffusion process, we recorded the maximum distance
369: reached by the right most firm as a function of time, and present it on a log-log
370: scale for different $sel$ values, from 0.3 till 1.8, in Fig. 5. Recall that due
371: to the lattice topology, critical values are to be found at $x$= 50, 100 and
372: 150. Studying Fig. 5, starting from high $sel$ values, it appears that a plateau
373: occurs (for $sel$ =1.8 and 1.5) near $L_{x2}$=100, also somewhat seen for $sel$ =
374: 1.1. The flow into region II and III seems more continuous at small $sel$
375: values. The small $tails$ sticking upwards (above $x$= 50) for $t> 1000$ witness
376: unsuccessful attempts to colonize regions II and III, suggesting that a
377: $sel$-dependent nucleation-growth process exists in the model. Moreover, for high
378: $sel$ values it appears that the region near the border at $L_{x1}$= 50 is
379: sometimes depleted., indicating a gain that high $sel$ values might destroy the
380: population at all. From Fig. 5 (and others not shown here) we can calculate the
381: time necessary for the first firm to reach the region right boundary, i.e.
382: respectively $L_{x2}$=100 and $L_{x3}$ = 150, as a function of the selection
383: pressure (Fig. 6). The curves look like high order power laws (with exponents
384: equal ca. 4 and 5 respectively), but more complicated functions might be tried.
385:
386: \subsection { Asymptotic values}
387:
388: It is of interest to observe beyond the transient regimes the equilibrium values.
389: In Fig.7 (a) the normalized asymptotic values of the born firms in the three
390: regions are shown as a function of $sel$. The ''normalization'' is made with
391: respect to the total asymptotic number of firms in the region shown in Fig. 7(b).
392: ''Asymptotic'' meaning here that we have averaged the last 500 entries for 10
393: 000 MCS runs. The same type of data is obtained $quantitatively$ for the number
394: of disappearing firms (not shown). A maximum occurs near $sel$ = 0.8 with
395: different analytic behaviors on both sides of this value. The increase goes like
396: a stretched exponential as for diffusion-controlled growing entities\cite{Avrami}
397: at low $sel$ ($<0.8$), i.e. $ sel^{0.45} exp(-0.5/\sqrt(sel))$, but the decay at
398: high $sel$ has been found to be markedly a power law $sel^{-0.40}$, like for
399: critical systems \cite{StanleyRMP}. Recall that each region may contain at most
400: 10 050 entities, a number to be compared to the number of asymptotically existing
401: firms, as displayed on Fig. 7 (b). For completeness, the asymptotic fitness in
402: the three regions is shown vs. $sel$ in Fig. 8. This indicates that only for
403: large $sel$ ($>0.8$) has the asymptotic average fitness almost reached the
404: expected one from the field conditions.
405:
406:
407:
408: \subsection { Regional behaviors}
409:
410: We show the short time dependence ($t$ $\leq $800 MCS) of the (average over a
411: ''vertical column'') firm concentration vs. the column number ''$x$'' for rather
412: strong (1.3) and very strong (1.7) $sel$ values in Fig. 9. A drop in the overall
413: concentration is well seen in region I at short time, but the concentration
414: recovers later on, as seen from previous paragraphs and figures. For $sel$ =1.8
415: the selection pressure is too high to maintain any entity in region I after a few
416: MCS.
417:
418: The dependence of the firm (vertical) concentrations on position ''$x$'' at
419: different (longer) times is shown in Fig. 10 (a-c) for $sel$ values just above
420: the $critical$ one, i.e. $\simeq$ 0.8.
421:
422: Interesting and unexpected features occur. At very low $sel$ (not displayed) a
423: saturation level is reached right from the beginning of the simulation,
424: indicating ''easy adaptation'' everywhere. At high $sel$ the third region is
425: invaded $ca.$ after 5000 MCS, while region II is invaded after 800 MCS. We
426: emphasize the occurrence of large well marked ''oscillations'' in the local
427: concentrations. Dips do occur, see Fig. 10 near the region borders, - likely due
428: to the difficulty for firms to reach the most appropriate fitness together with
429: their neighbors, due to the interregion field gradient.
430:
431:
432: Finally the (vertically averaged) fitness $vs.$ position along the $x$-axis
433: taken at three time steps and for two $sel$ values on both sides of the critical
434: one are shown in Fig. 11. Notice the different time(s) chosen for such snapshots.
435: For medium $sel$ (=0.7), the scatter around the expected theoretical fitness
436: imposed by the external field is not large but the average values nevertheless do
437: not closely correspond to the expected ones from the imposed field, in all three
438: regions, - the more so on the average , and asymptotically as seen above. At high
439: $sel$ (=1.7), the data is more scattered but the expected values are better
440: recognized as fulfilling the external field condition.
441:
442:
443: \section{CONCLUSION}
444:
445: The dynamics of an economic firm population has been considered through a model
446: considering entities characterized by a scalar number, diffusing on a regular
447: lattice, merging, collapsing or creating spin-offs. The economic environment is
448: described by a static regionally dependent field. The adaptation and evolution of
449: the firms to the field condition have been mimicked by a Darwinian-like selection
450: rule but with a fitness evolution equation for the newly appearing firms taking
451: into account that of the parents. The firm fitness distribution is shown to have
452: some best adaptation difficulty, as discussed by Yamasaki et al. \cite{Kazuko}.
453: We have found an unexpected and interesting feature: the local changes of the
454: environment is leading to sharp variations, almost discontinuous ones, in the
455: fitness and concentrations, in particular when the field gradient is strong. A
456: lack of self-organization is thus seen at region borders. This is due to the fact
457: that for an entity attempting an adaptation, the learning process is quite hill
458: climbing as in NK models\cite{Kauffman1998}. Indeed the evolution of an economy,
459: in which the functioning of companies is interdependent and depends on external
460: conditions, through natural selection and somewhat random mutation is similar to
461: bioevolution on NK fitness landscapes as described by
462: Kaufmann\cite{Kauffman1998}. In these evolutionary economics
463: models\cite{Frenken}, economic agents randomly search for new technological
464: design by trial-and-error and run the risk of ending up in sub-optimal solutions
465: due to interdependencies between the elements in a complex system. As argued by
466: Frenken\cite{Frenken} these models of random search are legitimate for reasons of
467: modelling simplicity, but remain limited as these models ignore the fact that
468: agents can apply heuristics.'' We totally agree. Indeed there is no mimicking nor
469: endogeneous search for optimisation as in
470: \cite{Kauffman2000,Auerswald2000,Frenken} in the present model, the algorithm
471: being only geared toward reducing the distance between the company fitness and
472: the external field value.
473:
474:
475: One might take into account in a more proper way the number of business contacts,
476: the evolving firm neighbors : the development of a market or a business plan, or
477: a macroeconomy depends on the available number of contacts (bonds between nodes
478: on a network) whence to the creation and destruction of contacts. A large stock
479: of relational capital (business contacts) usually increases the sold output of a
480: representative firm. Other lattices or network structures should be usefully
481: studied.
482:
483: It would be of interest to consider specific (historical) cases, and connect the
484: MC simulations time scales to real cases. The time scale of changing environments
485: is an available parameter which could indicate how robust the model can be with
486: respect to realistic adaptation and evolution. Some flexibility for defining
487: time scales exist in the (i) $b$ and (ii) $sel$ values: (i) The value of the
488: spin-off creation probability might be too large for previous economies. It has
489: been really shown in the line of the ACP model that for the best fitted company a
490: small $b$ value leads to chaotic behavior while a large $b$ leads to a monopoly
491: like situation\cite{JMMALadek}. (ii) A $sel$ (space and time dependent) value
492: could be found by examining macroeconomy data. A sort of $critical$ selection
493: pressure exists separating different birth-death regimes. This might be used as a
494: time and field scale definition when counting bankruptcies in countries,
495: including per industry type.
496:
497:
498:
499: Contrary to the Bak-Sneppen model \cite{BakSneppen}, the average fitness(es) or
500: concentration(s) do not seem to appear as power laws (like those characterizing
501: avalanches), except in the high selection pressure regime and for long
502: (asymptotic) times. The power law dependence, which is a usual signature of
503: self-critical organization is only seen for a high selection pressure in the
504: asymptotic behavior of the birth-death number. Nevertheless from the set of
505: results, as presented here above, we observe that in such a Darwinian evolving
506: economic world: (i) there are relatively well marked effects due to the
507: "selection pressure"; (ii) temporal field changes can imply a stable density
508: distribution, as in Ausloos {\it et al.}\cite{ACP}; (iii) a diffusion process
509: together with a business plan, and the selection pressure smoothly leads to
510: asymptotic equilibrium states with respect to birth and death processes; (iv) the
511: fitness does not always reach the {\it a priori externally} imposed field value;
512: (v) the role of the gradient (economic) field at borders might indicate either
513: complex oscillations or chaotic behaviors in such regions.
514:
515: We are aware that further improvements are needed. We are drastically
516: caricaturing macro and micro economy field conditions, as well as the description
517: of the "internal" interactions sequence(s). Surely a company, or a set of
518: industries, should not be described by one scalar number $f_i$, but rather a
519: vector (or matrix) model coupled to a (so called external) vector (or matrix)
520: field should be examined. Moreover the birth and death process description
521: through merging and spin off's could also be stochastic or evolutive. The spatial
522: distribution in the distinct regions might be also on interest\cite{Aoki3} as in
523: biology\cite{AMPV}.
524:
525: \vskip 0.5cm {\bf Acknowledgements}
526:
527: \vskip 0.3cm MA thanks Masano Aoki, Mich\`ele Sanglier, Mieko Tanaka-Yamawaki
528: and Kazuko Yamazaki for stimulating discussions, references and comments.
529:
530: \begin{thebibliography}{92}
531:
532:
533: \bibitem{ACP} M. Ausloos, P. Clippe, A. P{\c e}kalski, Simple model for the
534: dynamics of correlations in the evolution of economic entities under varying
535: economic conditions, Physica A 324 (2003) 330.
536:
537: \bibitem{ACP04} M. Ausloos, P. Clippe, A. P{\c e}kalski, Evolution of economic
538: entities under heterogeneous political/environmental conditions within a
539: Bak-Sneppen-like dynamics, Physica A 332 (2004) 394(physics/0309007).
540:
541: \bibitem{Richards} G.R. Richards, Reconciling econophysics with macroeconomic
542: theory, Physica A 282 (2000) 325.
543:
544: \bibitem{Gligor} M. Gligor, M. Ignat, A kinetic approach to some quasi-linear
545: laws of macroeconomics, Eur. Phys. J. B 30 (2002) 125.
546:
547: \bibitem{Aoki2} M. Aoki, A simple model of asymmetrical business cycles:
548: Interactive dynamics of a large number of agents with discrete choices,
549: Macroeconomic Dynamics 2 (1998) 427.
550:
551: \bibitem{AokiShirai} M. Aoki, Y. Shirai, A New Look at the Diamond Search Model:
552: Stochastic Cycles and Equilibrium Selection in Search Equilibrium, Macroeconomic
553: Dynamics 4 (2000) 487.
554:
555: \bibitem{Kalecki} M. Kalecki, A Macroeconomic Theory of the Business Cycle,
556: Econometrica 3 (1935) 327; ibid., A Theory of the Business Cycle Review of
557: Economic Studies 4 (1937) 77; ibid., Theory of Economic Dynamics: An essay on
558: cyclical and long-run changes in capitalist economy, Monthly Review Press, New
559: York, 1965.
560:
561: \bibitem{Freeman} C. Freeman, Schumpeter's Business Cycles Revisited, in A.
562: Heertje and M. Perlman (eds) {\it Evolving Technology and Market Structures} pp.
563: 17-38 (1990)
564:
565: \bibitem{Kazuko} K. Yamasaki, K. Kitakaze, M. Sekiguchi, Adaptation under the
566: Dynamic Environment and Application to Financial Time Series, IPSJ SIGNotes
567: Intelligence and Complex Systems Abstract No.127 - 014
568:
569: \bibitem{Alchian} A. Alchian, Uncertainty, Evolution, and Economic Theory, J.
570: Pol. Econ. 58 (1950) 211.
571:
572: \bibitem{Hirshleifer} J. Hirshleifer, Economics from a Biological Viewpoint,
573: Journal of Law and Economics 20 (1977) 1.
574:
575: \bibitem{Boulding} K. Boulding, What is Evolutionary Economics?, J. Evol. Econ.
576: 1 (1991) 9.
577:
578: \bibitem{Eldredge3} N. Eldredge, Evolution in the Market Place, Structural
579: Change and Economic Dynamics 8 (1997) 385.
580:
581: \bibitem{Kelm} M. Kelm, Schumpeter's Theory of Economic Evolution: A Darwinian
582: Interpretation, Evolutionary Economics 7 (1997) 97.
583:
584: \bibitem{Hogdson} G.M. Hodgson, ''Darwinism in economics: from analogy to
585: ontology, J. Evol. Econ. 12 (2002) 259.
586:
587: \bibitem{BerghGowdy01} J. van den Bergh, J. Gowdy, The Microfoundations of
588: Macroeconomics: An Evolutionary Perspective, Cambridge Journal of Economics 27
589: (2003) 65.
590:
591: \bibitem{NelsonWinter82} R.R. Nelson, S. G. Winter, An Evolutionary Theory Of
592: Economic Change, Cambridge, Belknap Press, 1982.
593:
594: \bibitem{BakSneppen} P. Bak, K. Sneppen, Punctuated equilibrium and criticality
595: in a simple model of evolution, Phys. Rev. Lett. 71 (1993) 4083.
596:
597: \bibitem{Foster97} J. Foster, The Self-Organization Approach in Economics,
598: Structural Change and Economic Dynamics 8 (1997) 427.
599:
600: \bibitem{Silverberg} G. Silverberg, G. Dosi, L. Orsenigo, Innovation, Diversity
601: and Diffusion: A Self-Organization Model, The Economic Journal 98 (1998) 1032.
602:
603: \bibitem{Radner} P. Dutta, R. Radner, Profit Maximization and the Market
604: Selection Hypothesis, Review of Economic Studies 66 (1999) 769.
605:
606:
607: \bibitem{Weron} R. Weron, B. Przyby{\l}owicz, Hurst analysis of electricity
608: price dynamics, Physica A 283 (2000) 462.
609:
610:
611: \bibitem{Aoki1} M. Aoki, Assymetrical Cycles and Equilibrium Selection in
612: Finitary Evolutionary Economic Models, in {\it Cycles, Growth, and Structural
613: Changes}, L. Punzo Ed., (Routledge, London, 2001) Ch. 8
614:
615: \bibitem{JMMALadek} J. Mi\'skiewicz, M. Ausloos, Logistic map approach to
616: economic cycles, Physica A, submitted.
617:
618: \bibitem{APKSV} A. P{\c e}kalski, K. Sznajd-Weron, Population dynamics with and
619: without selection, Phys. Rev. E 63 (2001) 31903.
620:
621: \bibitem{Kauffman2000} S.A. Kauffman, J. Lobo, W.G. Macready, Optimal search on
622: a technology landscape, J. Econom. Behav. Organiz. 43 (2000) 141.
623:
624: \bibitem{Auerswald2000} P. Auerswald, S. Kauffman, J. Lobo, K. Shell, "The
625: Production Recipes Approach to Modeling Technological Innovation: An Application
626: to Learning By Doing," Cornell University, CAE Working Paper 98-10, 1998.
627:
628: \bibitem{Frenken} K. Frenken, Innovation rules: the NK-model as a framework for
629: memetics, unpublished (2001); K. Frenken, L. Marengo, M. Valente,
630: Interdependencies, near-decomposability and adaptation, pp. 145-165 in: Brenner,
631: T. (ed.), Computational Techniques for Modelling Learning in Economics ( Kluwer,
632: Boston, 1999); $http://www-ceel.gelso.unitn.it/papers/papero99_{-}03.pdf$
633:
634: \bibitem{Aoki3} M. Aoki, Cluster Size Distribution of Economics Agents of Many
635: Types in a Market, J. Math. Anal. Appl. 249 (2000) 32.
636:
637: \bibitem{Avrami} D. Crespo, T. Pradell, M. T. Clavaguera-Mora, N. Clavaguera,
638: Microstructural evaluation of primary crystallization with diffusion-controlled
639: grain growth, Phys. Rev. B 55 (1997) 3435.
640:
641: \bibitem{StanleyRMP} H. E. Stanley, Scaling, universality, and renormalization:
642: Three pillars of modern critical phenomena, Rev. Mod. Phys. 71 (1999) S358.
643:
644: \bibitem{Kauffman1998} S.A. Kauffman, S. Levin, Towards a general theory of
645: adaptive walks on rugged landscapes, J. Theor. Biol. 128 (1987) 11.
646:
647: \bibitem{AMPV} M. Ausloos, I. Mroz, A. P{\c e}kalski, N. Vandewalle, Lattice
648: gas model of gradual evolution, Physica A 248 (1998) 155.
649:
650:
651:
652: \end{thebibliography}
653: \vskip 0.3cm
654:
655:
656: \section*{Figure captions} \vspace*{0.2cm} Fig. 1 -- Number of new firms born in region I, II, III ( +, x, dotted square,
657: respectively) as a function of time, (a) $sel$ = 0.7 ; (b) $sel$ = 1.7
658:
659: \vskip 0.6cm
660:
661: Fig. 2 -- Number of eliminated firms in the three regions versus time, for (a)
662: $sel$ = 0.7 ; (b) $sel$ = 1.7
663:
664: \vskip 0.6cm
665:
666: Fig. 3 -- Number of firms in the three regions $vs.$ time (in MCS) for (a)
667: $sel$ = 0.4; (b) $sel$ = 0.9; (c) $sel$ = 1.3; (d) $sel$ = 1.7
668:
669: \vskip 0.6cm
670:
671: Fig. 4 -- Average fitness in the three regions I, II, III ( +, x, dotted square,
672: respectively) vs. MCS time for several values of the selection pressure (a) =
673: 0.4; (b) = 0.9;(c) = 1.3; (d) = 1.7
674:
675: \vskip 0.6cm
676:
677: Fig. 5 -- Maximum distance reached by the right most firm as a function of time,
678: on a log-log scale for different $sel$ values
679:
680: \vskip 0.6cm
681:
682: Fig. 6 -- Shortest time necessary for a firm to reach the region II (100) or III
683: (150) right boundary, i.e. $L_{x1}$=100, $L_{x2}$=150, as a function of the
684: selection pressure ; curves indicate the best power law fit, i.e. $ca.$ 4 (top)
685: and 5 (bottom) respectively
686:
687: \vskip 0.6cm
688:
689: Fig. 7 -- (a) Normalized asymptotic values values of the born firms in the three
690: regions as a function of various $sel$; ''Normalization'' is with respect to
691: (b) total asymptotic number of firms in the respective region
692:
693: \vskip 0.6cm
694:
695: Fig. 8. -- Asymptotic fitness in regions I, II and III $vs.$ $sel$ ; recall
696: that the field values are imposed to be 0.3, 05. and 0.6 respectively
697:
698: \vskip 0.6cm
699:
700: Fig. 9 -- Short time dependence of the firm concentration in regions I, II, III
701: vs. spatial coordinate ''$x$'' for $sel$ values (a) = 1.3, (b) = 1.7 and at
702: different time moments - 10,20,50 and 800 MCS.
703:
704: \vskip 0.6cm
705:
706: Fig. 10 -- Dependence of the firm concentrations on position ''$x$'' at different
707: times for $sel$ = (d) 1.1; (e) 1.5; (f) 1.7
708:
709: \vskip 0.6cm
710:
711: Fig. 11 -- Vertically averaged fitness $vs.$ position along the $x$-axis at three
712: time steps and for two $sel$ values, i.e. (a) 0.7 and (b) 1.7
713:
714:
715: \end{document}
716: