Simple regs rdc ineq

arXiv:1007.0461v2[physics.soc-ph]13Jul2010
How simple regulations can greatly reduce inequality
J. R. Iglesias1
1
Programa de Pós-Gradua¸cão em Economia and Instituto de F´ısica,
Universidade Federal do Rio Grande do Sul, 91501-970 Porto Alegre, Brazil
(Dated: July 14, 2010)
Many models of market dynamics make use of the idea of wealth exchanges among economic
agents. A simple analogy compares the wealth in a society with the energy in a physical system,
and the trade between agents to the energy exchange between molecules during collisions. However,
while in physical systems the equipartition of energy is valid, in most exchange models for economic
markets the system converges to a very unequal “condensed” state, where one or a few agents
concentrate all the wealth of the society and the wide majority of agents shares zero or a very tiny
fraction of the wealth. Here we present an exchange model where the goal is not only to avoid
condensation but also to reduce the inequality; to carry out this objective the choice of interacting
agents is not at random, but follows an extremal dynamics regulated by the wealth of the agent.
The wealth of the agent with the minimum capital is changed at random and the difference between
the ancient and the new wealth of this poorest agent is taken from other agents, so establishing a
regulatory tool for wealth redistribution. We compare different redistribution processes and conclude
that a drastic reduction of the inequality can be obtained with very simple regulations.
PACS numbers: 89.65.s, 89.65.Gh, 05.20.y, 05.65.+b
I. INTRODUCTION
Empirical studies focusing the income distribution of workers, companies and countries were first presented more
than a century ago by Italian economist Vilfredo Pareto. He observed, in his book “Cours d’Economie Politique” [1],
that the distribution of income does not follow a Gaussian distribution but a power law. That means that the
asymptotic behavior of the distribution function is not exponential (as it should be for a Gaussian distribution) but
follows a power function that decreases, for big values of the wealth, as w−α
, being α > 1 the exponent of the
power law. Non-gaussian distributions are denominated Levy distributions [2], thus this power law distribution is
nowadays known as Pareto-Levy Distribution. Then, Pareto asserted that in different countries and times the income
distribution follows a power law behavior, i.e. the cumulative probability P(w) that an agents have an income which
is at higher or equal to w is given by P(w) ∝ w−α
[1]. The exponent α is named Pareto index. The value of this
exponent changes with geography and time, but a typical values are close to 3/2. The bigger the value of the Pareto
exponent the higher the inequality in a society.
However, recent data indicate that, even though Pareto distribution provides a good fit in the the high income
range, it does not agree with the observed data over the middle and low income ranges. For instance, data from
Japan [3, 4], Italy [5], India [6], the United States of America and the United Kingdom [7–9] are fitted by a log-normal
or Gaussian distribution with the maximum located at the middle income region plus a power law for the high income
strata. This kind of behavior is also observed in the wage distribution in Brazil. We have represented on Fig. 1 the
number of families as a function of the annual income and it is clear that for low and middle salaries the distribution
follows an almost Gaussian law, while for high salaries the curve clearly departs from the exponential and may be
described, in spite of the high noise, as a power law distribution. The Pareto exponent for this distribution is 2.7.
If we consider that the results in Fig. 1 are not cumulative, The integration of the curve will deliver the cumulative
distribution and the high income region will show a exponent 1.7 very near Pareto’s value of 3/2.
One way of justifying in a qualitative way the two regimes: Gaussian + power law, is by considering that in the low
and middle income classes the process of wealth accumulation is additive, causing a Gaussian-like distribution, while
in the high income range, wealth grows in a multiplicative way (interests or other mechanism of multiplying assets),
generating the observed power law tail [4].
Another verification of Pareto-Levy’s law is shown on Fig. 2, where the accumulated Gross Domestic Product of
the 100 richest cities in Brazil [11] is plotted as a function of the ranking. The county of São Paulo concentrates the
biggest fraction of the total Brazilian domestic product, around 12%, followed by Rio de Janeiro, Bras´ılia and Belo
Horizonte. All in all the first 100 counties in the ranking concentrate almost 60% of the Brazilian GDP. The log-log
plot used on Fig. 2 makes the power law explicit , as a power function becomes a straight line when plotted in double
logarithmic scale (while exponentials, or Gaussians, look as parabolas, for example the low – middle income region
in Fig. 1). This power law is valid for the richest counties. However, as there are more than 5500 in Brazil, the vast
majority (not included in the graphic) shares the other 40% of the GDP, so, it is certain that the distribution deviates
form the power low and becomes a Gaussian distribution for the big majority of cities and towns.

2
103
104
105
102
103
104
105
106
Numberoffamilies
Income
Brazil, income data 2004
Power law, exponent -2.7
Gaussian
FIG. 1: Number of families per salary class (in Brazilian Real, annual value). Data collected from IBGE, PNAD 2004 [10]
1 10 100
10
20
30
40
50
60
ParticipationintheGDP(%)
Ranking
SP
RJ
BSB BH
FIG. 2: Cumulative participation in the GDP of the 100 richest counties in Brazil. The red line is an interpolation with a linear
law, indicating, in the double logarithmic scale, that the distribution follows a power law. Data extracted from IBGE [11]

3
Power laws are not rare in nature, so it is not surprising that the wealth distribution follows a power law. The energy
liberated in earthquakes, the size of avalanches, the size of cities, the frequency of all these phenomena are described
by power laws [12]. The quiz with the income and wealth distribution is not the power law, but how this distribution
is generated through the dynamics of the agents interacting. On the other hand, a Pareto-Levy distribution is more
unequal than a Gaussian: when the distribution follows a power law there are more affluent agents than in the case of
a Gaussian distribution, but also more poor agents. And when the Pareto exponent increases the middle class tends
to disappear.
In order to try to describe the processes that generate a given profile for the wealth distribution, in recent years
diverse exchange models have been widely applied to describe wealth and/or income distributions in social systems.
Different mathematical models of capital exchange among economic agents have been proposed trying to explain these
empirical data (For a review see ref. [13]). Most of these models consider an ensemble of interacting economic agents
that exchange a fixed or random amount of a quantity called “wealth”. This wealth represents the agents welfare.
The exact choice of this quantity is not straightforward, but one can think that it stands for the exchange of a given
amount of money against some service or commodity. Within these models the amount of exchanged wealth when
two agents interact corresponds to some economic “energy” that may be randomly exchanged. If this exchanged
amount corresponds to a random fraction of one of the interacting agents wealth, the resulting wealth distribution is
– unsurprisingly – a Gibbs exponential distribution [7].
Aiming at obtaining distributions with power law tails, several methods have been proposed. Numerical
procedures[13, 16–19, 21], as well as some analytical calculations [22, 23], indicate that one frequent result of that
kind of models is condensation, i.e. concentration of all available wealth in just one or a few agents. This result
corresponds to a kind of equipartition of poverty: all agents (except for a set of zero measure) possess zero wealth
while few ones concentrate all the resources. In any case, an almost ordered state is obtained, and this is a state of
equilibrium, since agents with zero wealth cannot participate in further exchanges. The Gini coefficient [29] of this
state is equal to 1, indicating perfect inequality [13]. Several methods have been proposed to avoid this situation, for
instance, exchange rules where the poorer are favored [6, 13, 21, 23, 24] but in all circumstances the final state is one
with high inequality, i.e. very near condensation.
The exchange rules that produce condensation consider that, when two agents interact, the exchanged amount
∆w is proportional to the wealth of one of the participants or to both [22]. One particular example is ∆w =
min{(1 − β)w′
; (1 − β)w′′
}, where w′
and w′′
are the wealth of the two interacting agents, and β is the capital fraction
that the agents risk during an exchange [13, 21, 23]. It is worth noting that even approaching a condensed state,
in the intermediate stages the wealth distribution goes through a series of power law distributions where the Pareto
exponent increases as a function of time [23]. As real societies are not in a condensed state (since such a state would
represent the “thermal death” of the economy), some kind of regulation must be present to guarantee that resources
in the power law tail are re-injected back in the region of Gaussian distribution.
A few years ago we presented an alternative model for wealth distribution, the Conservative Exchange Market
Model (CEMM), inspired by the ideas of John Rawls [26] and Amartya Sen [27] and also on the Bak-Sneppen model
for extinction of species [28]. The main point of the model is that some kind of action should be taken to change the
state of the poorest agent in the society. The idea of a society that take measures in order to improve the situation
of the most impoverished is compatible with the propositions of John Rawls, in his book “A Theory of Justice” [26],
directed towards an inventive way of securing equality of opportunity as one of the basic principles of justice. He
asserts that no redistribution of resources within a state can occur unless it benefits the least well-off: and this should
be the only way to prevent the stronger (or richer) from overpowering the weaker (or poorer). The practical way to
carry out this proposition in a simulation was adapted from the Bak-Sneppen model of Self-Organized Criticality
applied to the extinction of biological species [28]. In this model the less fitted species disappears and is replaced by
new one with different fitness, and the appearance of this new species affects the environment changing the fitness
of the neighboring species. In 2003 we developed a similar model where the fitness is substituted by the assets of
a particular agent, and the model is now conservative, the difference between the new and the old wealth is taken
from (if positive) or given to (if negative) the assets of the neighbors of the poorest agent [14, 15]. The distribution
obtained follows an exponential law as a function of the square of wealth and a poverty line with finite wealth is also
obtained, i.e. the poorer agents do not have zero wealth (as it happens in the most exchange models). Also, the
Gini coefficient obtained is relatively low and compares well with the values of the Gini coefficient of some European
countries as Denmark or Sweden [15]. This suggest a path to decrease inequality in real societies [15].
Here we revisit this model including regulatory mechanisms in order to further diminish inequality. We will verify
that proportional taxes have a important effect on redistribution, while “uniform” taxes, like consumption taxes,
exhibit a lesser effect. But in both situations the impact on the poverty line and the Gini coefficient is really
impressive.
In the next section we present a very short review of the original model: the Conservative Exchange Market Model
(CEMM) and its main conclusions. Then, we introduce regulatory tools in the model in order to compare with the

4
previous results: Two kind of regulatory mechanisms will be discussed and the obtained results will be compared.
Finally, in the last section we present our conclusions.
II. THE CONSERVATIVE EXCHANGE MARKET MODEL - CEMM
The Conservative Exchange Market Model (CEMM) [14] is a simple macroeconomic model that consists of a one-
dimensional lattice with N sites and periodic boundary conditions. That means that each site represents an economic
agent (individuals, industries or countries) linked to two neighbors. Periodic boundary conditions denote that the
lattice closes on itself like a ring, being the last site in the chain neighbor of the first one. To each agent it is assigned
some wealth-parameter that represents its welfare, like the GDP for countries or accumulated wealth for individuals.
One chooses an arbitrary initial configuration where the wealth is a number between 0 and 100 distributed randomly
and uniformly among agents. The dynamics of the system is supported on the idea that some measure should be taken
to modify the situation of the poorest agent. In this context, this process is simulated by an extreme dynamics [28]:
at each time step, the poorest agent, i.e., the one with the minimum wealth, will perform (or be the subject of) some
action trying to improve its economic state. Since the outcome of any such measure is uncertain, the minimum suffers
a random change in its wealth, ∆w [14, 15]. In the first version of the model it is assume that whatever wealth is
gained (or lost) by the poorest agent it will be at the expenses of its neighbors and that ∆w will be equally deducted
from (or credited to) its two nearest neighbors on the lattice, making the total wealth constant. Numerical simulations
on this model showed that, after a relatively long transient, the system arrives at a self-organized critical state with a
stationary wealth distribution (Fig. 1 of Ref. 14) in which almost all agents are above a certain threshold or poverty
line.
Another possibility is to subtract ∆w from two agents picked at random. This situation has also been considered [15,
20], it is the annealed or mean field version of the model, and corresponds to a situation in which the agents with
which the exchange takes place are chosen at random and not based on geographical proximity.
In the nearest-neighbor version of the model, one founds a minimum wealth or poverty line that is ηT ≈ 40% of
the maximum initial wealth and above this threshold the distribution of the wealth of the agents is an exponential
P(w) ≈ exp(−w2
/2σ2
), with σ = 22.8 [20]. The obtained Gini coefficient, G [29], is very low, of the order of G = 0.1.
In the mean-field case the model exhibits a lower threshold, ηT ≈ 20%, and, beyond it, also an exponential distribution
with a higher value of σ ≈ 56.7 [20] and of the Gini coefficient G ≈ 0.25.
In the present article we would like to discuss a similar model still using extremal dynamics, but making some
changes in the way the wealth is redistributed. In the original CEMM model [14, 15] the quantity of money of the
agent with the minimum wealth is changed to an arbitrary value between the limits {0, 100} and the difference between
the new and the old wealth, ∆w, is taken from the neighbors in the local version of the model or from agents picked
at random in the mean-field version.
Now, we will include regulations on the form of taxes, and we will consider two scenarios: 1) Uniform taxes and 2)
Proportional taxes.
III. CEMM WITH UNIFORM TAXES
Let us first consider the situation where the amount attributed to the poorest agent is equally collected from the
full population. In this way each of the agents (including the agent of minimum wealth) will contribute to improve the
situation of the less favored ones. This kind of redistribution simulates a tax that is the same for every agent. Taxes
on consumption, like TVA in Europe, ICM in Brazil, IVA in Argentine, etc., are of this type, every agent contributes
the same amount independent of his available resources. The obtained wealth distribution in this case of uniform
and global taxes, is of the same type as in the original model [14] but now the poverty line is lower than in the local
case (≈ 25% of the maximum wealth) and higher that in the mean-field case. Also, the Gini coefficient is in between
G ≈ 0.2. Finally the wealth distribution is almost linear and it is represented in Fig. 3, where we show the results
for a number of agents N = 104
and 107
time-steps (we have verified that in this limit the system has attained the
state of self-organize criticality and there are no more changes neither in the poverty line nor in the Gini coefficient).
So, a system with uniform taxes exhibit less inequality than the previous mean-field model, but higher than the local
model.

5
20 40 60 80 100
0
200
400
600
Agents
Wealth
Gini=0.2
FIG. 3: Plot of the wealth distribution for N = 10000 agents considering that taxes are uniformly distributed among all agents.
The poverty line is ηT = 25
0 2000 4000 6000 8000 10000
20
40
60
80
100
Wealth
Agents
Poverty Line
FIG. 4: Plot of the wealth of each agent in a system with 10000 agents with local proportional taxes. “Islands” of poverty and
aﬄuence are observed, but the poverty line is high, of the order of 40
IV. CEMM WITH PROPORTIONAL TAXES
Let us now assume a more equitable way to apply taxes: proportional taxes, equivalent to the taxes on income
or on fortune. Here we will assume a linear proportionality, but it is very easy to modify the model to consider
progressive or regressive taxation. As this case has not been previously studied we will consider both the local and
global scenarios.
A. Local taxes
The dynamics here is very similar to the original CEMM. In the original model if the agent with minimum wealth
changes his fortune by a quantity ∆w, this capital (positive or negative) is equally subtracted from his neighbors.
Each one of them have a deduction ∆w/p where p is the number of neighbors considered[21]. We will consider now
that the deduction is not equally performed but proportional to the wealth of the neighbor. We deﬁne a factor

6
30 40 50 60 70 80 90 100
0
200
400
600
800
1000
1200
1400
1600
Agents
Wealth
Gini=0.1
FIG. 5: Histogram of the wealth distribution with local proportional taxes. There is a strong middle class and the distribution
is exponential, as if refs. 14, 21
λ = ∆w/Knn where Knn is the total capital of the p neighbors. Each neighbor will suffer a subtraction (if ∆w > 0)
or addition (if ∆w < 0) on his wealth equal to λwi. That means that if the i-agent is one of the neighbors his wealth
at time (t + 1) will be
wi(t + 1) = (1 − λ)wi(t), (1)
i.e., his wealth will be reduced if ∆w > 0 or will increase if ∆w < 0. It is straightforward to verify that with this
condition the total wealth is conserved.
In Fig. 4 we have represented the wealth of the agents in a scale from 0 to 100. We have again considered 104
agents and p = 4 neighbors. Because the transfer of wealth is local, some “geographical” segregation appears in the
wealth distribution. This is clear in Fig. 4 where it is possible to see that the poverty line is relatively higher than
in the previous situation, of the order of 40, but also that there are geographical differences: on the left one observes
a “burst” of poor agents below the poverty line, but it is also possible to see that there are wealthy and middle
class regions in the plot, indicating that the dynamics induces regional differences. Certainly this kind of behavior
should be visible in a more emphatic way if one consider a complex lattice instead of the very simple one-dimensional
model that we describe here. In any case, the Gini coefficient is very low, of the order of G = 0.1, as for the local
model previously studied [20]. Therefore, even if the results are encouraging they are not very different from the
original CEMM where the difference received by the poorest agent were equally shared (and not proportional) by the
neighbors. Also, the wealth distribution is exponential, as in ref. [21] and it is represent5ed in Fig. 5: there is a peak
in the number of agents just above the poverty line, and then the number of agents decreases exponentially as the
wealth increases.
It is when one examines the application of global taxes that differences appear, as well in the poverty line and the
Gini coefficient than in the type of distribution.
B. Global taxes
Let us now consider that the amount ∆w received by (or deduced from) the poorest agent is subtracted from (or
added to) all the agents (including the recipient) but in proportion to their wealth. Equation (1) is still valid but now
λ = ∆w/Ktot, being Ktot the wealth of the full society.
Globalization has its shortcomings when compared to local models but still the results are impressive. The poverty
line is a little lower than in the local case, 32 compared to 40 and the Gini coefficient higher, of the order of G = 0.16.
We have represented on Fig. 6 the wealth of each agent and we can verify that the distribution is more homogeneous

7
0 2000 4000 6000 8000 10000
20
30
40
50
60
70
80
90
100
Poverty Line
Wealth
Agents
FIG. 6: Plot of the wealth of each agent in a system with 10000 agents, for the case of global proportional taxes. The distribution
is uniform, without islands, and the poverty line is of the order of 32
from a geographical point of view. However, if one examines the wealth distribution, plotted on Fig. 7, the figure
suggest a power law distribution, even if we do not have enough statistics, particularly for the tail of affluent agents,
to determine the value of the exponent. In any case the distribution is not exponential and this is an expected when a
multiplicative redistribution is considered. Finally if we compare the distributions for local taxes (Fig. 5) and global
taxes (Fig. 7) it evident that the number of poor agents is lower and the number of middle class agents is higher in
the situation with global taxes, while the poverty line is higher in the case of local taxes.
Then when proportional taxes are applied to the full population, each agent contributes with a very small amount,
but we obtain a society where there are no zero-wealth agents, neither geographical disparities, as it can be observed
in Fig. 6. Also, the wealth distribution is a power law, then the number of affluent agents is bigger than in the case
of an exponential distribution, nevertheless the Gini coefficient is still very low, as it is shown on Fig. 7.
V. DISCUSSION AND CONCLUSION
Inequalities are related to social and economic phenomena and its origin is a subject of much debate. The fraction of
rich and poor people in a country depends on the inequalities of the income distribution but also on the remuneration
of labor, profits, taxes, etc. The model here presented is certainly a very simplified one; for example, taxes do
not play just a redistributive objective, but also are used by governments to assure the infrastructure needed to
economic development. Nonetheless, in spite on its simplicity the present model describes a very strong redistributive
mechanism, that coincides with some public policies, like the “bolsa-familia” (family fellowship, allocate to very poor
family groups) in Brazil, or the small loans to jobless people in order to develop their own business. Even in the global
case, and with taxes proportional to the possessions of each agent, the contribution of each one of them is very small
but results in a low Gini coefficient, even lower than the one observed in the most egalitarian countries, like Denmark
or Japan (where the Gini coefficient is of the order of G = 0.25). Finally, the poverty line is much higher that the one
obtained with purely exchange models.
We remark also that the local version of the model generates a wealth distribution with a very low Gini index, so
very close to an full equitable society, but with geographical differences. In the global case, the Gini coefficient is a
little higher , but from a geographical point of view the distribution is more uniform, so maybe this is a point in favor
of globalization. And coming back to the inspiration of this work, it is interesting that the minimum dynamics favors
wealth redistribution when acting on the poorest agents, in the same sense defined by Rawls [26]. It seems that this
kind of dynamics, when applied to markets, is a road to ensure to the poorest agent a chance to improve his situation.
We conclude that the model, in spite of its simplicity, is able to reproduce some properties of modern economies,
particularly the wealth distribution of welfare societies, and it indicates a way to improve the situation of extreme
inequality in some countries with very high Gini indexes.

8
40 60 80 100
0
200
400
600
800
1000
Agents
Wealth
Gini= 0.16
FIG. 7: Histogram of the wealth distribution in the case of global proportional taxes. There is still a strong middle class and
the distribution looks like a power law
Acknowledgements
We acknowledge fruitful discussions with G. Abramson, S. Gon¸calves, F. Laguna, S. Souza, R. Donángelo and J.L.
Vega. We acknowledge financial support from Brazilian agencies CNPq and CAPES.
[1] Pareto V (1897), Cours d’Economie Politique, Vol. 2, F. Pichou, Lausanne, Switzerland
[2] Mantegna R M and Stanley H. E. (2000), An Introduction to Econophysics, correlations and complexity in Finnaces,
Cambridge University Press, Cambridge, UK
[3] Aoyama H Souma W and Fujiwara Y (2003) Physica A 324:352
[4] Nirei M and Souma W (2004) Two factor model of income distribution dynamics, Santa Fe Institute (USA) report sfi0410029
[5] Clementi F and Gallegati M (2005) Physica A 350:427
[6] Sinha S (2006) Physica A 359:555
[7] Dragulescu A and Yakovenko V M (2000) The European J. of Physics B 17:723–729
[8] Dragulescu A and Yakovenko V M (2001) The European J. of Physics B 20:585
[9] Dragulescu A and Yakovenko V M (2001) Physica A 299:213
[10] http://www.ibge.gov.br/home/estatistica/populacao/trabalhoerendimento/pnad2004/default.shtm
[11] http://www.ibge.gov.br/home/estatistica/economia/pibmunicipios/2006/default.shtm
[12] Bak P (1999), How Nature Works, Springer-Verlag, New York
[13] Caon G M , Gon¸calves S and Iglesias J R, The European Physical Journal - Special Topics, 143:69 (2007)
[14] Pianegonda S, Iglesias J R, Abramson G and Vega J L (2003) Physica A 322:667
[15] Pianegonda S and Iglesias J R (2004) Physica A 342:193
[16] Sinha S (2003) Physica Scripta T106:59
[17] Chatterjee A, Chakrabarti B K and Manna S S (2004) Physica A 335:155
[18] Chakrabarti B K and Chatterjee A (2004), Ideal Gas-Like Distributions in Economics: Effects of Saving Propensity, in
“Applications of Econophysics”, Ed. H. Takayasu, Conference proceedings of Second Nikkei Symposium on Econophysics,
Tokyo, Japan, 2002, by Springer-Verlag, Tokyo. Pages 280–285. Preprint available: cond-mat/0302147.
[19] Chakraborti A and Charkrabarti B K (2000) The European J. of Physics B 17:167
[20] Iglesias J R, Gon¸calves S, Pianegonda S, Vega JL and Abramson G (2003) Physica A 327:12
[21] Iglesias J R, Gon¸calves S, Abramson G and Vega J L (2004) Physica A 342:186
[22] Bouchaud J.-P. and Mézard M., Physica A 282:536 (2000)
[23] Mourkazel C F, Gon¸calves S, Iglesias J R, Achach M, and Huerta R The European Physical Journal - Special Topics,
143:75 (2007)

9
[24] Scafetta N, West B J and Picozzi S (2003) cond-mat/0209373v1(2002) and cond-mat/0306579v2.
[25] Hayes B, American Scientist (2002), 90:400
[26] J. Rawls, A Theory of Justice, The Belknap Press of Harvard University Press, Cambridge, MA, 1971.
[27] A. Sen: The Idea of Justice, The Belknap Press of Harvard University Press, Cambridge, MA, 2009.
[28] P. Bak, K. Sneppen, Phys. Rev. Lett. 71 (1993) 4083
[29] http://en.wikipedia.org/wiki/Gini coeﬃcient

Simple regs rdc ineq

More Related Content

Viewers also liked (9)

Similar to Simple regs rdc ineq (20)

Recently uploaded (20)

Simple regs rdc ineq