International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
Economic Agent Based Models: Review
Nawfal NACIRI* and Mohamed TKIOUAT
Islamic Financial Engineering Laboratory (IFELab).
Studies and Research Laboratory in Applied Mathematics (LERMA).
Mohamed V University-Mohammadia school of Engineering. Rabat, BP 765, Morocco.
*Corresponding author
Arthur: « Over the last twenty years a different way of doing
economics has been slowly emerging. It goes by several
labels: complexity economics, computational modeling, agentbased modeling, adaptive economics, research on artificial
economies, generative social science — each of these with its
own peculiarities, its own followers, and its own nuances.
Whatever the label, what is happening, I believe, is more than
just the accumulation of computer-based or agent-based
studies. It is a movement in economics » [3].
According to Seppecher [41], the real precursors of the multiagent approach in the economic field are, Simon [42] [43],
Cohen [15], Cohen and Cyert [16], and Cyert and March [17].
They are the real introducers of computer programming in
economics, not only from a technical point of view (for model
implementation and simulation) but also as "theoretical
language natural” for the description of the processes at work
in complex economic systems.
Other recent work of several researchers based on the results
of these precursors have joined this movement, we can
mention: LeBaron [28] [29] [30] Axtell [6], Axelrod and
Tesfatsion [4], Axtell [7] Arthur [3], Farmer and Foley [21],
Rivero et al. [40] Pajares et al. [37], Phan [39], and Yildioglu
[53] [54] [55]
Based on this theoretical movement, we can imagine
massively and ultra-realistic multi-agent models, able to
explore quantitatively the likely reactions of economic
systems under different scenarios. These models could contest
conventional neoclassical models, constituting models making
and prediction from economic policy-makers [41]. According
to Farmer and Foley “it might even be possible to create an
agent-based economic model capable of making useful
forecasts of the real economy, although this is ambitious”
[21].
Abstract
Originally, multi-agents systems (MAS) appear in the field of
computer science as a subject of research in the field of
distributed artificial intelligence. Agent based computational
economics is the computational study of economic processes
modelled as dynamic systems of interacting agents. Their
applications to economic modeling are developed gradually
since the 1980s from individual and dispersed initiatives, until
constitute a "movement". Based on this theoretical movement,
we can imagine massively and ultra-realistic multi-agent
models, able to explore quantitatively the likely reactions of
economic systems under different scenarios. In this paper, we
presented this approach: agent-based computational
economics (ACE) and its building blocks and we give a
survey of economic agent based models.
Keywords: Multi-agent simulation; Complexity economics,
Agent computational economics; Economic building blocks;
Economic agent based models.
Introduction
ABM is now developing as an important tool for policy and
decision-making and for designing tools for human activity
systems. Artificial society, the economy and markets are also
areas of application for ABM [19]. ABM applied to economic
systems by considering them as complex systems, populated
by a large number of heterogeneous agents, autonomous and
competitors, in direct interactions and without superior
control.
This approach is named, agent-based computational
economics (ACE). For Axtell [5] there are three distinct uses
of ACE: (1) classical simulations, (2) as complementary to
mathematical theorizing and (3) as a substitute for
mathematical theorizing.
In this paper, we presented agent-based computational
economics (ACE) approach and its building blocks and we
give a survey of economic agent based models, specifically, in
financial market, monetary economics, economic policy,
innovation and consumer behavior, strategies for competing
brands, and in the Islamic economic model.
Complexity economics
The basic idea of this approach is that economic systems are
complex systems, According to Al-Suwailem [1], “an
economy is one of the most complex self-organized systems.
Through simple interactions of agents based on local
information, an economy emerges where production and
distribution takes place across the whole economy without
central authority. Agents could enjoy higher welfare through
the economy than they would obtain if they were scattered in
isolated islands”.
According to Arthur [2], Complexity economics holds that the
economy is not necessarily in equilibrium, that computation as
well as mathematics is useful in economics, that increasing as
well as diminishing returns may be present in an economic
situation, and that the economy is not something given and
Multi-Agent Simulation in economic systems
Originally, MAS appear in the field of computer science as a
subject of research in the field of distributed artificial
intelligence. Their applications to economic modeling are
developed gradually since the 1980s from individual and
dispersed initiatives, but they are gradually gaining strength
and cohesion until constitute a "movement" according to
5492
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
“True” dynamics Partly as a consequence of adaptive
expectations (i. e. agents observe the past
and form expectations about the future on
the basis of the past)
Direct
Agents interact directly: The decisions
(Endogenous)
undertaken today by any agent directly
Interactions
depend − through adaptive expectations −
on past choices made by subgroups of
other agents in the population
Endogenous and Socio-economic systems are inherently
persistent novelty non-stationary.
Agents
face
“true
uncertainty”, as they are only partly able
to form expectation
Selection-based
Agents are typically selected against −
market
over many different dimensions − by
mechanisms
market mechanisms.
existing, but forms from a constantly developing set of
institutions, arrangements, and technological innovations.
Agent Computational Economics: ACE
Agent-based computational economics (ACE) is the
computational study of economies modeled as evolving
systems of autonomous interacting agents. Starting from
initial conditions, specified by the modeler, the computational
economy evolves over time as its constituent agents
repeatedly interact with each other and learn from these
interactions. ACE is therefore a bottom-up culture-dish
approach to the study of economic systems [45]. She
describes the multi-agent simulation as a third way of doing
science, in addition to deduction and induction. She
emphasizes four main objectives of the use of multi-agent
simulation:
1.
Agent-based models (ABM) can provide empirical:
understanding of macroscopic features nature
without top-down control.
2.
ABM can provide normative investigation, testing
the qualities of different designs, looking for one that
gives desirable system performance.
3.
ABM can provide heuristic investigation of market
phenomena, understanding of economic system
behaviors under alternatively initial conditions. ABM
sheds some new light on causal mechanisms in social
systems.
4.
Agent-based modeling can help researchers to get
advance in logical method issues, as it provides the
methods and tools needed to undertake the rigorous
study of social systems through controlled
computational experiments. This axis covers the
necessity in testing of experimentally generated
theories against real-world data.
Economic Agent based Models: Overview
Simulation of the financial market
Agent-based artificial financial markets can be mathematical
or computational models, and are usually comprised of a
number of heterogeneous and boundedly rational agents,
which interact through some trading mechanism, while
possibly learning and evolving. These models are built for the
purpose of studying agents‟ behavior, price discovery
mechanisms, the influence of market microstructure, the
reproduction of the stylized facts of real-world financial timeseries (e. g. fat tails of return distributions and volatility
clustering) [34]. In this part we will present an overview of
agent based artificial financial market.
NASDAQ
Market dynamics results from the behavior of many agents
acting with each other, leading to emergent phenomena. There
has been in recent years a growing interest in this type of
modeling markets, stimulated by the work of WB Arthur and
his colleagues. Commercial application was developed by
Organic Group for the National Association of Securities
Dealers Automated Quotation (NASDAQ) [W1].
The NASDAQ stock exchange used a model that simulated
(with agents) the impact of regulatory changes on the capital
market under various conditions. The model allowed
regulators to examine and predict the effects of different
strategies, observe the behavior of agents in response to
changes, and monitor developments, providing early warning
of unintended consequences newly implemented faster in real
time and without the risk of initial tests on the market. In their
model, agents‟ market maker and investor (institutional,
pension funds, traders and casual investor) buy and sell shares
using various strategies. Agent access to information of prices
and volumes are similar to those of the real market, and their
behavior varies from the simplest to the most complicated
study strategies.
The use of dynamic models (a popular method of financial
modeling using sets of differential equations) could not
produce the results as lucid and profound as the simulation
with multi-agents, because the market behavior emerges out
of the interactions of actors, which in turn can change their
behavior in response to market changes. Interactions between
Economics Agent-Based Models: Building Blocks
Economics Agent-Based Models have a common set of
qualitative assumptions that reflect their underlying modelling
philosophy. In what follows, we will present the most
important ones given by Payka and Fagiolo [38]. We have
synthesized these building blocks in the following table:
Building Blocks
Bottom-up
Philosophy
Explanations
Any satisfactorily, account of a
decentralized economy must be addressed
in a bottom-up perspective. Aggregate
properties must be viewed as the outcome
of micro dynamics involving basic entities
(agents).
The
Evolving Agents live in complex systems evolving
Complex System through time [26]. Therefore, aggregate
(ECS) Approach properties are seen to emerge out of
repeated interactions among simple
entities,
Heterogeneity
Agents are (or might be) heterogeneous in
almost all their characteristics.
Bounded
Agents are assumed to behave as
Rationality
boundedly rational entities with adaptive
expectations.
5493
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
investors, market makers, and the exploitation of the
NASDAQ rules make the dynamics of the whole system
difficult to understand [W1].
At the end of the period, one time step later, this portfolio
becomes worth
wi‟(t+1) = (1+r)Mi(t) + hi(t)p(t+1) + hi(t)d(t+1)
where the three terms are the money in the bank with interest,
the new value of the stock, and the dividend pay-out.
The trading process is managed by a specialist inside the
market. The specialist also has the job of setting the p(t+1). If
there are more bids than offers, then the price is raised, so the
bids drop and the offers increase, until they match closely [29]
[27].
SHOPBOTS
Electronic auctions can also employ multi-agents technical.
EBay allows buyers to use a robot that will bid for them
automatically, using agent systems to testing varieties of robot
behavior. Designing intelligent agents that have the overall
desired properties may be making this application the
preferred economic transactions cyber world. The Shopbots
are the Internet agents that automatically search for
information regarding the price and quality of goods and
services [W1].
A team of corporation ecosystem simulated market ISPs
(Internet Service Provider or ISP) with MAS. Each ISP is an
agent and each client is an agent. The ISP offers are faced
with the needs and expectations of customers. Customers
make decisions (adopt from, change) according to the ratio
between their profile and ISP. One of the attributes of ISP,
among other, is their monthly charge for services. It
eliminates the ISP who does not make enough money after
selecting "evolutionary". The system has produced two
significant results: First, the free ISP business model with no
monthly fees. On the other hand, found that this model is
unstable. The first ISP of this type has emerged in the
simulation and differs by providing services without charge
monthly fees or makes money on advertising. These two
properties emerge out of the dynamic interaction between the
ISP on the market. As ISPs learn and evolve, it would have
been difficult to obtain this knowledge by using other
simulation methods [W1].
Genoa Artificial Stock Market (GASM)
GASM is, characterized with heterogeneous agents, which
exhibit random buy or sell patterns and interact with each
other [36].
The purpose of the market maker is to fix the price of the
stock. It does so from the demand and supply curves. Demand
curve gives the price per stock against the demand for the
stock (ordered quantity). Similarly supply curve gives the
price per stock against the ordered quantity. The price
formation process is given by the intersection point of these
two curves [36].
Agent based model for investment (ABMI)
ABMI used for modeling and studying stock markets. The
model is based on an order-driven market. trades take place at
t, t + 1, and so on, using the following price-formation rule
[Farmer, 2001]:
where
ωi = market order of agent i
p = log price
ω^i(t + 1) = x (t + 1)-x (t)
λ = liquidity
X = market-maker position
β = market-maker risk aversion
x = position of agent i
Santa Fe Artificial Stock Market (SF-ASM)
Consists of a central computational market and a number of
artificially intelligent agents. The agents choose between
investing in a stock and leaving their money in the bank,
which pays a fixed interest rate. Agents make their investment
decisions by attempting to forecast the future return on the
stock using genetic algorithm to generate test and evolve
predictive rules [45].
The basic structure of the market is N agents, ranging from 50
to 100, interacting with the central market A single stock
exists with price p(t) per share at time t. The stock pays a
dividend of d(t+1) per share at the end of time period t. The
dividend time series d(t) is a stochastic process independent of
the market and the agents‟ actions. The dividend d(t) is given
by the simple random process [29]
d(t+1) = pd(t) + α n(t).
where p and α are parameters and n(t) is a Gaussian random
variable, chosen independently at each time t from a normal
distribution with mean 0 and variance σ.
There is also a fixed-rate asset, the bank, which pays a
constant rate or return r per period. The agents have to decide
how much money they want to put into the stock and how
much money they want to leave in the bank. At any time t,
each agent i holds some number of shares, of stock h (t) and
has some amount of cash M (t) in the bank. Its total wealth is
then given by
wi(t) = Mi(t) + hi(t)p(t)
The change in the logarithm of the price is proportional to the
sum of the net order imbalance. The first term is the sum of
the orders that are placed by the agents at time t, and the
second term is the order placed by the market maker, which is
always a fraction β of the market maker‟s current position.
The variable X, the market maker‟s total position, is the total
amount that supply and demand are out of balance in the
market. The constant of proportionality1/λ, can be thought of
as liquidity, and it determines the amount that an order of a
given size will move the price. Any particular family of agent
behaviors can be studied by specifying a set of functions x (t)
and iterating the resulting equations [Farmer, 2001]
The behavioral model involves four classes of investors:
Market makers as seen above, fundamental (or value)
investors, technical traders (or chartists), and liquidity
demanders (or noise traders) [Farmer, 2001].
Business School (BS)
The Business School [BS]is an agent-based model of a socalled “school” (actually strategies) which is used to forecast
5494
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
future values and then evolves over time as a function of their
performance. Investors update from time to time the forecast
function selected from the school if it does not predict
satisfactorily [11], [12] and [13].
especially on the social aspects of investor behavior and study
consequences of two distinct network topologies.
The U-Mart Project: a Virtual Market [52]
The U-Mart Project is a forum for interdisciplinary research
using virtual markets. The organizing committee of the UMart has developed a virtual market simulator of a futures
market of an existing stock index, demonstrated it in several
academic conferences, and carried out virtual market
experiments using it involving participants (human
traders/software agents) all over Japan via the internet. The UMart simulator is a system that participant traders access to
the U-Mart servers via the internet, and its salient feature is
that the U-Mart system enables hybrid simulation involving
both human traders and the software trading agents.
From the viewpoint of economics, to design virtual markets
and to carry out simulation using them help understanding of:
(1)
behaviors of human in trading security,
(2)
behaviors of the emerging market, and institutional
devices to control them. From the viewpoint of
computer science, they contribute to
(3)
study of evolution, learning, collective intelligence
through developing software trading agents.
Baron’s Model (BM)
The investment decisions of agents are based upon an
information set. A Walrasian auction is adopted to determine
the price. The model contains two assets for investment
LeBaron [31], [32] and [29]-cash and equity. Cash pays a
constant guaranteed rate of return rf (risk free). The equity
pays a dividend at each time step. This is random and the logdividend follows a random walk:
log (dt+1) = log (dt) + ɛ t
wheredt is the dividend and ɛ t is a Gaussian random variable
N(μ, σ). The equity is available in a fixed supply of one share
for the population. If si is the share holding of agent i, the
summation of holding of all agents will be always maintained
at unity. The equity price arises through the interactions of the
agents.
Kim and Markowitz-Portfolio Insurers Model.
Although the model of Kim and Markowitz [25] is not the first
computational study in the area of finance, it is considered one
of the first modern agent-based models of the financial
market. The motivation for this model was the stock market
crash of 1987, and the main focus of the study was exploring
the link between portfolio insurance strategies and market
volatility[34].
Further, the U-Mart project aims at the third mode of study as
alternative to theory and experiments in social sciences
through interdisciplinary activity of education and research.
Simulation of the Monetary Economics
The research team CEMAFI from the University of Nice
Sophia Antipolis (before 1 January 2012, at present CEMAFI
is no more in place [51]), built a multi-agent model for the
monetary economy. This model is composed of three types of
agents: firms, which are associated with the production
function, households, which are associated with the
consumption function, and the bank which is associated with
the management of means of payment. Businesses and
households are multiple agents, heterogeneous, autonomous
and competing direct and indirect interaction-only the banking
sector is composed of a single representative agent.
Interactions of agents operate through a particular structure,
consisting of two coupled systems, one representing the real
sphere, the other monetary sphere [41].
The real sphere is composed of elementary objects
representing the forces of labor and commodities; the
monetary sphere is composed of elementary objects
representing loans and deposits. Each of the two spheres
encapsulates a fundamental principle which applies to the
agents: in the real sphere, the production of goods takes time;
in the monetary sphere, the money is created by bank credit to
the production for a limited time. These two principles are
expressed in a multitude of parallel and asynchronous real and
monetary process (fully decentralized). At the interface
between the real and the monetary agents provide and undergo
interactions of the two spheres. Markets-the labor market and
goods market-are simple meeting places where households
and businesses establish direct relations [41].
Levy, Levy, Solomon-Microscopic Simulation.
Levy, Levy, Solomon model (LLS) is a prominent model of
the financial market based on the microscopic simulation
approach which has roots in physics. It is a numerical model
developed in the framework of expected utility maximization.
[33].
Lux and Marchesi-Stochastic Interaction and Scaling
Laws.
The model of Lux and Marchesi [35] is an agent-based model
of the financial market that follows the tradition of earlier
attempts to capture herding behavior by means of stochastic
modeling. An example of this is the ant recruiting model of
Kirman [26], which has also been proposed as an analogy for
herding behavior of investors in the financial markets.
Takahashi and Terano-Investment Systems Based on
Behavioral Finance.
Although various behavioral aspects of agents, including
investor biases, have been studied in earlier literature, the
model of Takahashi and Terano [44] is to our knowledge one
of the first agent-based models that explicitly studied a
number of investor biases proposed in the behavioral finance
literature.
Hoffmann et al.-SimStockExchange Model.
Following the tradition of Takahashi and Terano [44] the
paper of Hoffmann [23] is another study that combines a
number of behavioral phenomena within an agent-based
simulation of the financial market. Hoffmann [23]focus
5495
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
Passive agents
The passive agents are [46]:

Local outlet malls: agent that performs the selling
role of the firms in the region. It receives inventory
stock from each firm, then sells the goods and
collects sales revenues that are reported back to the
firms.

Eurostat: collects aggregate statistics at the macro
level and publishes it to the public.

Market research entity: collects data on regional
markets (anticipated local demand and market share
for several outlet malls) and reports this data to the
firms.

Clearinghouse: collects stock orders, computes
clearing prices and transactions. It replaces the
market clearing equation on the financial market.

Financial Advisor: an agent who collects information
about the past performance of a set of portfolio
allocation rules (investment strategies), and then
distributes this information to the individual
investors (this is used in the learning algorithm).
EURACE economic model
EURACE is a major European attempt to construct an agentbased model of the European economy with a very large
population of autonomous, purposive agents interacting in a
complicated economic environment [18],
FLAME. Flexible Large-scale Agent Modeling Environment
Euracewill be implemented using the Flexible Large-scale
Agent Modeling Environment (FLAME) developed by Simon
Coakley, Mike Holcombe, and others at the University of
Sheffield (see http://www. flame. ac. uk. FLAME provides a
computational framework allowing modellers to easily create,
exchange, include and couple models written in a high-level
modeling language. [18 ]
In the FLAME framework, agents act within contexts, that is
preeminently, within markets. The agents can have different
roles in different contexts, e. g., an agent can be a buyer on
one market, a seller on another. The framework is based on
the logical communicating extended finite state machine
theory (Xmachine) which gives the agents more power to
enable writing of complex models for large complex systems.
The agents are modelled as communicating X-machines
allowing them to com-municate through messages being sent
to each other as per designed by the modeler [14].
Main markets in Eurace
Figure 2. 3 presents a general picture containing the agents
populating, the Eurace model is represented along with the
main markets where they are involved.
Categorization of agent classes
In the Eurace model we have fundamentally two classes of
agent: active agents who take decision, and passive agent who
do not [47].
Active agents
Table 2. 3 lists the main classes of active agents in Eurace, the
contexts in which they operate and the main messages they
exchange [46],
Table 5. 3: Main agents, contexts, roles, and messages in the
Eurace model [46]
Figure 5. 3: A general scheme of the main agents and markets
included in the Eurace model.
The arrows show the main communication network of the
agents through the different markets [14].
In the following we present brief definitions for the different
type of markets included in the Eurace model [46]:
1
Labor Market: local, decentralized. Workers only
search locally for jobs and there is bilateral
interaction between the job searchers and firms.
2
Investment Goods Market: global, decentralized. We
assume there are only few investment good
producers. Consumption good producers can go to
any investment good producer to purchase capital
goods, so the market consists of global interaction.
5496
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
3
4
5
There is bilateral interaction between the buyers and
sellers.
Consumption Goods Market: local, decentralized.
Consumers can go to any local outlet mall to buy
consumption goods, and there is bilateral interaction
between the buyers and sellers.
Credit Market: global, decentralized. Firms can get
loans from any bank, and they can go to any bank
they wish depending on the credit conditions (e. g.
the interest rate) that is offered.
Financial Market: global, centralized. There is only a
single global financial market on which all stocks
and bonds can be traded, and the market mechanism
is either a clearinghouse or a limit-order book.
X-Economy
The X-Economy project develops tools for multi-agent
simulation of artificial economic markets. The system
developed is composed of an X-Economy server and multiple
clients connected to this server module such as banking, stock
exchange, stock market, and a mediator negotiation. The
server manages the flow of information circulating between
various clients in the simulation. It routes messages sent by
clients to update their states and control the different phases of
the simulation. The project aims to develop from this first
implementation operating several extensions to test the system
on real financial markets [Bouron et al, 2004 ].
Simulation of innovation and consumer behavior
InfoSumers and Sphere of Influence
TheInfoSumers systems and Sphere of Influence [10], are
multi-agent simulation performed to study the diffusion of
innovation fashions in a population and the influence of
interactions between suppliers and consumers on the structure
of the textile market. Implementation of agent behavior
remains ad hoc. It does not describe a method applicable to
different market situations and does not introduce behavioral
models based on generic principles of consumption. Modeling
interactions between agents is limited to the application in
question and does not apply to a competitive environment [9 ].
The EURACE model follows a fundamentally different
approach from the dynamic stochastic general equilibrium
models (DSGE) which are currently employed by the
European Central Bank (ECB) or the European Commission
on Economic and Financial Affairs (ECFIN). The project
objectives are characterized by scientific, technological and
societal scopes [49].
From the scientific point of view, the main effort regards the
study and the development of multi-agent models that
reproduce, at the aggregate economic level, the emergence of
global features as a self-organized process from the complex
pattern of interactions among heterogeneous individuals.
From the technological point of view, the project will develop,
with advanced software engineering techniques, a software
platform in order to realize a powerful environment for largescale agent-based economic simulations. Key issues will be
the definition of formal languages for modeling and for
optimizing code generation, the development of scalable
computational simulation tools and the standardization of data
with easy to use human-machine interfaces.
From the social point of view, the agent-based software
platform for the simulation of the European economy will
have an outstanding impact on the economic policy design
capabilities of the European Union. It will be a powerful tool,
enabling to perform “what-if” analysis, optimizing the impact
of regulatory decisions that will be quantitatively based on
European economy scenarios.
Consumat
Consumat is a multi-agent simulation of behavioral rules that
tests consumer under the management of shared resources.
This simulation, study particularly the retroactive effects of
satisfaction and the impact of uncertainty on the behavior of
individuals. This project is located on the margins of the
simulation of consumer populations to the extent which it
implements a twenty agents. The conceptual model proposed
is composed of four systems describing: the driving forces of
behavior, a set of states on cognitive processes, the results of
the individual behaviors, and strategies for behavioral change.
This conceptual model of consumer behavior provides a
probable tool analysis of consumer behavior but its
complexity makes it unsuitable for simulations requiring
several hundred individuals [9 ].
Simulation of strategies for competing brands: CUBES
(Customer Behaviors Simulation)
CUBES (Customer Behaviors Simulation) is a software-based
agent that simulates a population of several thousand
consumers and test the reactions of a market face different
brand strategies, putting more or less emphasis on the
innovation, promotion, branding. CUBES offers a new
experimental approach to marketing to help marketing
managers and strategic managers to develop their brand
strategies [8]. Design model CUBES relies on several
assumptions [9]:

The study of consumer behavior is based essentially
on observation, analysis and modeling of interactions
between a consumer individual and his environment.

Differences in consumer behavior are largely due to
their basic social and behavioral characteristics of the
individual that are not specific purchasing behavior.
Simulation of economic policies
ASPEN
ASPEN is a micro-simulation of the U. S. economy that can
analyze and compare different economic policies. This system
models the impact of changes in laws, monetary policy and
taxation policies on a population of households.
Macroeconomic variables such
as inflation and
unemployment rates are calculated as the aggregation results
of individual decisions. Agents are used to represent the
various actors involved in decision-making in economics.
They simulate the behavior of banks, different types of
businesses (food, buildings and other products), government,
Federal Reserve, the stock exchange and households. ASPEN
implements a mechanism of interaction between agents in the
system. Genetic algorithms are used to simulate learning in
some agents [ 9].
5497
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com

The design of a behavioral model of the consumer
may be based on factors related to the individual and
his environment regardless of market conditions and
the specific product as opposed to a rationalizing
approach behavior relative to price criteria or
qualitative assessment of the product.
ATOM can be used to evaluate execution strategies and can
be tuned to serve as ”best-execution” expert-system;
Artificial price motions span over different time horizons and
granularities: from ultra-high frequency (tick-by-tick data) to
daily closing prices;
The platform is multi-asset oriented and any kind of matching
or execution mechanism can be implemented.
ATOM
This section provides a definition of this platform, its various
features and components, as well as its methodology and
applications. It is derived mainly from the following reference
[24]
Scientific research
ATOM is modular. It can be viewed as a 3 meta-components
system:
i)
Agents and their behaviors;
ii)
Market microstructure and
iii)
Artificial Economic World.
Definition
ATOM (ArTificial Open Market) is a general environment for
agent-based simulations of stock markets. Platform
independent, fully flexible, ATOM can perform distributed
simulations as well as local-host, fast simulations. It can also
be used for experiments mixing human beings and
autonomous, artificial traders. Agents are endowed with
Artificial Intelligence and range from simple zero-intelligence
robots to highly evolved cognitive agents.
Each of these components can be used independently as well
as in complete interaction.
ATOM can be used for the evaluation of new regulation or
market procedures, or to assess the potential effects of taxes or
new trading strategies in a sophisticated artificial financial
environment.
Multi-agent simulation in the Islamic economic model
Al-Suwailem [1] provided in his book “Islamic economics in
a complex world”, in the first time, an application of agentbased simulation in Islamic economics that shows how riba
(interest), markup finance, and zakah (Charity) affect
economic performance.
The model is implemented using an agent-based environment.
The one chosen is NETLOGO. There are 1225 agents
modeled as patches on a landscape. Each agent receives an
income and decides his consumption accordingly.
Main Features
The characteristics and properties of the platform ATOM are
presented below:
1.
Agents trade on a single market defined by its
microstructure. Competition among several trading
systems can be programmed;
2.
Market hosts as many financial commodities as
needed;
3.
Each commodity is traded through a central order
book mimicking the Nyse-Euronext stock-exchange
(this microstructure can be swapped for another one
if necessary);
4.
Order formulation covers any usual type of orders,
from the most basic ones (limit orders) to
sophisticated orders (stop orders, mid-match point
orders... );
5.
Validity rules are taken into account as well as order
updates and cancel orders;
6.
Any simulation can include human agents interacting
with Artificial Intelligence Agents;
7.
The platform can serve as a replay engine for a
whole trading day (using real-world order flows). It
then delivers the same results as the original stock
market;
8.
Users can monitor the entire agent‟s population and
get access to each of the artificial agents‟ parameters.
Financing results
The results include three models: interest-free lending,
interest-based lending, and markup-based financing. Each
model was run for 1500 periods, and repeated 50 times (each
with a different random seed number). The averages of the 50
runs are summarized in the following table.
Standard denotes the model with interest-free lending. Std
denotes standard deviation scaled by the mean. Gap denotes
the difference between the mean value of the upper half of the
population and that of the lower half, scaled by the median.
The last row represents number of agents with value below
50% of the maximum in the population. Net-income
represents income plus agent‟s share in markup or interest,
minus amount due [1].
Applications
Technological research and Applied Quantitative Finance
ATOM can be used in various fields of quantitative Finance:
Algorithmic Trading, Risk Management, and Portfolio
Management among others;
The platform can serve as a replay-engine using real-world
order flows or generate new typical market regimes (”high
volatility”, ”bear market”... ). In both of these cases, ATOM
offers an environment to test algorithmic trading methods;
5498
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
4.
5.
financing economically indistinguishable from
interest financing.
The interest model demonstrates several undesirable
results. These include: cycles and instability, wealth
concentration, low fund utilization and thus high
inefficiency, and high costs of borrowing. In
contrast, the markup model shows stability, high
efficiency, less concentration of wealth, and
relatively low costs of financing.
The markup model is slightly more concentrated than
the standard model, but it shows higher savings and
liquidity. Although not modeled at this stage, higher
savings potentially promote higher investment and
higher economic growth.
Charity
The assumption of relative behavior also predicts that charity,
i. e. transfers from high net-worth agents to low net-worth
ones, might improve consumption of both, donors and
receivers. The results below are obtained from simulations
[1]. The following table summarizes statistics of consumption
at different charity structures.
Summary of Financing Models
According to Al-Suwailem, the above results show that
different financing models have dramatically different
economic impacts. This is generally consistent with Muslim
economists long standing on the subject. But the results
provide further insights that were not easily deducible [1]:
1.
Markup financing is fundamentally different from
interest or riba financing. Despite some apparent
similarities, they have starkly different impacts on
economic variables.
2.
Interest is not merely a price; it is a mechanism for
debt growth. This is not the case for markup
financing. Markup restricts debt to actual
consumption, thus controlling the growth of debt to
that of available resources. Interest in contrast allows
debt to grow exponentially. Debt growth in this case
is controlled not by interest but through bankruptcy.
3.
We have seen that delayed payments are treated quite
differently in the two models. In the interest model,
they represent a source of income to lenders. This
leads to very low turnover of funds, explosion of
debt and consequently to economic cycles. The
markup model does not allow for generating profits
from these obligations. This puts an incentive to
creditors not to extend credit beyond a level that
would permit for late payments to appear in the first
place. This can be contrasted to some legal artifices
that allow creditors in practice to gain from late
payments. Not only this contradicts the principles
and logic of Shari‟ah, but also makes markup
There is an overall increase in average and median
consumption under all structures. Minimum consumption has
increased notably, while standard deviation and gap decreased
notably also. Number of agents with consumption below 50%
of the maximum level has dropped significantly [1]. Overall,
charity was effective in reducing inequality in consumption
and raising its aggregate level.
The table also compares consumption of both donors and
receivers. Consumption of receivers has increased
substantially, while that of donors increased only marginally.
The increase for the 2. 5% and 5% rates however is
statistically significant at the 5% level (test statistics are not
reported) [1].
5499
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
Bibliographies
Interest of study
Al-Suwailem demonstrated in his study using multi-agent
simulation in the Islamic economic model that from an
Islamic point view, usury or riba have a radically different
impact in presence of relative considerations, giving another
rationale for its legal prohibition. Riba allows agents to evade
independent variables governing the economic system,
thereby making the economy dominated by relative behavior,
and pushing the system away from the phase of complexity
towards the phase of chaos. Markup financing in contrast does
not suffer these consequences, and thus no legal restrictions
are needed. Finally, charity appears to have a greater positive
impact in a complex economy, reinforcing the old wisdom [1].
1.
2.
3.
4.
Conclusion
In this part, we have we presented the origin of multi-agent
simulation applications in the economic field, and its real
precursors. This approach considers that economic systems
are complex adaptive systems (CAS) as they have supported
[2] “Complexity economics builds on the proposition that the
economy is not necessarily in equilibrium: economic agents
(firms, consumers, investors) constantly change their actions
and strategies in response to the outcome they mutually
create. ” And because they have diverse producers, firms
which are connected to suppliers in a supply network and
interdependent between them; also they‟re adapting to a
changing environment.
We have also synthesized the building blocks of an economic
agent based model, namely: bottom-up philosophy, the
evolving complex system approach, heterogeneity, bounded
rationality, true dynamics; direct endogenous) interactions;
endogenous and persistent novelty and selection-based market
mechanisms
The overall objective of this work is to give a survey of
economic agent based models in the various branches,
specifically, in financial market, monetary economics,
economic policy, innovation and consumer behavior,
strategies for competing brands, and in the Islamic economic
model.
This paper reinforces interest of the use Agent-based
computational economics (ACE). Thanks to agent-based
models (ABM), we can provide, empirical understanding of
macroscopic features nature without top-down control;
normative investigation, testing the qualities of different
designs, looking for one that gives desirable system
performance; and heuristic investigation of market
phenomena, understanding of economic system behaviors
under alternatively initial conditions. ABM sheds some new
light on causal mechanisms in social systems.
Agent-based modeling can help researchers to get advance in
logical method issues, as it provides the methods and tools
needed to undertake the rigorous study of social systems
through controlled computational experiments. This axis
covers the necessity in testing of experimentally generated
theories against real-world data.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
5500
Al-Suwailem, Sami Ibrahim 2008: Islamic
Economics in a complex world. Islamic
Development Bank,
Arthur, 2015: W. Brian Arthur “Complexity and the
Economy”. Oxford University Press 2015
Arthur, 2006: Out-of Equilibrium Economics and
Agent-Based Modeling, ” dans Handbook of
computational economics, ed. L. Tesfatsion, et K. L.
Judd, vol. 2, p. 1551-1564
Axelrod et Tesfatsion 2005: A guide for newcomers
to agent-based modeling in the social sciences.
Appendix in Handbook of Computational
Economics. Agent-Based Computational Economics,
2. 6, 19.
Axtell R. 2000 “Why Agents? on carried motivations
for agent computing in social sciences” WP17 Center
on Social and Economic Dynamics The Brooking
Institution
Axtell 2006: Multi-Agent Systems Macro: A
Prospectus. Post Walrasianmacroeconomics: beyond
the dynamic stochastic general equilibrium model.
pp. 203-220, 2006
Axtell 2007: What economic agents do: How
cognition and interaction lead to emergence and
complexity, ” The Review of Austrian Economics,
20, 105-122.
Ben said, 2003: Simulation multi-agents des
consommateurs dans un contexte concurrentiel, thèse
de l'Université Paris 6, 13 juin 2003.
Bouron et al, 2004: Thierry BOURON et Lamjed
BEN SAID: Simulations multi-agents d'écosystèmes
d'entreprises. 2004
Brannon, 1994: Artificial life simulation of the
textile apparel marketplace: An innovative approach
to strategizing about evolving markets, http://www.
ntcresearch. org/pdf-rpts/AnRp96/R96A9520. pdf,
1994.
Chen and Yeh2001: S.-H. Chen, C.-H. Yeh,
Evolving traders and the business school with genetic
programming: a new architecture of the agent-based
artificial stock market, Journal of Economic
Dynamics and Control 25 (2001) 363-393.
Chen and Yeh 2002a S.-H. Chen, C.-H. Yeh, On the
emergent properties of artificial stock markets,
Journal of Economic Behavior and Organization 49
(2002) 217-239.
Chen and Yeh 2002b S.-H. Chen, C.-H. Yeh, C.-C.
Liao, On AIE-ASM: a software to simulate artificial
stock markets with genetic programming, in: S.-H.
Chen (Ed. ), Evolutionary Computation in
Economics and Finance, Physica-Verlag, Heidelberg,
2002, pp. 107-122.
Cincotti et al, 2011. SilvanoCincotti, Marco Raberto,
Andrea Teglio: The EURACE macroeconomic
model and simulator. 16th IEA World Congress 2011
Cohen 1960: Simulation of the Firm, The American
Economic Review, 50(2), 534-540. 1960
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
Cohen and Cyert 1961: Computer Models in
Dynamic Economics, ” The Quarterly Journal of
Economics, 75(1), 112-127
Cyert and March 1963: A Behavioral Theory of The
Firm. Prentice-Hall, Englewood Cliffs, New Jersey.
1963
Deissenberg et al 2008. Christophe Deissenberg,
Sander van der Hoog, Herbert Dawid: EURACE: A
massively parallel agent-based model of the
European economy. Applied Mathematics and
Computation 204 (2008) 541-552.
Deguchi Hiroshi 2004: “Economics as an AgentBased Complex System. Toward Agent-Based Social
Systems Sciences” Originally published by SpringerVerlag Tokyo in 2004
Diletta 2011: DilettaGrella: coping with financial
crises: A multi-disciplinary agent based research
project. CRISIS Brochure 2011
Farmer and Foley 2009: The economy needs agentbased modelling, Nature, V460/6 august 2009
Farmer, J. (2001), Toward agent-based models for
investment, benchmarks and attribution analysis,
Working paper, Association for Investment
Management and Research
Hoffmann et al. 2007 (A. O. I. Hoffmann, W. Jager,
and E. J. H. V. Eije. Social simulation of stock
markets: Taking it to the next level. Journal of
Artificial Societies and Social Simulation, 10(2),
2007)
Iryna et al, 2010: Iryna VERYZHENKO and Olivier
BRANDOUY and Philippe MATHIEU 2010: Iryna
VERYZHENKO and Olivier BRANDOUY and
Philippe MATHIEU 2010: ATOM: Agent‟s minimal
intelligence calibration for realistic market dynamics
Kim and Markowitz 1989: G. Kim and H. M.
Markowitz. Investment rules, margins, and market
volatility. Journal of Portfolio Management,
16(1):45-52, 1989
Kirman 1993: Ants, rationality, and recruitment. The
Quarterly Journal of Economics, 108(1):137-156,
1993
Kumar et al, 2010: PN Kumar, Ashutosh Jha,
Gautham TK, Jitesh Mohan, Rama Subramanian AJ,
VP Mohandas. “A survey of call market (discrete)
agent based artificial stock market”. (IJCSE)
International Journal on Computer Science and
Engineering Vol. 02, No. 09, 2010, 3025-303.
LeBaron 2001: A builder‟s Guide to Agent-based
Financial Markets" quantitative finance, V1 n° 2 p
254-261
LeBaron 2002: Building the Santa Fe Artificial Stock
Market, Brandeis University 2002
LeBaron 2010: Active and Passive Learning in
Agent-based Financial Markets, International
Business School Brandeis University July 2010
LeBaron[2] 2001. Empirical regularities from
interacting long-and short-memory investors in an
agent based stock market. IEEE Transactions on
Evolutionary Computation, 5(5):442-455, Oct 2001.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
5501
LeBaron[3] 2001. Evolution and time horizons in an
agent based stock market. Macroeconomic
Dynamics, 5:225-254, 2001.
Levy et al. 2000 (M. Levy, H. Levy, and S. Solomon.
Microscopic Simulation of Financial Markets: From
Investor Behavior to Market Phenomena. Academic
Press, 2000).
Loyri 2011: Milan Lovri´: “Behavioral Finance and
Agent-Based Artificial Markets”. THESIS. Erasmus
University Rotterdam. 2011
Lux and Marchesi1999: (T. Lux and M. Marchesi.
Scaling and criticality in a stochastic multi-agent
model of a financial market. Nature, 397(6719):498500, 1999.
Marchesi et al 2005: Michele Marchesi,
SilvanoCincotti, Sergio Focardi, Marco Raberto,
Development and testing of an artificial stock
market, Genoa Artificial Stock Market, September,
2005.
Pajares et al, 2004: J. Pajares, C. Hernández-Iglesias,
and A. López-Paredes. Modeling learning and r&d in
innovative environments: a cognitive multi-agent
approach. J. ArtificialSocieties and Social
Simulation, 7(2), 2004.
Pyka and Fagiolo 2005: Andreas Pyka, Giorgio
Fagiolo. “Agent-Based Modelling: A Methodology
for Neo-Schumpeterian Economics”. Beitrag Nr.
272, Februar 2005
Phan, 2004: Cognitive Economics, chapter From
Agent-Based Computational Economics towards
Cognitive Economics, pages 371_398. Springer
Verlag, 2004.
Rivero et al, 1999: S. L. M. Rivero, B. H. Storb, and
R. S. Wazlawick. Economic theory, anticipatory
systems and artifficial adaptive agents. Brazilian
Electronic Journal of Economics, 2(2):19, 1999.
Seppecher2011: modélisation multi-agents d‟une
économie monétaire de production (thèse). 2011
Simon 1955: Simon 1955:“A behavioral model of
rational choice, ” The Quaterly Journal of
Economics, 69, p 99-118.
Simon, 1959: Herbert Simon. Theories of decisionsmaking in economics
Takahashi and Terano 2003: H. Takahashi and T.
Terano. Agent-based approach to investor‟s behavior
and asset price fluctuation in financial markets.
Journal of Artificial Societies and Social Simulation,
6(3), 2003. )
Tesfatsion, L. (2004), „Notes on the Santa Fe
artificial stock market model‟ http://www. econ.
iastate. edu/classes/ econ308 /tesfatsion/sfistock. htm
van der Hoog and Deissenberg 2007: Sander van der
Hoog,
Christophe
Deissenberg:
Modelling
Specifications for EURACE. 2007
Van der Hoog and Deissenberg 2008: Sander van der
Hoog,
Christophe
Deissenberg:
Modelling
Guidelines for EURACE. 2008
http://www. limsi. fr/Actualites/These/patrice jacque
http://www. eurace. org/
http://www. crisis-economics. eu/
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 8 (2016) pp 5492-5502
© Research India Publications. http://www.ripublication.com
51.
52.
53.
54.
55.
http://www. unice. fr/CEMAFI/
W5: http://www. u-mart. org/html
Yildizoglu 2001 (1): Connecting adaptive behaviour
and expectations in models of innovation: The
potential role of artifficial neural networks. European
Journal of Economics and Social Systems, 15(3),
2001.
Yildizoglu, 2001 (2): Modeling adaptive learning: R
and d strategies in the model of nelson and winter
(1982). In DRUID's Nelson and Winter Conference,
Aalborg, Danemark, 12-15 Juin 2001.
Yildizoglu, 2002: Competing and strategies in an
evolutionary industry model. Computational
economics, 19:51_65, 2002.
Biographical notes
Nawfal NACIRI, earned his Polyvalent state
Engineering Degree (2006) from ENSAM
(L‟ÉcoleNationaleSupérieured‟ArtsetMétiers)
Morocco-France. PhD Student in Studies and
Research Laboratory in Applied Mathematics
(LERMA) and Islamic Financial Engineering
Laboratory (IFELab), in Mohammadia school of
Engineering, Mohammed V University, Rabat. He
is in Charge of studies at the Head of Government
of the Kingdom of Morocco. His main areas of
research include Complex systems, artificial
intelligence, Agent Based Modelling and
simulation, complexity economics, Financial
Engineering.
Mohamed Tkiouat, earned his doctorat of
Mathematic
Sciences
(1991)
from
the
UnivertsitéLibre de Bruxelles (ULB) on markovian
and semi-markovian models and applications,
Belgium. Previously, he completed his third cycle
thesis (1980) in Operations Research at Faculté des
Sciences de Rabat on perturbed markovian chains
models
and
application
to
hydropower
management, prepared at INRIA France . He is a
professor at Ecole Mohammadia d‟Ingénieurs,
Université Mohamed V Agdal à Rabat. His main
areas of research include Markovian models and
applications to reliability and finance, risk
management and multi-agent games theory models.
5502