4th World Congress of Social Simulation, TAIPEI, 2012
Special Session: Applications of Social Simulation in Politics and Public Spheres
Macroeconomic forecasting with agent-based models.
Prediction and simulation of the impact of public policies on SMEs.
Federico Pablo-Martí
Antonio García-Tabuenca
Juan Luis Santos
María Teresa Gallo
María Teresa del Val
Tomás Mancha
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
Institute of Economic and Social Analysis, University of Alcala (Madrid, Spain)
Plaza de la Victoria, 2. 28802 Alcala de Henares
Tel. no.: 91 885 52 25 Fax: 91 885 52 11
Abstract
The aim of this paper is to show and discuss an integrated scheme of the MOSIPS project. This
project, under an agent-based model methodology, focuses on simulating and evaluating
policies for small and medium enterprises within a local environment. Its purpose is to conduct
experiments on the implementation of policies according to different socio-economic scenarios.
The results of these experiments allow stakeholders and citizens knowing how the measures
proposed by governments affect them and enable them to interact in the decision-making
process by relying on first-hand information. Participation through social networks reinforces
this interaction.
This perspective allows analyzing and implementing the actions, interactions and results
between agents. At the micro-level, the model provides two main kinds of agents: individuals
and firms, and some other entities: the public sector, the financial system, and the external
sector. It also determines rules for actions and interactions. Simulations on variables and
parameters lead to the analysis of the evolution and the forecast of the system at the
macroeconomic-level. The trajectories of the agents offer a dynamic standpoint that overcomes
the standard equilibrium models.
All this process drives to complex models, even long term, through the successive performance
by individuals and firms, in each period, of the short-term models. Thus, the behavior of the
model in the long-term becomes the update, period by period, of the trajectories followed by the
agents in the short-term models.
Keywords: Agent-based modeling in macroeconomics, prediction and simulation, policy
evaluation.
JEL: E17, L53, R12, R58
Acknowledgements The work presented in this document has been conducted in the context of the EU Framework
Programme project with Grant Agreement 288833 MOSIPS (Modeling and Simulation of the Impact of public
Policies on SMEs). MOSIPS is a 36 months project started on September 1st, 2011.
The project consortium is composed by: Anova IT Consulting (ANOVA), Universidad de Alcalá (UAH), Research
Studio Austria Forschungsgesellschaft (RSA), University of Reading (UoR), Opera 21 (Opera21), University of
Konstanz (Konstanz), European Institute of Interdisciplinary Research (EIIR), Ayuntamiento de Madrid
(MUNIMADRID) and Comune di Verona (VERONA)
1. Introduction
The traditional forecasting models assume that a representative agent—by maximizing
its utility—solves a social problem with which it is faced. However, this neglects the
fact that the economy is a complex and evolving system made up of diverse and
heterogeneous interacting agents (Wolf et al., a, 2011). As a result, such models present
major shortcomings. One of the most important ones is that often, by their nature, such
models fail to provide an overview of the phenomenon studied, as they are concerned
with analyzing specific aspects of reality without factoring in the actions and
interactions between the participating agents. Thus, these non-general models often
involve the concern of many endogenous aspects as fixed variables.
Even those models provide an overview, such as the dynamic stochastic general
equilibrium (DSGE) models, often refer to global rather than local areas and reveal
further important constraints such as the non-inclusion of the heterogeneity of agents.
These models sometimes include various types of individuals, but they do not capture
individual differences (Colander et. al, 2008). Other limitations are the controversial
assumptions of rationality and perfect information (Kirman, 2010) or, in a
macroeconomic area, difficulties in incorporating the endogenous emergence of crises
into the analysis (Committee on Science and Technology, 2010).
Moreover, the implementation of policies generally produces an impact that goes
beyond the considered area of analysis (e.g., spillover effects, both positive and
negative) and such impact is not usually considered in the models or is considered just
partially.
These reasons justify the need to find an ideal modeling that will make it possible to
study in detail and forecast the effects that economic policies have on the business
sector.
Consequently, agent-based models (ABM) have increased their importance in
Economics. In particular, the last financial crisis was not predicted by standard
macroeconomic models. Due to several of their assumptions, they were not able to
represent that significant deviation from the equilibrium growth path predicted. In
contrast, if the approach is bottom-up, starting with the specification of the agents
involved in the economy, it appears an emergent behavior of the system which cannot
be explained from the behavior of the representative agent. This allows the appearance
of ‘bubbles’ followed by a sharp reduction in prices and a lowering in expectations.
Multi-agent models have been used to study economic systems in several ways: we can
find examples of conceptual works on agent-based economic models, such as Tesfatsion
and Judd (2006); a variety of agent-based models focused on a part of the economy, for
example, leverage effects in financial markets (Farmer and Foley 2009). Multi-agent
models of the economy as a whole are infrequent, examples are the models by Gintis
(2006), Dosi et al. (2008), Madel et al. (2009) and the EURACE model (Dawid et al.,
2011).
The MOSIPS project includes a number of features of the previous referred models, but
it represents the economy making the emphasis in actions and interactions of two basic
types of agents: individuals and firms (SMEs basically), and also some other entities
(public sector, the financial system, and external sector) participate in the process.
Therefore, at the micro-level, the model provides these agents and entities. It also
determines rules for actions and interactions. Simulation on variables and parameters
used lead to the analysis of the evolution and forecast of the system at the
macroeconomic-level. The trajectories of the agents offer a dynamic standpoint that
overcomes the standard equilibrium models.
Considering the aforementioned, the aim of this paper is to show and discuss an
integrated scheme of the MOSIPS project. Under an agent-based model methodology,
MOSIPS focuses on simulating and evaluating policies for small and medium
enterprises within a local environment. It tries at conducting experiments on the
implementation of policies according to different socio-economic scenarios. The results
of these experiments allow stakeholders and citizens know how the measures proposed
by governments affect them particularly and enable them to interact in the decisionmaking process by relying on first-hand information. Participation through social
networks reinforces this interaction.
The paper’s structure is inspired in the formal criteria and guide recommendations of
the Dahlem Conference at the time of writing and presenting a work on new challenges
for the development of agent-based models. It also takes as a methodological guide the
multi-agent Lagom Regio model (Wolf et al., b, 2011) and includes a number of
features of the multi-agent models as the ones of Lagom family multi-agent model
developed within the World Climate Forum. The second section provides an overview
that includes the model’s rationale, the type of relationships that determine the
interactions and activities developed by the agents. The third section concerns the policy
design in MOSIPS model. Starting with the Small Business Act of the European
Commission, we arrive to policy domains, which are implemented in every module of
the model. Finally, the fourth section draws the main conclusions.
2. Overview of MOSIPS model
2.1 Rationale
MOSIPS model represents the dynamics of behavior and decisions of agents, and their
interactions. It forecasts the evolution of an economic system over a time horizon of one
quarter to several years. It is based on a multi-agent approach at the micro-economic
level. It can be used to model macro-economic features of a system and allow focusing
in a specific part of the economy, at sector and spatial level. Taking into account the
firms and individuals’ characteristics, the model evaluates the effects of policies at a
local level over them.
The main components included in MOSIPS model are the following (see figure 1):
AGENTS have these properties: reactivity, proactivity, social skills and autonomy.
They are:
i) INDIVIDUALS: people in the place under study the previous period.
ii) IMMIGRANTS: people arriving in period t, to the place under study.
iii) HOUSEHOLDS: groups of individuals of the same family living together.
MOSIPS system aim is not to simulate and foresee the effects of public policies on
these agents, although is also computable. They are taken into account as a kind of
agent as significant as firms as they are the owners of the firms. They are the workers,
and the final purchasers of goods and services produced by enterprises. They are located
in the territory, both in their residence and their workplace, and they are interacting
most of the time with a number of firms close to their location, and with some other
individuals as their family or their work colleagues. The model uses information about
how many there are, where are located, where they work, who are their relatives, which
is their level of education, their mobility patterns, their income and their expenses.
iv) PRODUCTIVE FIRMS, which are owned by individuals.
v) ESTABLISHMENTS are the places where the firms develop their activity.
Establishments and firms are the key agent in MOSIPS system. In fact, the model is
created to know the effect of public policies over them, which can took place in several
ways, directly or indirectly to individuals and firms. MOSIPS uses information relative
to the activity level of the agent, their surplus, the location of their establishments, R&D
investment, ownership, workers, competitors and if they cooperate with other firms.
In addition to these two basic types of agents, there appear OTHER complementary
ENTITIES for their activities involved to a higher or lesser extent in the modeling
process but which are pivotal in the composition of agents. They do not make decisions
directly in the process, but the evolution of their behaviors in time clearly impacts on
the creation of the expectations and decisions of firms and individuals. These entities
are:
i) FINANCIAL FIRMS play the mail role to facilitate funding to whichever
agents.
ii) PUBLIC SECTOR represents the set of different authorities.
iii) EXTERNAL SECTOR is the aggregation of firms and individuals not
located in the place under study.
Several actions taken by agents in period t have effects over other kind of agents in the
following period t+1 (pointed in the figure in purple, red and orange)
There are three MARKETS where agents and other entities relate to each other: labor,
financial and goods and services market. Agents, other entities and markets are affected
by the MACROECONOMIC ENVIRONMENT and PUBLIC POLICIES. The
macroeconomic environment and public policies have a lot of influences over the
decisions of the individuals and enterprises. The model uses a high number of
macroeconomic information in order to create agents’ expectations.
Within a local, regional or national economy, both firms and people should be placed
with their individual characteristics. In addition, government, financial sector and the
external sector are represented as well, as they interact with SMEs establishing policies,
giving access to finance, competing with them or allow selling part of their production
abroad. However, MOSIPS represents the whole economy of a territory emphasizing in
SMEs and the factors faced in their creation and growth. SMEs' suppliers, workers and
consumers tend to be near their location. In addition, the demand SMEs face is
determined by their location and their size, which conditions their visibility. Then, it is
crucial to locate every agent in its actual place to allow determining realistic interaction
networks and the correct performance of every firm.
The raster of locations allows locating every agent, and supports to establish the links
between them in a more realistic way in comparison with other possibilities that do not
have the location into account. Then, families only demand goods and services near
their residence and their workplace; however they can buy often in other areas as the
downtown. Individuals working far for their residence will tend to move either from
their house of or from their job, as they tend to be displeased. The degree of competence
in an area and the density, among other factors, are determinants of the level of prices.
The more items on location we consider in the model the more faithful representation of
a region.
MOSIPS model provides the framework to test the accuracy of micro-foundations
specified outside the scope of the representative agent paradigm. The model tries to
reproduce a virtual reality to evaluate the effects of economic policy. The obtained
results have a range of error due to the randomness of individual processes and the
building of the database. This approach can be seen as an extension to Arrow-Debreu
general equilibrium theory (Arrow and Debreu, 1954), as the unique result computed by
standard models is one of the possible outcomes: the optimal trajectory excluding part
of the heterogeneity of the agents, and not having into account spatial issues with a
sufficient degree of accuracy.
Figure 1. Main components included in MOSIPS model
Source: Own elaboration
2.2. Interactions
The types of relations that structure the interactions among agents are of a diverse
nature. These are developed based on the various activities conducted by entrepreneurs
and firms in their processes of recruitment and procurement of inputs, human resources
management, production, innovation and technology management, product
development, financial management, and marketing and sales strategy. Also influential
are the possible strategies for growth and territorial expansion of investments (within or
outside the territory).
In general, the market itself sets a ‘virtual’ kind of network relationships among firms in
their pursuit of needs/opportunities for personnel, intermediate inputs, production
equipment and technology, and sales niches, which are provided by the different types
of markets (labor, goods and services…) (Coviello and Munro, 1995; Slotte-Kock and
Coviello, 2010). But the general market also establishes the relationship of rivalry and
competition between them. In sum, these are network relations which provide
information for cooperation or competition as appropriate (Gulati, 1999, Meyer et al.,
2004).
Contractual relationships on the labor market are structured between individuals and
firms. They are based on search processes, where firms in a context of imperfect
information choose the best candidates and individuals offer their work to firms that
provide the most attractive terms. In general, these relationships are normative since
they are based on labor legislation.
In the modeling process, the properties of the protocols that govern the interaction
between individuals and companies are based on relations of production and
consumption, employment and lending, adopting a microeconomic perspective (see
figure 2). In this sense, the price of competitors’ products in relation to themselves
constitutes one of the main signals received by the agents. They act taking the suitable
decisions for the acquisition of inputs, recruitment of factors, production and sales.
These relations take place in markets.
Figure 2. Agents and environment incorporated in MOSIPS
Source: Own elaboration based on Albino et.al (2007).
Individuals face the same interaction protocols and information flows, but applied to
their decisions. They obtain most of the information from firms which they are linked,
but also from the aggregate behavior (e.g. the unemployment rate, GDP growth, price
index). Individuals’ decisions are also conditioned by the performance of other agents
who are linked with. For example, a potential entrepreneur will decide to create his own
enterprise with a higher probability if both their acquaintances and the information
about the general performance of the economy are promising for the success.
2.3. Activities
Individuals are born into households where they grow, consume, pursue an education,
and, eventually, die. Households may change their location and increase or decrease the
number of their members. Upon reaching working age, individuals choose in each
period whether to join the workforce and, if so, become employees or entrepreneurs. Job
seekers offer their work to firms in their environment recruiting workers, which take the
decision regarding who to hire. Individuals who choose to become entrepreneurs create
firms and choose the location.
Firms produce and sell their products on the market. They also choose the cheapest and
most reliable suppliers. Additionally, they change in terms of size through internal
growth or by acquiring other firms, provided they have adequate funding. They can
apply for funding from the financial system, based on their repayment capacity and on
the financial market conditions. Further, firms decide the level of their commitment to
innovation, both in terms of processes and products. They modify their workforce by
hiring or laying off workers according to their labor skills or to production needs. Firms
disappear if they go bankrupt or if the businessperson so chooses.
The public sector incorporates its rules and policies by modifying the attributes and
behaviors of agents. Banks procure funds from agents with a funding capacity, and they
provide funds to those who require them. The financial resources available for firms and
households can differ from the level of savings of agents due to the financial flows with
other countries and the circumstances of the financial system. The local environment
includes aspects that affect the agents from a territorial perspective such as closeness
from infrastructures, the existence of firm clusters, or congestion problems.
3. Public Policies
3.1. Small Business Act and MOSIPS policy domains
The MOSIPS project takes the Small Business Act (SBA) (European Commission,
2008) as the core area for policy investigation, analysis and modeling. The SBA forms
the ‘enabling framework’ of the EU for improved SME performance and policy quality.
It is the EU flagship SME policy initiative comprising ten principles that should guide
the design and implementation of policies in the EU and its Member States. These
principles are central to the conceptual, theoretical and empirical scope of MOSIPS. The
SBA “aims to improve the overall policy approach to entrepreneurship, to irreversibly
anchor the “Think Small First” principle in policymaking from regulation to public
service, and to promote SMEs’ growth by helping them tackle the remaining problems
which hamper their development”. These principles are the following:
1. Create an environment in which entrepreneurs and family businesses can thrive
and entrepreneurship is rewarded
2. Ensure that honest entrepreneurs who have faced bankruptcy quickly get a
second chance
3. Design rules according to the “Think Small First” principle
4. Make public administrations responsive to SMEs’ needs
5. Adapt public policy tools to SME needs: facilitate SMEs’ participation in public
procurement and better use State Aid possibilities for SMEs
6. Facilitate SMEs’ access to finance and develop a legal and business environment
supportive to timely payments in commercial transactions
7. Help SMEs to benefit more from the opportunities offered by the Single Market
8. Promote the upgrading of skills in SMEs and all forms of innovation
9. Enable SMEs to turn environmental challenges into opportunities
10. Encourage and support SMEs to benefit from the growth of markets
We identify the following ten policy domains according to the rationale of MOSIPS
model (see figure 3). The purpose of this categorization is to include all the possible
different areas in which policies on SMEs can be carried out. Moreover, this can be
translated to the principles presented by the Small Business Act developed by the
European Commission, strengthening the project by demonstrating a strong ‘rationale’
behind the categorization presented.
This approaching make possible to provide aggregate results easily understandable for
policy makers, public administrators and citizens. In addition, every policy could be
integrated in the system only in a specific manner avoiding errors in the inclusion. It is
possible to identify a lot of public policies that directly or indirectly affect SMEs. This
fact makes it intractable to develop a model that allows specifying individually each
policy. To solve this problem we classify public policies into several groups that cover
all the possibilities, but not so many that can lead to a policy to be included in several
groups.
Figure 3. Policy domains of MOSIPS model implementation-oriented
Source: Own elaboration
The following examples show how policies can be integrated in MOSIPS model. First, a
decrease in the percentage charged in the value added tax (VAT) affects SMEs through
a change in the macroeconomic environment, because the policy incentive might affect
consumption behavior, and hence greater potential sales from local SMEs. Also, it will
affect firms and individuals trough a change in their expectations. That includes an
increase in the demand that will provoke an increase in the supply, the employment and
other macro variables. In conclusion, it affects SMEs by a changing in data. Thus, a
change in VAT percentage charged does not affect behavior of enterprises nor
consumers. They act in the same way as before, but with a different data for taking their
decisions. In the same way, the cost functions of enterprises are not shifted, this change
is produced around the cost curve.
Second, we could imagine two kinds of grants that affect innovation: a grant intended to
increase the potential of SMEs to acquire high-tech machinery and another one that is
granted for training their employees in a R&D course. In the first case, it affects the cost
function by reducing the costs of producing a certain amount of product. Instead, the
second measure affects the behavior of SMEs, making them more able to produce and
adopt innovations, but in the next period they will produce with the same cost function
and data (employees, machinery…) as if the policy had not taken place. However, in the
following periods, it would behave in a more pro-innovative way, adapting and
developing new procedures that will reduce costs.
Thus, the effect of public policies on SMEs, in accordance with the 10 principles of the
European SBA, is included in the previous presented categorization. Those principles
are connected to MOSIPS policy fields as follows:
Conversion of the SBA principles into functional domains
However, two of the principles are not connected in one-to-one. First, single market is a
principle more related with rules of standardization. Then it is linked somehow with
innovation, but not only. In an indirect way also will be part of funding (financial
support to standardization) and an administrative effort on this issue should be consider
the correct implementation. Secondly, internationalization is included in two different
domains: Inter-firm relations and Internal managing. The first one catches the
importance of business relationships (e.g. temporary joint ventures, Chambers of
Commerce membership). The second one takes into account the relevance of the
decision of internationalize the enterprise undertaken by the entrepreneur.
3.2. Design of policy in MOSIPS model
Then, to illustrate how these ten policy domains are included in the model, it is
presenting one of the modules of MOSIPS model: Firm demography (see figure 1 to
view the position in the model). After adults decide to become entrepreneurs, or they
continue being entrepreneurs, they are asked about if they have any firms, if they want
to open or close anyone and if the answer affirmatively, the system determines if the
firms is finally opened, that is, if licenses and funding are obtained.
Figure 4. Firm demography
FIRM DEMOGRAPHY
Entrepreneur
Does the
entrepreneur
have any
firms?
NO
Entrepreneur (Firm owner)
YES
Evaluates if the
businessman wants
another firm
Does the
businessman
wants to create
other firm?
YES
Determines the
characteristics of the
new company
NO
Evaluates the
perfomance of their
firms
Does it get
funding?
NO
NO
End
Does the
individual
want to close
any firms?
YES
YES
He/she closes the
firm(s)
Does it get
licenses?
YES
The entrepreneur
creates the firm
YES
Does the
individual have
any remaining
firms?
Entrepreneur -> Firm
End
NO
Stops being an
businessman
End
Source: Own elaboration
The main operational assumptions for firm demography are:




NO
Entrepreneurs who have no firm trying to create it.
Entrepreneurs who have at least one firm are businessmen.
Businessmen may own more of a business.
Entrepreneurs determine the desirable characteristics of their new firms before
creating them.
 If they have funding so they create them.
 To create a firm is mainly a legal issue. It’s possible a firm without
establishments.
 If the performance of firms is too low then they close. The establishments of the
closed firms become free.
It is assumed that in each period only one firm can be created (with a single
establishment) by each businessperson. A firm exists from the moment that it has been
legally incorporated even if it lacks physical space for production. An establishment is
needed for production. For those firms that do not have any physical space for
production (street vendors, professionals without a registered office, etc.), the owner’s
fiscal address is allocated as the physical space.
Figure 4 also shows how the different policy domains are incorporated to each part of
the module:
The macroeconomic environment affects the decision of start a new firm
through expectations about its success, and if the entrepreneur wants to create a
firm, have effects on the determination of its characteristics (sector, size,
location,…).
Labor market has effects on the decision of open a new business. A strong labor
regulation could discourage the entrepreneur, while if there are public aids to
contracting new employees or to becoming self-employed the entrepreneur will
decide to open a new firm.
Funding is present in the determination of funding possibilities of new potential
firms and in the evaluation of the performance of current firms.
Regulation and red tape has three different effects on firm demography:
o It is the main issue in the chance of getting licenses to open a new firm.
o It affects the characteristics of new firms, as public administration
discourages or even forbids some sectors and locations.
o Red tape and regulation also affects the current undergoing of firms,
having effects on their profitability in a plenty of ways.
4. Main conclusions
This work discuss the policy design in an agent-based model, in a wider project called
MOSIPS, which consists on model with a number of modules and it is aimed at the
simulation, and evaluation of policies for small and medium enterprises in a local or
regional environment.
By contrast to traditional models, which give a partial view of the economic reality, and
to the DSGE models, broadly used to evaluate and predict the effects of policies, which,
even though they offer a global sphere of analysis, hardly provide for the heterogeneity
of agents, the ABM model approach adopted by this study focuses on the actions and
interactions of these agents in their local environment.
The model includes ten policy domains, based on the ten principles of the SBA, but
different to cover any possible policy taken by the public administration. Every domain
can affect the behavior of the agents, the data which they face or the cost function of
each establishment. The functioning of policy design is showed with one of the modules
of MOSIPS model, firm demography.
References
Albino, V., Carbonara, N. and Giannoccaro, I. (2007). “Supply chain cooperation in industrial
districts: A simulation analysis”. European Journal of Operational Research, 177, 261-280.
Arrow, K.J. and Debreu, G. (1954). “The existence of an equilibrium for a competitive economy”.
Econometrica, 22, 265-290.
Colander. D, Howwit, P., Kirman, A., Leijonhufvud, A. and Mehrling, P. (2008). “Beyond DSGE
Models: toward an Empirically Based Macroeconomics”, American Economic Review: Papers
and Proceedings, V. 98, 2: 236–240.
Committee on Science and Technology. (2010). “Building a Science of Economics for the Real
World”, Serial No. 111–106, available at:
http://www.gpo.gov/fdsys/pkg/CHRG111hhrg57604/pdf/CHRG 111hhrg57604.pdf
Coviello, N. and Munro, H. (1995), "Growing the entrepreneurial firm: networking for international
market development", European Journal of Marketing, Vol. 29 Iss. 7, pp. 49 – 61.
Dawid, H., Gemkow, S., Harting, P., van der Hoog, S., and Neugart, M. (2011). “The Eurace@Unibi
Model: An Agent-Based Macroeconomic Model for Economic Policy Analysis”. Available from
www.wiwi.uni-bielefeld.de/vpl1/projects/eurace/eurace-unibi.html
Dosi, G., Fagiolo, G., and Roventini, A. (2008). “Schumpeter Meeting Keynes: A Policy-Friendly
Model of Endogenous Growth and Business Cycles”. Laboratory of Economics and
Management, SantAnna School of Advanced Studies, LEM Working Paper Series, 21.
European Commission. 2008. “Think small first: A Small Business Act for Europe”. DG Enterprise,
Brussels, Belgium.
Farmer, J. and Foley, D (2009). “The economy needs agent-based modelling”. Nature, 460, 685-686.
Gintis, H. (2006). “The emergence of a price system from decentralized bilateral exchange”. Journal
of Theoretical Economics, 6, 1302-1322.
Gulati, R. (1999), "Network location and learning: The influence of network resources and firm
capabilities", Strategic Management Journal, Vol. 20, Iss.5, pp. 397-420.
Kirman, A. (2010). "The Economic Crisis is a Crisis for Economic Theory". CESifo Economic
Studies, Oxford University Press, V. 56, 4: 498-535.
Mandel, A., Fürst, S., Lass, W., Meissner, F., and Jaeger, C. C. (2009). “Lagom generiC: an agentbased model of growing economies”. ECF Working Paper, 1. Available from
www.europeanclimate-forum.net/index.php?id=ecfworkingpapers
Meyer, M., Aderhold, J. and Duschek, S. (2004), “Organizing social complexity in production
networks”, Journal of Academy of Business and Economics, Vol. 3, Iss. 1, 1 January.
Slotte-Kock, S. and Coviello, N. (2010). “Entrepreneurship Research on Network Processes: A
Review and Ways Forward”, Entrepreneurship Theory and Practice, Vol. 34, Iss. 1, pp. 31–57.
Tesfatsion, L. and Judd, K. (2006). “Agent-Based Computational Economics”. in Handbook of
Computational Economics. Elsevier, North-Holland.
Wolf, S., Bouchaud, J-P., Cecconi, F., Cincotti, S., Dawid, D., Gintis, H., Hoog, S. Jaeger, C.C.,
Kovalevsky, D.V., Mandel, A., Paroussos, L. (2011). “Describing economic agent-based models,
Dahlem ABM documentation guidelines. Proceedings of the 100th Dahlem Conference, New
Approaches in Economics after the Financial Crisis”, 2010 Aug 28-31.
Wolf, S., Fürst, S., Mandel, A., Lass, W., Lincke, D., Pablo-Marti, F., Jaeger, C. (2011), “Lagom
RegiO – a multi-agent model of several economic regions”. (Published forthcoming).