BookPDF Available

Abstract

Complex Systems can be defined in a broad manner and embrace concepts from different fields of science, from physics to biology, to computing and social sci-ences. Mainly, the definition includes nonlinear dynamical systems that contain large number of interactions among the parts. These systems learn, evolve, and adapt, generating emergent non-deterministic behavior. Public policies are to be applied upon a vast range of issues that involve the public, the broad community of citizens and communities, firms and institutions. Public policies are also to be employed on a number of sectorial issues which are intertwined, asynchronous, and spatially superposed. This coupled understanding of complex systems and public policies suggests that most objects of public policies – be them of economic or urban nature, be them of environmental or political consequences – can be viewed as complex systems. Thus, if public policies’ objects can be seen as complex systems, their understanding may benefit from the use of associated methodologies, such as network analysis, agent-based modeling, numerical simulation, game theory, pattern formation and many others within the realm of complex systems. These methodologies have been applied to different aspects of science, but less frequently to public policy analysis. We hypothesize that the use of these concepts and methodologies together improves the way policies of complex objects are viewed, adjusted, and operated upon from a public point of view.
Editors
Bernardo Alves Furtado
Patrícia A. M. Sakowski
Marina H. Tóvolli
MODELING COMPLEX SYSTEMS
FOR PUBLIC POLICIES
MODELING COMPLEX SYSTEMS
FOR PUBLIC POLICIES
Editors
Bernardo Alves Furtado
Patrícia A. M. Sakowski
Marina H. Tóvolli
Federal Government of Brazil
Secretariat of Strategic Affairs of the
Presidency of the Republic
Minister Roberto Mangabeira Unger
A public foundation affiliated to the Secretariat of
Strategic Affairs of the Presidency of the Republic,
Ipea provides technical and institutional support to
government actions – enabling the formulation of
numerous public policies and programs for Brazilian
development – and makes research and studies
conducted by its staff available to society.
President
Jessé José Freire de Souza
Director of Institutional Development
Alexandre dos Santos Cunha
Director of Studies and Policies of the State,
Institutions and Democracy
Daniel Ricardo de Castro Cerqueira
Director of Macroeconomic Studies
and Policies
Cláudio Hamilton Matos dos Santos
Director of Regional, Urban and Environmental
Studies and Policies
Marco Aurélio Costa
Director of Sectoral Studies and Policies,
Innovation, Regulation and Infrastructure
Fernanda De Negri
Director of Social Studies and Policies
André Bojikian Calixtre
Director of International Studies, Political
and Economic Relations
Brand Arenari
Chief of Staff
José Eduardo Elias Romão
Chief Press and Communications Officer
João Cláudio Garcia Rodrigues Lima
URL: http://www.ipea.gov.br
Ombudsman: http://www.ipea.gov.br/ouvidoria
MODELING COMPLEX SYSTEMS
FOR PUBLIC POLICIES
Editors
Bernardo Alves Furtado
Patrícia A. M. Sakowski
Marina H. Tóvolli
Brasília, 2015
© Institute for Applied Economic Research – ipea 2015
The opinions expressed in this publication are of exclusive responsibility of the authors, not necessarily
expressing the official views of the Institute for Applied Economic Research and the Secretariat of
Strategic Affairs of the Presidency.
Reproduction of this text and the data it contains is allowed as long as the source is cited. Reproductions
for commercial purposes are prohibited.
Cover photo
National Aeronautics and Space Administration (Nasa)
Available at: https://www.flickr.com/photos/nasa2explore/8380847362
Modeling complex systems for public policies / editors: Bernardo
Alves Furtado, Patrícia A. M. Sakowski, Marina H. Tóvolli.
Brasília : IPEA, 2015.
396 p. : il., gráfs. color.
Includes bibliographical references.
ISBN: 978-85-7811-249-3
1. Public policies. 2. Development Policy. 3. Complex Systems.
4. Network Analysis. 5. Interaction Analysis. 6. Planning Methods.
I. Furtado, Bernardo Alves. II. Sakowski, Patrícia A. M. III.
Tóvolli, Marina H. IV. Institute for Applied Economic Research.
CDD 003
CONTENTS
FOREWORD ...............................................................................................7
PREFACE.....................................................................................................9
PART I COMPLEXITY: THEORY, METHODS AND MODELING
CHAPTER 1
A COMPLEXITY APPROACH FOR PUBLIC POLICIES ...........................................17
Bernardo Alves Furtado
Patrícia Alessandra Morita Sakowski
Marina Haddad Tóvolli
CHAPTER 2
COMPLEX SYSTEMS: CONCEPTS, LITERATURE, POSSIBILITIES AND LIMITATIONS ..37
William Rand
CHAPTER 3
METHODS AND METHODOLOGIES OF COMPLEX SYSTEMS ...............................55
Miguel Angel Fuentes
CHAPTER 4
SIMULATION MODELS FOR PUBLIC POLICY.....................................................73
James E. Gentile
Chris Glazner
Matthew Koehler
CHAPTER 5
OPERATIONALIZING COMPLEX SYSTEMS ........................................................ 85
Jaime Simão Sichman
PART II OBJECTS OF PUBLIC POLICY AND THE COMPLEX SYSTEMS
CHAPTER 6
UNDERSTANDING THE ENVIRONMENT AS A COMPLEX, DYNAMIC
NATURAL-SOCIAL SYSTEM: OPPORTUNITIES AND CHALLENGES IN
PUBLIC POLICIES FOR PROMOTING GLOBAL SUSTAINABILITY ......................... 127
Masaru Yarime
Ali Kharrazi
CHAPTER 7
THE COMPLEX NATURE OF SOCIAL SYSTEMS ................................................ 141
Claudio J. Tessone
CHAPTER 8
THE ECONOMY AS A COMPLEX OBJECT .......................................................169
Orlando Gomes
CHAPTER 9
MODELING THE ECONOMY AS A COMPLEX SYSTEM .....................................191
Herbert Dawid
CHAPTER 10
CITIES AS COMPLEX SYSTEMS .....................................................................217
Luís M. A. Bettencourt
PART III COMPLEX SYSTEMS APPLICATIONS TO OBJECTS
OF PUBLIC POLICIES
CHAPTER 11
COMPLEXITY THEORY IN APPLIED POLICY WORLDWIDE .................................239
Yaneer Bar-Yam
CHAPTER 12
COMPLEX SYSTEMS MODELLING IN BRAZILIAN PUBLIC POLICIES ...................261
Bernardo Mueller
CHAPTER 13
COMPLEXITY METHODS APPLIED TO TRANSPORT PLANNING .........................279
Dick Ettema
CHAPTER 14
EDUCATION AS A COMPLEX SYSTEM: IMPLICATIONS FOR
EDUCATIONAL RESEARCH AND POLICY .......................................................301
Michael J. Jacobson
CHAPTER 15
COMPLEX APPROACHES FOR EDUCATION IN BRAZIL ................................... 315
Patrícia A. Morita Sakowski
Marina H. Tóvolli
CHAPTER 16
OVERCOMING CHAOS: LEGISLATURES AS COMPLEX ADAPTIVE SYSTEMS ........337
Acir Almeida
CHAPTER 17
THE TERRITORY AS A COMPLEX SOCIAL SYSTEM ..........................................363
Marcos Aurélio Santos da Silva
FOREWORD
Policy is the means and end of the Institute for Applied Economic Research. Policy
evaluation, policy design and monitoring along with advising the State using the
best scientic knowledge are at the core of the Institute. us, tools that enable
policy-makers and academia alike to foster a deeper understanding of policy
mechanisms and their intertwined, asynchronous, and spatially-bound eects are
at the forefront of our interests.
Complexity is a relatively new approach to science, which has integrated
knowledge from dierent elds, trying to understand collective behavior in living
systems and complex phenomena such as emergence. It has brought important
insights for science, but little has been done trying to explore the policy aspects
of this new approach both in Brazil and worldwide.
is book tries to help building this bridge between complexity and public
policies, by bringing together an international group of prominent researchers,
stemming from the very Santa Fe Institute, University of Maryland, University of
Tokyo, University of Sidney, ETH Zurich, Bielefeld University, Utrecht University,
New England Complex Systems Institute, Polytechnic Institute of Lisbon, MI-
TRE Corporation, University of Brasilia, University of São Paulo and EMBRAPA
and Ipea researchers. By introducing the major concepts, methods and state-of-
the-art research in the area, the book is intended to be a seminal contribution to
the application of the complexity approach to public policies, and a gateway for
the world of complexity.
As an Institute whose middle name is policy, I think it is high time for us to
look more and more at the policy aspects of this new approach and to explore the
insights and applications they can bring into policy making and analysis. You are
invited to join us in this journey.
Jessé Souza
President of the Institute for Applied Economic Research
PREFACE
Scott E. Page
1
In Norman Juster’s classic e Phantom Toolbooth, the protagonist Milo and his
companion, a large dog named Tock, cannot gure out how to get their wagon to
move forward. A Duke arrives and tells them that if the wagon to go, they must sit
quietly, that it (the wagon) goes without saying. e same might be thought about
the relevance of complexity theory to public policy – that it too goes without saying.
Given the complexity of the political and bureaucratic processes that generate
policies and the complexity of the systems within which most policies are applied,
it would seem that complexitys relevance should go without saying. Yet, that’s not
the case. e patchwork of models, concepts, and ideas that comprise the eld
of complexity studies rarely enter into policy discussions and when they do, they
primarily engage at the fringes.
2
erefore, unlike Tock and Milo, complexity scholars cannot sit quietly.
If complexity scholars want their ideas to advance and improve public policy,
they must speak clearly and loudly. In this volume, many leading scholars
choose to do just that. eir impact should be substantial.
What follows includes contributions from many of the leading scholars in
the eld of complex social systems. It should then come as no surprise that the
volume achieves multiple, ambitious goals: it introduces the concepts and tools of
complex systems, it demonstrates complexity theory’s relevance to public policy,
it contrasts the complexity approach to public policy to traditional methods, and,
nally, it presents case studies and examples that demonstrate proof of concept by
focusing on specic policy domains in Brazil and elsewhere.
So what are complex systems? Complex systems consist of diverse, adaptive
actors who interact with their neighbors and over networks. ese interactions
produce both additive outcomes – aggregate oil consumption or the average
price of #2 red wheat – as well as emergent phenomena such as traveling waves in
trac patterns, stock market crashes, and even Spanish culture. ese aggregate
1. University of Michigan, Santa Fe Institute.
2. Climate change models, which might be seen as a counterexample, can be seen as a type of complex system
model, but they tend to be mash ups of standard economic models with geophysical models, lacking many of the core
components of complex systems models.
Modeling Complex Systems for Public Policies10
|
phenomena become part of the world and induce adaptations at the micro level.
ese in turn create new macro level phenomena.
e resulting dynamics can take many forms. ey can converge to equilibria.
ey can produce cycles or simple patterns, such as the near linear trend in the
global output of oil over the past hundred years. ey can produce complex time
series, as is the case with oil prices. Finally, they can produce data that appears
random, a property nearly satised by detrended stock prices. In brief, complex
systems can produce anything. And because they can, they can help us understand
almost anything.
e pitch then, the two-minute elevator speech for why we need to bring
complexity research to policy domains, partly relies on this resonance: if we want
to understand a policy domain produces complex outcomes such as constantly
changing stock and housing prices, trac patterns, or divergent paths of school
success, then we should use models capable of producing similar types of complexity.
Cast in a comparative frame, this logic challenges the predominant equilibrium
paradigm: why would someone base policy in a complex domain on a model that
assumes equilibrium? e argument extends what William Rand in this volume
relates as the “standing under the streetlight” criticism of neoclassical economic
models: they shine light, but not where we should be looking. Complex systems
models represent ashlights to guide us to new locations in modeling space.
e pitch also relies on the interconnectedness of policy actions. Education
policy, environmental policy, zoning policy, infrastructure decisions, and energy
policies all bump into one another. Put metaphorically, policies do not operate in
silos. Put mathematically, nonzero cross partials abound. Eorts to reduce wealth
inequality by extending home loans induce residential sorting which inuences
school quality, trac density, crime rates, and so on. As described by Furtado
et al. in this volume, complex systems’ approaches, “enable public policies to be
considered comprehensively and simulated explicitly in all their multiplicity of
sectors and scales, of cause and eect.” An observation echoed in and elaborated
on in the excellent chapter by Claudio Tessone.
Advocacy notwithstanding, the volume takes a measured stance. No one
denies that standard approaches to evaluating policies – equilibrium models and,
when possible, natural experiments – are useful and often powerful tools. Complex
systems do not represent a silver bullet, but another arrow in the policy makers
quiver. More accurately, all of these tools put together can be thought of as multiple
imperfect arrows that provide insight into what is likely to happen, what could
happen, and how what happens might spill into other domains.
Preface
|
11
Consider, for example, a gold standard natural experiment that reveals a
policy to be a success. Complex systems models might suggest that the policy could
create multiple types of outcomes. e success of the policy might well have been
good luck – like picking up a die and rolling a six. Rather than roll out the policy
nationwide, a prudent policy maker might run a few more experiments to see if
in fact, the outcome was a lucky roll.
Alternatively, complex systems models might show that the policy,
though successful, produces long-term negative feedbacks. An analysis of
these feedbacks, as William Rand demonstrates in his chapter, provides us
with a deeper understanding of the full effects of a policy.
Notice that these feedbacks produce a type of nonlinear eect which along
with heterogeneity can make a model may intractable using game theoretic or
mathematical optimization techniques. For decades, these tractability constraints
limited the dimensionality and realism of models. Policy makers had to rely on
models they could solve. ose models were not complex.
Owing to increases in computing power and the introduction of a new meth-
odology, agent based models (ABM), tractability has become less of a constraint.
Any model that can be coded can be explored. Yet, we should be skeptical, dubious
(dare I say dismissive) when someone claims “I have a simulation that shows (ll in
the blank).” A vast continent of poorly constructed, spaghetti coded, invalidated,
unveriable, non-calibrated models surrounds a much smaller region of useful
models. As Gentile, Glazner and Koehler’s chapter makes clear, ABMs have enor-
mous potential as a tool for policy comparisons, but ABMs must be constructed
by people well versed in the methodology.
e resulting models can include agents who use sophisticated learning
algorithms (see Jaime Sichmans chapter) or they can rely on relatively simple
rules. No one ABM model will tell us with one hundred percent certainty
the full eects of policy, but many models with multiple levels of granularity
and domains of interaction will give us a better understanding of the set of
the possible and ensure more robust policies. And, isnt avoiding surprise an
important aim of policy makers?
Economic policy is one domain where surprise events can have dramatic
consequences. More than two decades ago, several leading economists advanced the
notion that the economy would be more accurately thought of as a complex adap-
tive than as an equilibrium system. ough equilibrium models still predominate,
those models include networks, learning, and heterogeneous agents who do not
always make optimal decisions. Furthermore, state of the art equilibrium monetary
models (dynamic stochastic general equilibrium models – DSGE) spend almost
all of their time out of equilibrium.
Modeling Complex Systems for Public Policies12
|
At their core though, the DSGE models rely on equilibria to characterize the
dynamics. e economy is always headed toward an equilibrium. In other words,
the modeler and the actors in the model know where they economy is headed.
In contrast, ABM models of the economy make assumptions and then, echoing
ideas from Orlando Gomes’ and Rand’s chapters, the economy emerges from the
bottom up.
Two chapters in this volume clarify this complexity approach to modeling the
economy. Gomes both makes the intellectual case for a complexity approach and
presents a (relatively) simple complex systems model. In contrast, Herbert Dawid
shows how one can embrace complexity in full. He provides an introduction to
the elaborate Eurace@Unibi model of the economy. is intendedly realistic model
includes spatially situated consumers with budgets and rms with suppliers and
inventories has produced meaningful policy insights, among them: policy can be
relevant away from equilibrium, individual responses may dier from aggregate
eects,
3
outcomes can be path dependent, and institutional details can matter.
ese are not empty claims. Evidence suggests that in some domains ABMs can
make better predictions than standard models.
But ABMs can also make worse predictions. And while its tempting to stage
a horse race between complexity models and equilibrium models of the economy,
doing so misses the earlier point about multiple arrows in the policy makers quiver.
Economic models consider the economy. Complexity models have the potential
to see the economy within a broader system in which people engage in social
movements, confront political regime changes, respond to threats of epidemics,
natural disasters, and climactic change. All of this can be seen as operating within
one system. We can try to peel o the economy and study it in isolation, just as we
could study only the circulatory system, the nervous system, the immune system,
or the digestive system, but if we do, we miss the real show.
e real show occurs at multiple scales: from family, to city, to nation, to
world. Cities oer one useful scale as, Luis Bettencourt shows in his chapter sum-
marizing years of scholarship. Cities, as many have noted, are the engines of the
economy. As Paul Krugman once quipped, almost anyone can identify cities from
an airplane on a clear night, but almost no one could draw country boundaries.
Cities, therefore, might be a important level of activity to analyze. Bettencourt
shows this to be true, highlighting provocative ndings of scaling laws – produc-
tivity scales superlinearly and infrastructure scales sublinearly – and juxtaposing
implications from the complexity paradigm with historical views of the city that
3. This would be the case when an outcome is emergent as opposed to additive. When feedbacks and nonlinearities are
present, agent heterogeneity can produce aggregate results that differ from what would be produced by an economy
composed of identical agents.
Preface
|
13
take an engineering approach. He embraces the roles of information and learning
in thinking through policy eects and identies criteria when local adaptation
should outperform top down implementation.
Yaneer Bar-Yam takes on a multiple scales – from single markets to the world
writ large. Deregulation of a commodity (a national scale economic policy) results
in global scale price changes. ese in turn can depress rm scale revenues, which
could under certain conditions, result in regional scale uprisings. Quoting Bar-Yam:
“One nations energy subsidies can cause global food prices to spike, setting o
political unrest halfway around the world.” e general phenomena to which he
speaks is captured in the famous lyrics of Disneys Richard and Robert Sherman,
it is in fact “a small world after all.
Within that small world, policy makers must make choices. Inevitably, there
will be successes and failures. e raison d’être of the volume is to increase the former
and decrease the latter. Perhaps the single most powerful statement in the book
appears in the chapter by Bernardo Mueller describing two case studies. Writing
of the Brazilian policy apparatus, he writes: “I have not found any example of an
explicit use of complex thinking in any policy in this country.
In Brazil, the complexity wagon does not go without saying despite the fact
that legislative institutions within Brazil appear to be quite complex, as proven by
Acir Almeida, who shows the relative contributions of a complexity perspective on
legislative activity. Using models from political science and complex systems, he
shows how ideas from complexity theory add to our understanding of emergent
patterns of law-making in the Brazilian Congress.
at law making occurs within a Brazilian system in which, according to
Mueller, the Executive wields enormous power. Of course, a structure of checks
and balances, reigns in that power, but what’s most relevant is how the policies are
formulated. Mueller nds fault with what he calls a reductionist, i.e. non-complex,
approach. e policy domains in question: land use, public health, the environ-
ment, and transportation, these are all complex domains. Policies are developed
and evaluated as if they were not. In his opinion, that’s a mistake.
e subsequent chapter by Dick Ettema unpacks this line of criticism in even
greater detail. He describes the engineering approach to transportation policy with
its focus on meeting individual level criteria of success or utility such as avoidance
of congestion and pollution. ese models level out the eects on housing markets,
equity, and social exclusion.
Most people accept that transportation systems and stock markets are complex.
People experience congestion and trac jams. ey watch stock prices rise for
weeks with only small changes and then drop five to ten percent in a few hours.
Modeling Complex Systems for Public Policies14
|
Other systems, such as educational systems, are less obviously complex.
Time
unfolds more slowly. Phenomenological changes are more abstract and less
easily measured. Yet, as Michael Jacobson demonstrates, schools can be usefully
seen as complex systems.
Policies can try to improve them by pulling levers – reducing class size or
increasing teacher quality. Policies can also try to improve mechanisms. Both types
of policies have a linear orientation and are presented as such leading to claims
that an decrease of X percent in pupil teacher ratios will lead to an increase of Y
percent in student test scores. is is yet another example of a misplaced focus on
a single partial derivative within a complex system, a point reinforced by Sakowski
and Tóvolli in their analysis of Brazilian education policy.
Policies have interactions in other domains. In constructing an eective policy,
one cannot proceed dimension by dimension. Step 1: minimize average commute
time. Step 2: maximize student test scores. Step 3: reduce ination. Step 4: produce
sustainable forest management plan. Policies occur within systems and those systems
interact. In particular, the social and the physical interact, as made abundantly clear
in Marcos Aurélio Santos da Silvas chapter on socioterritorial systems.
In sum, whether we focus our lens on the forests or students of Brazil or the
world writ large, we cannot help but see the inherent complexity. We see diverse,
purposeful connecting people constructing lives, interacting within institutions,
and responding to rules constraints, and incentives created by policies. ese activi-
ties occur within complex systems and when the activities aggregate they produce
feedbacks and create emergent patterns and functionalities. By denition, complex
systems are dicult to describe, explain, and predict, so we cannot expect ideal
policies. But we can hope to improve, to do better. Having more tools, especially
the evolving and maturing tools of complexity science, can only make us better.
Complex scholars can move the needle. But they can no longer sit quietly.
PART I
Complexity: theory, methods
and modeling
CHAPTER 1
A COMPLEXITY APPROACH FOR PUBLIC POLICIES
Bernardo Alves Furtado
1
Patrícia Alessandra Morita Sakowski
2
Marina Haddad Tóvolli
3
1 INTRODUCTION
Complex Systems can be dened in a broad manner and embrace concepts from
dierent elds of science, from physics to biology, to computing and social sci-
ences. Mainly, the denition includes nonlinear dynamical systems that contain
large number of interactions among the parts. ese systems learn, evolve, and
adapt, generating emergent non-deterministic behavior.
4
Public policies are to be
applied upon a vast range of issues that involve the public, the broad community
of citizens and communities, rms and institutions. Public policies are also to be
employed on a number of sectorial issues which are intertwined, asynchronous, and
spatially superposed. is coupled understanding of complex systems and public
policies suggests that most objects of public policies – be them of economic or
urban nature, be them of environmental or political consequences – can be viewed
as complex systems. us, if public policies’ objects can be seen as complex systems,
their understanding may benet from the use of associated methodologies, such
as network analysis, agent-based modeling, numerical simulation, game theory,
pattern formation and many others within the realm of complex systems. ese
methodologies have been applied to dierent aspects of science, but less frequently
to public policy analysis.
5
We hypothesize that the use of these concepts and
methodologies together improves the way policies of complex objects are viewed,
adjusted, and operated upon from a public point of view.
Given these broad denitions of complex systems and public policies, this
chapter further describes the concepts, methodologies and computing implementa-
tion of complex systems. en, it demonstrates the adherence of those concepts and
1. Researcher at Diset/Ipea, Productivity researcher at CNPq.
2. Reseacher and Chief of the Planning and Institutional Articulation Advisory Board (Aspla/Ipea).
3. Research Assistant at Dirur/Ipea.
4. A didactically complete discussion of Complexity is available at Mitchell (2011). The initial concepts that compose
the complexity sciences can be found in Furtado and Sakowski (2014).
5. Initially, one could look at Colander and Kupers (2014) who provide a review focused on economics. Edmonds and
Meyer (2013) give a detailed background. An earlier report can be found at OECD (2009).
Modeling Complex Systems for Public Policies18
|
methodologies to social policies, economic, urban and environment analysis; and
highlights some applications on transport planning, in education, on the study of
the legislative and on territorial analysis. After having described the concepts and
methods and why public policies’ objects can be easily viewed as complex systems,
the chapter lists the advantages of using complex systems’ approach specically to
public policies. us, this chapter introduces and summarizes the contents of the
book (and project) of the same name.
In short, it is the objective of this chapter to dene complex systems and
its more prominent attributes; list the more common methodologies associated
with complex systems; present a varied scope of applications of complex systems
modeling to public policies and discuss yet briey the advantages of applying these
approaches to a public policy context.
2 DEFINITIONS OF COMPLEX SYSTEMS AND PUBLIC POLICIES
Complex systems denition is usually attached to a specic context; however, it
usually incorporates the following set of features.
Firstly, the idea of interaction among parts from and across scales, space and
time is relevant. ese interactions, in turn, lead to a system that is not reducible;
a system that cannot be described by the attributes of the parts alone. Basically, to
quote Andersons classic "More is dierent" paper (1972, p. 395, our emphasis):
“In this case we can see how the whole becomes not only more than but very dierent
from the sum of its parts.
Secondly, the interaction among parts can lead to self-organization of the
system without the need of central control. is implies that local interactions can
generate bottom-up emergent behavior. is powerful concept can be illustrated
for the novice reader with the example of a bird ocking. No actual bird controls
the direction and position of all birds in a given ock ight. Each one bird only
observes those near it and synchronizes with their immediate neighbors. As a result,
coordinated ight emerges.
A third attribute to highlight is that complex systems can experiment feedback.
In complex systems, interactions have eects in time: actions in a given moment
reect on possibilities and constraints in the following moments. at is why
complex systems are said to be adaptive and evolutionary.
All these briey mentioned characteristics of complex systems seem to be use-
ful to the study of public policies. As stated below most objects of public policies
contain similar features and can be easily labeled complex systems. e relevance
of viewing objects of public policies as complex systems is that the associated
methods and methodologies available for the study of such systems could be ap-
plied to public policies, helping improve their analysis.
A Complexity Approach for Public Policies
|
19
Essentially that means that modeling and simulation can be used to inves-
tigate public policies. is is especially relevant in areas of public policies where
experiments are usually not simple, cheap or even viable.
To simulate means to model the action and the interaction among citizens, rms,
institutions, and the environment constrained by legislation and regulation, the
budget, politics and spatial boundaries (…) working with complex systems applied
to public policy means to create computational experimental environments in which
the essence of the systems is present and from which one can withdraw elements
of improvement of public policies in a relatively simple and cheap way, besides
increasing the understanding of the eects (spatially and temporally) of the policies
(Furtado and Sakowski, 2014).
us, complex systems methods have the potential to inform public policies
and help tackle their eects, eectiveness, direct and indirect costs.
roughout the book, similar denitions of both complex systems and public
policies are presented. However, each one emphasizes dierent perspectives which
add up to a more complete denition as the book progresses. Fuentes recovers the
denition of Murray Gell-Mann and Seth Lloyd of “eective complexity” as “the
length of a highly compressed description of its regularities” (p. 68). Sichman in
chapter 5 – focuses on interactions as “information processing.Tessone delves into
the concept of heterogeneity distinguishing idiosyncratic heterogeneity, such as
cultural heritage; from endogenous heterogeneity that surfaces as a complex system
unfolds. Dawid and Orlando Gomes discuss economics and remind the reader
of the relevance of non-equilibrium states. Bettencourt states that the problem of
interacting citizens inhabiting urban spaces comes down to “how to create a set
of processes in space that makes such interactions possible at a cost that is com-
mensurate with their benets” (Bettencourt, in this volume, p. 227). Mueller picks
up from Sichmans information processing idea advocating that
policy is information-intensive when information is scarce; it tries to centralize a
policy that is inherently local; it assumes the ability to control the process when in
reality it can only act reactively; it requires measurement and evaluation along a series
of diverse and subtle margins, while in reality a single and imprecise metric is used
(the number of settled families); it deals myopically with a policy area that unfolds
over the long-term (Mueller, in this volume, p. 268).
Consistently, the chapters refer to the dynamics of inuence between objects
and subjects in time; the dynamics of crossed-eect causalities in which “eects and
outcomes are, at the same time, causes and inputs of what had produced them
(Morin 2011, p. 74, apud Sakowski and Tóvolli, p. 321). us, methodologies
have to be able to explicitly “account [for] endogenous change” (Almeida, in
this volume, p. 345) or “explicitly capture the underlying causal hypotheses of
policy proposals in a way that allows us to experiment” (Gentile et al., in this
volume, p. 78).
Modeling Complex Systems for Public Policies20
|
is discussion of dynamics also leads to a debate over the timing of the
analysis and the timing of policy. Bettencourt argues that there are problems that
can be managed in “its simpler, shorter-term technical management”, but also phe-
nomena that yearn for “longer-term complex challenges” (Ettema, in this volume,
p. 222). For Ettema, transportation may fall into both categories: a performance
metric-based engineer-like problem, such as the frequency of a single bus line; but
transportation would also fall well into a city-mobility long-term intricate issue.
Still on the dynamics of processes, Jacobson says that cognitive learning takes place
in a varied number of places, moments and experiences through minutes, hours,
semesters and years.
Finally, Mueller also mentions that typical evaluation of public policy is im-
plicitly based on a denition of a system that can be easily tracked and measured
upon metrics that are known a priori. is assumption leaves no room for systems
that adapt, evolve and learn which are exactly what objects of public policies, such
as the economy, the environment, the society and the cities (chapters 6-9) are.
3 METHODS
e methods and methodologies
6
used in complex systems approach come from
already existing disciplines and are not new themselves. However, they reect the
principles and concepts discussed above.
us, a rst thing to point out is explicitly considering the nonlinearity of
systems. Put simply, nonlinear systems are those in which the outputs are not pro-
portional to the inputs. Nonlinearity is attached to the idea that interaction among
elements may generate emergent behavior. Also, the systems outcome cannot be
entirely deductible ex ante. Approaches that include nonlinearity have been used
in applications of physics (laser, superconductors, uid dynamics, and engineer-
ing), biology (biological rhythms, insect outbreaks, genetic studies), chemistry and
cryptography (Strogatz, 2014).
Network analysis studies interactions (edges) among parts (nodes). How
strong, how lengthy and how relevant are the links among people or institutions?
How connected is a given network so that a change in a specic node would af-
fect the connections signicantly? ose are some of the questions that network
analysis may help answering.
7
Strictly connected to the analysis of networks is information theory or, ac-
cording to Shannon (1948), theory of communication. Information theory was
proposed before network science and it is related to the denition of what infor-
6. A detailed description of methods and methodologies is found in chapter 3.
7. See Newman (2003), Newman et al. (2006) and William and Martinez (2000).
A Complexity Approach for Public Policies
|
21
mation is; to the quantication and denition of the elements involved in any
information exchange, and its storage and compression. It is from this theory (and
probability theory
8
) that quantities such as entropy and mutual information come
into play. ese quantitative measures are applied to dierent areas of science from
telecommunications to biology to probability theory to statistical physics, computer
science and medicine. A central aspect of information theory and its associated
measures is the quantication of uncertainty. Given past information, how uncer-
tain is the next bit? is is related to the notion of a measure of complexity and
also to the denition of entropy (Crutcheld and Feldman, 2001; Gell-Mann and
Lloyd, 2004; Szilard, 1964; Turing, 1952).
Two other very commonly used methodologies within complex systems are
cellular automata (CA) and agent-based models (ABM).
9
ey are similar in the
sense that both use agents – of free and ample design – that follow rules. e usage
of ABMs and CA is a way to simulate the interactions in the system and the ensuing
emergent properties. e dierence between CAs and ABMs is that the former is
xed in space and the latter may be mobile. CAs are more relevant to study spatial
analysis where local interactions, physically bounded, are relevant to the problem
at hand. ABMs, in turn, can be modeled to be xed or mobile and they can be in
such a framework that space is completely irrelevant. ey can even be thought
so that the agents are connected through links, thus resembling network analysis.
Finally, it is worth mentioning eorts arising from computing science and
contemporary availability of detailed, micro, spatially-precise data. is abun-
dance of data is fertile land for the use of methodologies such as data mining,
machine learning and articial intelligence, which are collections of techniques
that can be put together to help simulate complex systems and which are likely
to improve insightfulness.
3.1 Methodologies’ tools
Most methodologies are implemented using computational methods. Actually, it
is the availability of computing power along with databases that are temporally-
spatially-individually detailed that helped fuel complex systems in recent years.
10
ere is a number of customized software developed to run specic proprietary
and open-source models.
11
8. Such as in clustering and decision tree procedures.
9. A thoroughly review of the application of ABMs in social sciences is found in chapter 5.
10. Journals dedicated to complex systems include: Journal on Policy and Complex Systems, Complex Systems, The
Journal of Artificial Societies and Social Simulation, Complex Adaptive Systems Modeling, Ecological Modelling, Advances
in Complex Systems, Computers, environment and urban systems, Complexity, Computational Economics. A list of 41
complexity centers can be found at http://en.wikipedia.org/wiki/Complex_systems.
11. Examples include, not exhaustively: MASON, Swarm, RePast, NetLogo, Flame, MASS, and at least 78 others <http://
en.wikipedia.org/wiki/Comparison_of_agent-based_modeling_software>.
Modeling Complex Systems for Public Policies22
|
Models can also be simulated in typical program language such as C++, Java,
or statistical and modeling programs (Matlab or Mathematica). As a high-level,
exible language, Python has been used quite a lot for simulation and modeling
(Downey, 2012; McKinney, 2012; North, Collier and Vos, 2006) – for example
using the SimPy
12
library – or associated to spatial software, such as QGIS. Spe-
cically for network analysis, Pythons library NetworkX
13
is very useful for both
creating and analyzing networks.
A software program that has been around for some time now is NetLogo.
14
Based on Java, it contains a user-friendly set of commands that quickly takes the
beginner programmer to an operational modeler. It allows for cellular automata
spatially-bounded modeling as well as for full agent-based models. More recently
it has incorporated network-like link capabilities and it is easily coupled with other
languages and analysis programs such as Python, R or QGIS.
4 PUBLIC POLICIES AS COMPLEX OBJECTS
is section discusses the complex nature of objects of public policies, such as
social, economic, urban and environmental systems. e hypothesis is that all
these objects can be easily dened as complex systems. Chapters 6 to 9 of the
book deepen the arguments.
4.1 Social
Social systems can be described as a collection of heterogeneous agents (individu-
als, banks, countries etc.), whose state (opinion, liquidity, wealth, etc.) inuences
and is inuenced by the state of others, and whose interactions give rise to global
properties of the system that are more than the sum of individual behavior. ese
features characterize social systems as complex. Understanding how these systems
respond to external inuences is of particular interest for the analysis of public
policy. For example, how does a social system respond to an external signal such
as a change in policy? Simulating the eects of policy change is particularly useful
to steering policy measures.
4.2 Economy
An economic system is composed of heterogeneous actors, with dierent charac-
teristics, expectations and behavioral rules that interact with each other and with
the environment. Besides, the actors are in constant adaptive learning, generating
evolutionary systems. e traditional or classic view, based on the assumptions of
12. Full documentation is available at: <https://simpy.readthedocs.org/en/latest/>.
13. Full documentation is available at: <https://networkx.github.io/>.
14. Full documentation is available at: <https://ccl.northwestern.edu/netlogo/>.
A Complexity Approach for Public Policies
|
23
market clearing, perfect foresight, and equilibrium behavior, does not focus on
the aforementioned elements, producing a more abstract analysis, which makes it
dicult to comprehend the system as a whole.
In this context, alternative models that incorporate such elements are recom-
mended in order to increase the availability of alternate understanding of economic
processes. e heterogeneity of agents and the features of institutional setups that
drive economic interactions should not be ignored. Many methodologies, already
presented in section 2, have been used in order to capture such elements. One of the
most used methodologies in economic modeling has been the agent-based simulation
approach. is method is the basis of the Eurace@unibi model,
15
a closed agent-
based macroeconomic model that has been used as a unied framework for policy
analysis in dierent economic policy areas, such as scal policy, labor market, and
issues related to income inequality. Besides, not only Agent-based Computational
Economics models (Farmer and Foley, 2009; Le Baron and Tesfatsion, 2008), but also
network analysis (Jackson, 2010; Newman, 2010), and analytical approaches for the
analysis of agent models (Alfarano, Lux and Wagner, 2008; Dawid, 1996; Delli Gatti
et al., 2012), are useful for a clearer picture of the dynamics of economic systems.
4.3 Cities
Cities in particular or urban spaces in general are par excellence places where people
and institutions entangle themselves, usually, in productive and innovative ways
(Glaeser, 2012; Jacobs, 1970). However, to reach the most out of their potential,
people and institutions need to cover some basic functions within their shared
space: dwell, commute, work and play.
16
On top of it all, cities are politically
managed, which reinforces the fact that even those four basic actions cannot be
accomplished individually. All activities share a common space. Moreover, cities
are thought out to thrive, to harvest the best (and sometimes the worst) of societ-
ies. us, using sectorial policies, such as housing policies, sanitation policies or
transport policies with no theoretical and methodological background to rmly
go through the interactions – as mentioned above – makes applying policies to
cities very hard work.
Even the approach to cities as an object of science may dier signicantly.
Are cities to be viewed as machines to be “xed”, as markets to be regulated (or
freed), as organisms in a jungle ecosystem, or as a social exercise in which political
or religious values prevail above all?
Mainly, the message of relevance is that attempts to change the city – and
occasionally even inaction and omission on policies on the city – have to be made
15. See chapter 9 and Dawid et al. (2012, 2014) for details of the Eurace@unibi model.
16. Those are the four principles of the functional city proposed by architect Le Corbusier in Charte d’Athènes in 1943.
Modeling Complex Systems for Public Policies24
|
with clear view of its consequences across all aspects and layers of the city. In short,
city planning calls for integrated, connected, nonlinear, dynamic approaches.
As those attributes are typical of complex systems, it may be of interest to apply
them to the study and policy applications of cities.
4.4 Environment
Sustainable development is one of the major challenges for society today. How to
manage natural resources in a world that is more and more complex, and where
everything is interconnected? How to deal with sustainability problems, such as
climate change or biodiversity conservation that are too complex to be tackled by
a single discipline?
Complex systems views and methodologies can provide tools to help analyze
these social-ecological systems and to inform environmental and sustainability
policy making. Actually, many of the insights and concepts from complexity theory
come from the eld of biology.
Emergent behavior and information processing is often exemplied by the
way ants forage for food, or how neurons interconnect to produce global cognitive
behavior. e immune system is another example of self-organization, through
which the interaction of simple cells leads to complex behavior without the presence
of a central controller. Food webs and trophic dynamics are used to understand
biodiversity and to analyze the implications of dierent types of disruptions to
the ecosystem.
Modeling can be a valuable approach to understanding the dynamics of en-
vironmental systems. rough modeling, one aims to identify the key factors and
rules governing a system, allowing the simulation of dierent scenarios and the
performance of sensitivity analysis. is approach has been used to study climate
change, the spread of diseases and the change in land use over time.
Modeling can also help identifying dangerous tipping points
17
in the social
ecosystem. is can be useful, for instance, for the management of water resources,
which might have a turning point, after which water pollution becomes costly
and dicult to reverse.
Similarly, conservation policy can benet from the analysis of food webs and
the resilience of ecosystems to external shocks, such as an increase in deforestation
or in carbon emissions.
17. Mitchell (2011, p. 253) defines tipping point as “points at which some process (...) starts increasing dramatically
in a positive-feedback cycle.” See also Gladwell (2006).
A Complexity Approach for Public Policies
|
25
ese and other methodologies from complex systems can help gure out how
to manage natural resources, how to build sustainable cities, and how to promote
more eective environmental and sustainable policies.
5 OTHER SYSTEMS AND APPLICATIONS
5.1 Education
A considerable amount of research has been done exploring the complex nature
of educational systems, learning and teaching. A report from Organisation for
Economic Co-operation and Development – OECD (Snyder, 2013) investigates
how to operationalize a complexity approach to educational reforms, and provides
examples of educational reforms that have used complexity principles in dierent
countries. Other studies (Lemke et al. 1999; Morrison 2003, Batista and Salvi
2006; Santos 2008) focus on the complex nature of learning, with a focus on
curriculum development, calling attention to transdisciplinarity. One academic
journal
18
is dedicated exclusively to the study of education and complexity (Davis,
Phelps and Wells, 2004).
5.2 Transport
Transport is a typical example of a system composed by a large number of in-
teracting, independent agents, who follow some rules, and who react to their
local environment; a system from which emergent, collective behavior can be
observed. If a number of commuters have to travel a specic route across the city
and they have some window interval to do that, they might probabilistically just
decide to go at the same time. at (unlikely) decision is denitely suboptimal
as it decreases the total capacity of ow of the system. Also, if a central trac
controller established a specic, precise time of departure for all travelers, one
small disturbance might once again settle total congestion. On average, neither
will occur. Anyway, the example shows that transport systems are complex, within
the concepts described above.
Planners and transport engineers have used simulation models in order to
derive scenarios or possibilities that are not able to pinpoint exact ows of trac,
but that can predict the size of the demand on the system, specifying at times,
how the system has to be dimensioned.
A more recent usage of modeling in transport attempts to simulate both the
dynamics of the city – considered as density and land-use type – coupled with the
dynamics of commuters. UrbanSim (Waddell et al., 2007) is a pioneer example.
More sophisticated modeling also tries to compute location and change of the job
18. Complicity: an international journal of complexity and education.
Modeling Complex Systems for Public Policies26
|
market and the behavior of housing markets. Together, the models try to anticipate
the "movement" the city is taking – along with its possibilities or “vocation” – and
attach the planning of the transport system accordingly.
All in all, as most other modeling experience, modeling in transport may help
policy-makers envision scenarios in which key adjusting parameters are visible and
their consequences measured.
5.3 The legislative process
e process of law-making entails heterogeneous individuals (legislators), usually
under no centralized control, who strategically interact with each other in order
to produce collective decisions.
19
When this interaction occurs under a majority
rule institution, collective choice problems may arise. In this sense, complexity
theory might help explain why outcomes vary within the context in which they
are embodied, and how legislative institutions emerge and change.
6 COMPLEX SYSTEMS AND PUBLIC POLICIES
is section summarizes the main insights regarding the use of complexity concepts,
methods and methodologies to public policy.
First, complexity concepts can prevent an oversimplied view of the objects
of public policy. Complexity points out that, when thinking of public policy, one
has to consider that agents are heterogeneous.
6.1 Agents are heterogeneous
Assuming a representative agent, such as an average consumer or rm, can be highly
inaccurate and produce misleading insights for public policy. is is specially the
case in countries like Brazil, where inequalities of dierent types are prevalent.
As Claudio Tessone summarizes it, “heterogeneity can crucially aect the
observed properties of the system, and also be the source of a priori unexpected
phenomena in socio-economic systems.
6.2 Everything is interconnected
is is another way of saying that "the whole is more than the sum of the parts";
that non-trivial complex behavior emerges from the interaction among agents; or
that systems are nonlinear. In public policy, this brings awareness to the fact that
many traditional linear type analyses might be inadequate or insucient. is feature
also points out that the connections among agents, sectors, and scales should not
be neglected, suggesting an interdisciplinary and systemic view of policy objects.
19. This section is based on the contributions of Acir Almeida to the Project “Modeling Complex Systems for Public Policies.”
A Complexity Approach for Public Policies
|
27
e analysis when viewed by a multiplicity of sectors warrant that exter-
nalities, interests and perspectives are properly weighted among each other. e
multiplicity of scales links the microanalysis – at the level of individuals, rms or
the household – to the macro analysis of communities and parties, large sectors
of the economy, neighborhoods, and cities and metropolis.
e multiplicity of scales seems central given that the emergence of patterns
or, similarly, the eectiveness of public policies tends to be specic to one scale
and not automatically valid over other scales. ere are continuous interaction
and idiosyncrasies in interaction across scales. is is especially true when consid-
ering public policies objects, especially across federative levels. Macroeconomic
policy, such as interest rate setting, generates results that vary by regions, sectors,
and rm size. It may impact suppliers and buyers dierently. Further, actions of
multiple agents with multiple interests, means and views may generate results
that can also dier in scope, speed of occurrence, qualitative characteristics and
permanence of eects.
6.3 Policy does not work with clear, linear or immediate cause and effects
e hope for action-reaction policies might be somehow naïve, as complex systems
do not work in a mechanical way, but change, evolve, and adapt. ey are dynamic.
Policy should thus take into consideration multiple causalities and indirect eects
that arise as a consequence of the interaction among dierent agents.
Romanian philosopher Basarab Nicolescu (1999) lists three fundamental
principles of the hard sciences that are not easily applicable to human sciences.
ey are: i) the existence of general, fundamental laws; ii) the use of experiments
to decode such laws; and iii) the possibility that given the same conditions (coeteris
paribus), independently, it would be possible to replicate the experiments and thus
the laws that they attest.
e diculties to apply the fundamental laws, their experimentation and
replicability is clear in social phenomena and public policies by realizing the i)
discontinuities, jumps and ruptures; ii) unique, discrete events, that do not follow
a clear universal pattern which could be decoded into mathematics in any immedi-
ate way; and iii) uncertainties which together with subjectivity of actors and lack
of coherent and strict rationality leads to a non-deterministic social environment.
erefore, policy might be more eective if geared towards i) improving the
resilience of the system and decreasing its vulnerabilities; ii) avoiding (promoting)
dangerous (positive) tipping points, and iii) identifying the key actors in a network
that can promote changes in the system.
Modeling Complex Systems for Public Policies28
|
In other words, an OECD document states: “(…) it is not uncommon for
small changes to have big eect; big changes to have surprisingly small eects; and
for eects to come from unanticipated causes” (OECD, 2009, p. 2). is means that
policy-making should try to understand the underlying mechanisms of the system
under analysis in order to identify how to best steer it towards the desired path.
Second, complexity methods and methodologies can help take into account
the complex features of the systems under analysis.
1. Modeling is a good strategy to obtain better understanding of how a
system works, and one which allows incorporating the complex features
of the system. Modeling can help identify the important players in
the system under analysis (agents), their dierent characteristics (het-
erogeneity), their interrelations (interconnectedness), and how these
components together give rise to complex and sometimes unexpected
behavior. Examples of such modeling techniques are cellular automata
and agent-based modeling. Heemskerk and colleagues collect a clarifying
sequence of modeling denitions:
A model is an abstraction or simplication of reality. Scientists often use models to
explore systems and processes they cannot directly manipulate (Jackson et al. 2000).
Models can be more or less quantitative, deterministic, abstract, and empirical. ey
help dene questions and concepts more precisely, generate hypotheses, assist in test-
ing these hypotheses, and generate predictions (Turner et al. 2001). Model building
consists of determining system parts, choosing the relationships of interest between
these parts, specifying the mechanisms by which the parts interact, identifying
missing information, and exploring the behavior of the model. e model building
process can be as enlightening as the model itself, because it reveals what we know
and what we dont know about the connections and causalities in the systems under
study (Levins 1966; Jackson et al. 2000; Taylor 2000). us modeling can both sug-
gest what might be fruitful paths of study and help pursue those paths (Heemskerk;
Wilson; Pavao-Zuckerman, 2003).
2. Modeling permits simulating scenarios as a decision-support tool to
inform policy making. Models work as platforms for so-called in silico
experiments, by means of which dierent policy options can be compu-
tationally simulated and “cheaply” tested.
3. Modeling stimulates a forward-looking, prospective view of policy, by
allowing scenario building and testing. Models can enable prognosis that
are less based solely on probabilities but that include essential interactions
at various scales and with various agents’ interests considered. Policy-
makers can thus work with spaces of scenarios and realms of probabilities
that occur given known rupture points.
A Complexity Approach for Public Policies
|
29
4. Models can be continuously improved, as more knowledge is gained about
the system. Models can also be simple and provide general insights, or
specic to help tackle a particular problem.
5. Models are a means of communicating ones ideas and theories and can
work as a "meeting point" for collaborative work among interdisciplin-
ary teams. “Models not only help formulate questions, clarify system
boundaries, and identify gaps in existing data, but also reveal the thoughts
and assumptions of fellow scientists” (Heemskerk, Wilson and Pavao-
Zuckerman, 2003).
6. e notion of multiple models contributes to the understanding of social
phenomena in particular and of public policies in general because it is
based on the richness of diversity, dierence and dissimilarities (Page,
2007). As Page (2007) argues, no single model can independently cover
comprehensively the intricacies of some phenomena, especially those of
subjective nature, complex ones. He also states that models section the
analysis with specic parameters, be it from the theoretical, methodologi-
cal or procedural point of view. us, the diversity of models implies
a larger coverage of possible scenarios that are more keen to envelope
unexpected sequences, unlikely important events, unique tipping points.
ird, data are a valuable resource for policy making and complexity methods
give insights into how to use them to the best extent.
1. Data can help visualize, describe and identify features of the system to be
better explored. Social network analysis, for instance, relies on the visual
representation of networks to convey complex information.
2. Data mining, machine learning, network analysis and other association
studies can provide insights into the functioning of the system.
3. Data can help validate and improve models.
Finally, knowledge can be viewed as a feedback process, “an endless cycle of
proud proposing and disdainful doubting” (Mitchell, 2011, p. 295). Modeling
provides a way to structure this process and to improve the understanding of the
system one wants to impact. e cycle of data analysis, modeling, validation,
simulation, implementation, data analysis, remodeling and so on might be the
"strange loop" that can provide decision support for tackling complex problems
through public policy. If not a certain, determined path to be tread on, complex
systems may illuminate the key pathways to policy-makers, clarifying what is
likely to happen given choices of sets of paths, after so much has been traveled on.
Modeling Complex Systems for Public Policies30
|
7 MAPPING THE BOOK
e book is organized in three parts. Part I – Complexity: theory and methods
discusses the main concepts of complex systems, its methods and methodologies
and brings two chapters specically on the computational modeling needed to
implement such an approach.
is is the introductory chapter. e second chapter Complex systems: concepts,
literature, possibilities and limitations written by William Rand presents the main
concepts of complex systems, and briey describes some complex systems´ methods.
Besides, it discusses possibilities and limitations of complex systems analysis, in
contrast with traditional methods, indicating the advantages of complex systems´
applications to public policy.
e third chapter Methods and methodologies of complex systems by Miguel
Fuentes presents more technical details and a guided reading of the literature on the
methods that are commonly used in the complex systems approach. e chapter
provides a discussion at the conceptual level and references for further reading,
aiming to reach readers from dierent elds.
e fourth chapter, Simulation Models for Public Policy, is authored by
computer scientists James E. Gentile, Chris Glazner and Matthew Koehler. e
chapter presents an overview of modeling and simulation, with clear statements
for stakeholders and concerned audience. It argues that computational modeling
can be an interesting tool for policy analysts to compare policy options. It focuses
on ABM, and gives an overview of its benets for policy analysis. Besides, the
chapter discusses each step of model construction – implementation, verication,
validation, and renement –, pointing out the main challenges in each of them.
Chapter 5, Operationalizing complex systems, by Jaime Sichman presents some
concepts and tools of computational simulation, and provides a detailed panorama of
the main tools used in the complex systems approach. e chapter aims to help the
reader interested in implementing the methods and techniques of complex systems
computationally. e chapter focuses on the concepts and implementation tools
of Multi-Agent-Based Systems (MABS), but also discusses the implementation of
other methods, such as social networks and machine learning.
Part II contains four chapters that together qualify grand objects of public
policies as complex systems.
Chapter 6 Understanding the environment as a complex, dynamic natural-social
system, by Masaru Yarime and Ali Kharrazi discusses the coupling-uncoupling of
social natural systems and the implications of viewing sustainability from a systems
perspective. eir approach views both the quantitative and the qualitative dimen-
sions of concepts such as resilience, eciency and redundancy. After developing
A Complexity Approach for Public Policies
|
31
their conceptual framework, they move on to address actual governance cases of
networked systems and their public policies eects.
e complex nature of social systems, by Claudio Tessone, is the theme of chapter 7.
e chapter discusses why society should be viewed as a complex system, and
presents the challenges involved in modeling the complex behavior found in
social systems. Besides, the chapter discusses what the policy implications of
this view are. It also describes the characteristics of social systems that are most
relevant to the analysis of public policy, such as the heterogeneity of agents, the
dynamic evolution of society by means of interaction and feedback, the systemic
nature of society that renders its decomposition or break down in dierent aspects
inadequate, and the niteness of such systems.
Economics is discussed in two chapters. Chapter 8 e economy as a complex
object by Orlando Gomes presents a more general defense of economics as a complex
object. It reviews the contemporaneous literature on complexity economics, and
discusses why the macro economy should be analyzed as a complex system.
The chapter also provides an illustrative example of a complex economic environ-
ment simulated with a network model. Chapter 9 Modeling the economy as a complex
system by Herbert Dawid continues the line of thinking, focusing on how to model
the economy under a complex systems framework. e chapter emphasizes agent-
based modeling, and discusses the advantages and disadvantages of this modeling
approach for economic analysis. e chapter is particularly strong in discussing
issues related to modeling the economy for policy analysis and provides insightful
illustrations of applications to public policy, such as the Eurace@Unibi model.
Chapter 9, Cities as complex systems authored by Luis Bettencourt discusses
why cities should be viewed and analyzed as complex systems. It presents a brief
historical overview of the concepts of city, and how they have been perceived
in urban planning and policy. en it describes the main properties of cities as
complex systems, and discusses how this new understanding of cities reveals that
urban areas of dierent sizes pose dierent challenges to the planner. e chapter
discusses the implications of the complex systems approach for urban planning
and policy, and counterbalances it with problems of engineering solutions.
Part III presents applications in the world and in Brazil, besides a number of
applications in transport, education, the legislative process and a territorial approach.
e rst chapter of the applications, Complexity theory in applied policy world-
wide, by Yaneer Bar-Yam emphasizes the importance of analysis of the potential
eects of policy changes in one part of the world into another part, considering the
increasingly interdependent world. e chapter highlights some methodologies of
complex systems, such as multiscale analysis, networks, and patterns of behavior.
It also presents the applications of such methodologies in the analyses of nancial
Modeling Complex Systems for Public Policies32
|
and commodities markets, disease spread, and violence, and argues that complex
systems have been proven to explain and predict global phenomena.
Bernardo Mueller, in Complex system modeling in Brazilian public policies
presents two case studies of policy in Brazil, one success and one failure, and
builds on these examples to explain why a complex systems approach may be more
adequate to evaluate public policies of complex systems rather than usual metrics
widely disseminated. Further, the chapter presents a panorama of studies related
to public policies' issues that have used the complex systems approach in Brazil.
e mapping of such studies indicates the potential use complex systems’ meth-
odologies in relevant areas for the country, and also allows the reader to identify
researches on his/her topic of interest
Chapter 13, Complexity methods applied to transport planning by Dick Ettema,
looks into a specic area of public policies: transport planning. It discusses why trans-
port should be viewed as a complex system, and reviews the main characteristics of
existing complex methods in transport planning, the implementation issues related
to these methods, and the main implications for the transport system, for cities and
society. In general, the chapter provides an overview of trac and transport simulation
models, highlighting the innovations and challenges for complex transport models.
Two chapters focus on education. e rst one, Education as a complex system:
implications for educational research and policy, by Michael Jacobson, discusses why
education should be considered a complex system, and which are the methodologi-
cal implications of this view for researchers and policy makers. Besides, the chapter
provides an overview of applications of complexity methods in educational policy and
research. e second chapter, Complex approaches for education in Brazil, by Patrícia
Sakowski and Marina H. Tóvolli, contributes to the discussion on education made
on the previous chapter, adding to the conceptual discussion and looking speci-
cally to Brazil. e chapter presents what has been done in the complexity area in
Brazil, and discusses how this approach may contribute to education in the country.
Chapter 16, Overcoming chaos: legislatures as complex adaptive systems, by
Acir Almeida, describes the complex nature of legislatures and discusses why they
should be seen as complex adaptive systems. It presents two main models of leg-
islative organization and discusses the limitations of these traditional approaches
in explaining the evolution of legislative institutions. e chapter highlights the
potential contribution of the complex systems approach for the analysis of the
emergence and change of institutions. Looking specically to the Brazilian case,
the chapter points out how the complex systems approach may explain the recent
evolution of law-making patterns in the Brazilian Congress.
Finally, chapter 17, e territory as a complex social system, by Marcos Aurélio
Santos da Silva, focuses on the study of socioterritorial systems, and the need of
A Complexity Approach for Public Policies
|
33
interdisciplinary methods for the analyses of such systems. e chapter presents
the Sociology of Organized Action (SOA) theory to rethink the analyses of power
and dependence relations in socioterritorial systems. e chapter then presents the
Soclab method, which is a formalization of the SOA theory. e chapter highlights
how the Soclab method may contribute to the analysis of social relations in social-
territorial systems through computational simulation.
REFERENCES
ALFARANO, S.; LUX, T.; WAGNER, F. Time variation of higher moments in
a nancial market with heterogeneous agents: an analytical approach. Journal of
Economic Dynamics and Control, v. 32, n. 1, p. 101-136, 2008.
ANDERSON, P. W. More is dierent. Science Magazine, v. 177, n. 4.047,
p. 393-396, 8 Aug. 1972.
COLANDER, D.; KUPERS, R. Complexity and the art of public policy: solving
societys problems from the bottom up. Princeton: Princeton University Press, 2014.
CRUTCHFIELD, J. P.; FELDMAN, D. P. Regularities unseen, randomness ob-
served: levels of entropy convergence. Santa Fe Institute, v. 13, n. 1, p. 25-54, 2003.
DAVIS, B.; PHELPS, R.; WELLS, K. Complicity: an introduction and a welcome.
International Journal of Complexity and Education, v. 1, n. 1, p. 1-7, 2004.
DAWID, H. Adaptive learning by genetic algorithms: analytical results and
applications to economic models. New York: Springer, 1996.
DAWID, H. et al. Labor market integration policies and the convergence of regions:
the role of skills and technology diusion. Journal of Evolutionary Economics,
v. 22, n. 3, p. 543-562, 2012.
DAWID, H. et al. e Eurace@Unibi model: agent-based macroeconomic modeling
and policy analysis. Economics and Management, n. 1, 2014.
DELLI GATTI, D. et al. Reconstructing aggregate dynamics in heterogeneous
agents models. A Markovian Approach. Revue de l’OFCE, n. 5, p. 117-146, 2012.
DOWNEY, A. B. ink complexity: complexity science and computational
modeling. 1
st
ed. California: O’Reilly Media, 2012.
EDMONDS, B.; MEYER, R. Simulating social complexity: a handbook. New
York: Springer, 2013.
FARMER, J. D.; FOLEY, D. e economy needs agent-based modelling. Nature,
v. 460, n. 7256, p. 685-686, 2009.
Modeling Complex Systems for Public Policies34
|
FURTADO, B. A.; SAKOWSKI, P. A. M. Complexidade: uma revisão dos clássicos.
Brasilia: Ipea, 2014. (Discussion Paper, n. 2019).
GELL-MANN, M.; LLOYD, S. Eective complexity. Santa Fe Institute, p.
387-398, 2003.
GLADWELL, M. e tipping point: how little things can make a big dierence.
Boston: Little Brown, 2006.
GLAESER, E. L. Triumph of the city: how our greatest invention makes us
richer, smarter, greener, healthier, and happier. New York: Penguin Books, 2012.
HEEMSKERK, M.; WILSON, K.; PAVAO-ZUCKERMAN, M. Conceptual
models as tools for communication across disciplines. Conservation Ecology,
v. 7, n. 3, p. 8, 2003.
JACKSON, M. O. Social and economic networks. Princeton: Princeton University
Press, 2010.
JACOBS, J. e economy of cities. New York: Vintage Books, 1970.
LE BARON, B.; TESFATSION, L. Modeling macroeconomics as open-ended
dynamic systems of interacting agents. e American Economic Review, p.
246-250, 2008.
MCKINNEY, W. Python for data analysis: data wrangling with pandas, NumPy,
and IPython. 1
st
ed. California: O’Reilly Media, 2012.
MITCHELL, M. Complexity: a guided tour. Oxford: Oxford University
Press, 2011.
NEWMAN, M. Networks: an introduction. Oxford: Oxford University Press, 2010.
NEWMAN, M.; BARABÁSI, A. L.; WATTS, D. J. e structure and dynamics
of networks. 1
st
ed. Princeton: Princeton University Press, 2006.
NEWMAN, M. E. J. e structure and function of complex networks. Siam
Review Journals, v. 45, p. 167-256, 2003.
NORTH, M. J.; COLLIER, N. T.; VOS, J. R. Experiences creating three imple-
mentations of the repast agent modeling toolkit. ACM Transactions on Modeling
and Computer Simulation, v. 16, n. 1, p. 1-25, Jan. 2006.
OECD - ORGANISATION FOR ECONOMIC CO-OPERATION AND
DEVELOPMENT. Applications of complexity science for public policy: new
tools for nding unanticipated consequences and unrealized opportunities. France:
OECD, 2009.
A Complexity Approach for Public Policies
|
35
PAGE, S. E. e dierence: how the power of diversity creates better groups,
rms, schools and societies. Princeton: Princeton University Press, 2007.
SHANNON, C. E. A mathematical theory of communication. e Bell System
Technical Journal, v. 27, p. 379-423, 623-656, 1948.
STROGATZ, S. H. Nonlinear dynamics and chaos: with applications to physics,
biology, chemistry, and engineering. Colorado: Westview Press, 2014.
SZILARD, L. On the decrease of entropy in a thermodynamic system by the inter-
vention of intelligent beings. Behavioral Science, v. 9, n. 4, p. 301-310, Jan. 1964.
TURING, A. M. e chemical basis of morphogenesis. Philosophical Trans-
actions of the Royal Society of London, Series B, Biological Sciences, v. 237,
n. 641, p. 37-72, 1952.
WADDELL, P. et al. Incorporating land use in metropolitan transportation plan-
ning. Transportation Research, v. 41, p. 382-410, 2007.
WILLIAMS, R. J.; MARTINEZ, N. D. Simple rules yield complex food webs.
Nature, v. 404, n. 6774, p. 180-183, Mar. 2000.
CHAPTER 2
COMPLEX SYSTEMS: CONCEPTS, LITERATURE, POSSIBILITIES
AND LIMITATIONS
William Rand
1
1 INTRODUCTION
e goal of public policy is often to alter, or maintain the behavior of a large group of
individuals or organizations to achieve some societally desirable outcome. e
challenge
with evaluating public policy is that every individual in a population does
not react the
same way to the introduction of a new policy or a set of incentives.
Moreover, the overall
result of a public policy is not simply the sum of the individual
reactions; instead, those
reactions interact and feed o of each other. As a result, the
outcome of the implementation
of any public policy is an emergent product of many
individual decisions, and the way that
those decisions interact with each other and
the policy.
For example, a governmental organization or entity may put into practice a policy
such as a tax policy, a speed limit, or an urban renewal incentive. Dierent individuals that
are aected by that policy may react in dierent ways. Since individuals are
not always
perfectly rational, or necessarily law abiding, sometimes they will react in
ways that the
governmental organization never intended. For instance, some organizations may evade
taxes, some drivers may speed, and some residents may move
away from urban areas; while
other individuals act in exactly the way intended by
the policy. Moreover, individuals do
not just react to the policy they also react to
each other and may modify their behavior
based on what they see in others. e
eect of public policy is not just a one-time, static
event, but rather it is the result of a series of actions taken by both government and citi-
zens to achieve a desired
outcome. For instance, new policies may be enacted to attempt
to corral some of the
unintended behavior, or citizen action groups may form to attempt
to alter policy.
e aggregation of all these dierent actions results in an emergent, complex
pattern
of behavior, which will aect future policy making, and will also feedback to
aect individual level decisions. us, the eect of public policy is not just a product
of
government control, and it is not just a product of market forces, and it is not just a
product of citizen action, but rather it is a combined product of the interaction of all
ac-
tors (Colander
and
Kupers
, 2014).
1. Center for Complexity in Business, Robert H. Smith School of Business, University of Maryland, Van Munching Hall,
College Park, MD USA 20742. E-mail: <wrand@umd.edu>.
Modeling Complex Systems for Public Policies38
|
Studies of the kinds of complex interactions that are readily apparent in public
policy
are at the core of the study of complex systems. Complex systems are systems
of interacting,
autonomous components, where the outcome of the system is not
simply the sum of
the underlying parts (
Mitchell, 2009
; Waldrop, 1993;
Casti, 1994
). is makes complex
systems a
natural lens through which to study public policy.
One classic example of complex systems was described in chapter 1 when the
process
of bird ocking was described. Craig Reynolds illustrated in his “boids”
model that
birds can ock without any central leader dictating how the birds should ock (Reynolds,
1987). In the boids model, the agents (birds/boids) follow three simple rules: i) avoid
other birds; ii) head toward the center of mass of nearby birds; and iii)
align your head-
ing with other birds nearby. is model is robust and will generate
emergent patterns of
behavior that resemble ocks under a wide variety of situations. However, none of the
birds contains the notion of a “ock” and the ock as
an entity does not exist, instead it
is entirely composed of individual birds.
Another classic example of complex systems within public policy is something
that
has been explored using a variety of methods over the years, the trac jam
(
Resnick,
1994
). Highway trac is composed of many individual actors, i.e., the drivers of cars,
trucks, and other types of moving vehicles. None of these individual actors denes a
trac
jam. Instead a trac jam is the emergent product of many dierent individual
decisions.
However, the overall emergent pattern of stuck trac, feeds back to aect
individual deci-
sions. Drivers slow down, they change their routes, and they may even alter their decision
to drive in the rst place. is system, which seems simple
at rst, already contains the
basic components of a complex system, specically
emergent patterns of behavior that
feedback to aect individual decisions.
Because of the complex interactions of these systems and the nonlinear way in
which
the elements of a complex system give rise to overall patterns of behavior,
complex systems
can be very dicult to predict, control and manage. erefore the
best use of complex systems
analysis methods for public policy evaluation is not in
the context of perfect prediction,
but rather as a “ight simulator” (Holland, 1996; Sterman, 2000; 1994). A
regular ight
simulator is not the same as ying a plane, but nonetheless provides the potential pilot
with an education about how a plane might react in dierent conditions, and dierent
environments. In the same way, complex systems can give
an analyst or manager the ability
to understand how a policy might play out, and
even develop contingency plans as to
what actions to take in dierent contexts. Some systems cannot be easily manipulated or
changed. A policy ight simulator
can identify these places where no matter what policies
are implemented the system
still winds up in a pre-determined outcome. e incentive
structure or the forces
at work may be such that it is very dicult if not impossible to
alter the process
of the system. ough this may be frustrating, it tells the user of the
policy ight
simulator to look for alternative solutions, or to consider reprioritizing their
goals
and objectives.
Complex Systems: concepts, literature, possibilities and limitations
|
39
FIGURE 1
An Agent-Based Model of Flocking Behavior implemented in NetLogo
Source:
Wilensky (1998
;
1999)
.
FIGURE 2
An Agent-Based Model of Traffic implemented in NetLogo
Source:
Wilensky (1997;
1999)
.
e goal then of a complex systems analysis of public policy is to provide insight
and
understanding of how the complex system of society may be aected by the
application
of a policy. Moreover, by examining a suite of policies, it is possible to
identify the policies
that will have the greatest benet for the least cost. Additionally,
robustness and sensitivity
analysis can be carried out, and supplementary policies
can be examined to help adapt to
unforeseen circumstances.
In this chapter, we will present the basic concepts and ideas of complex systems,
including a brief description of the tools that complex systems employs.
2
We
will then
discuss the possibilities and the limitations of complex systems analysis.
Finally we will
end with a brief discussion of the future of the complex systems
approach to public
policy evaluation.
2.
A longer description of the tools of complex systems is available in chapter 3, and a detailed discussion of the application of
complex system tools to a wider variety of application areas is
available in chapters 6 through 17.
Modeling Complex Systems for Public Policies40
|
2 CONCEPTS AND TOOLS
e basic conceit of complex systems is that many systems that we observe and
want to
understand around us, are best described through methods that enable the
modeling and
examination of the interactions of dierent parts of the system. To
this extent a number
of dierent concepts and tools that are employed by complex
systems science focus on the
interactions and properties of a large number of interacting parts. To explain this in more
detail, we will begin this section by exploring
some basic concepts in this space, and then
move on to examine tools that are used to
examine complex systems. We will then nish this
section by describing some areas
where these concepts and tools have been applied in the
realm of public policy.
2.1 Concepts
ere are several standard features that complex systems regularly exhibit that are
useful to
understand from a public policy perspective. ese features help policy analysts and researchers
describe and comprehend the properties of these systems that
often make them dicult to
manage and predict. As highlighted in the introduction,
there are two main concepts that
are important in every complex system. ey are
emergence and feedbacks.
Emergence is the idea that “the action of the whole is more than the sum of the parts
(Holland, 2014). Complex systems are inevitably composed of many dierent entities or
individuals. ese individuals have their own properties and actions, but an emergent
property is something that cannot be discovered by inspecting any of the individual agents.
Instead it is a product of the interactions of the dierent agents and can only be observed
at the population level (Holland, 1999; Miller and Page, 2009). For instance, in the traf-
c jam example none of the agents denes a trac jam or contains the property
of a
trac jam” within it, however the emergent result of all the actions of many
individuals
is a trac jam. Similarly, no single individual can be responsible for
the development
of a city. Instead, the development pattern of a city is an emergent
result of developers,
residents, employers, politicians, the environmental landscape
and many other factors. No
single entity within this system contains the property
of city development”; instead, that
property is the emergent result of many actors
acting together. Moreover, these emergent
properties feedback to aect individual
decisions. For instance, within the context of city
development the evolution of the
city will eventually aect where developers build new
buildings, where residents
decide to live, what kinds of business will move into the city,
how politicians will
position their campaign platforms, and it will even transform the
physical environment of the city. In turn these decisions will result in new emergent
patterns of
behavior, which will in turn result in new feedbacks.
So how is a public policy analyst supposed to understand these systems? One
of the
best ways is to create a model of the underlying system. As discussed in the
introduction of
this chapter, the results of these models should not be used as perfect predictions or complete
understandings, but rather as ight simulators. In fact, one
of the best uses of complex
Complex Systems: concepts, literature, possibilities and limitations
|
41
systems analysis for public policy evaluation is in the
identication of leverage points
within the overall societal system (Holland, 1996). Leverage
points are places in a complex
system where the system can be altered or changed.
Modeling gives analytics the ability
to identify these leverage points by trying out
many dierent scenarios and interventions
and seeing what policies have the largest
positive eect on the goal that they are hoping
to reach. By identifying leverage
points, it is possible to explore a policy (Bankes, 1993;
Lempert, 2002) and to gure out when a policy and
at what magnitude a policy will be
most eective.
Leverage points are also related to another important concept in complex systems,
known as tipping points. Tipping points are when a system suddenly changes
state based
on a small change in a parameter of the system (Lamberson and Page, 2012;
Mitchell,
2009
; Schelling
, 1972
). In some
elds, this is also called a phase transition (Lamberson
and Page, 2012) or bifurcation (Drake and Grien, 2010). Systems with
tipping points
can sometimes seem like they are not responding at all to public policy that is attempting
to alter them, and then suddenly with just a few small changes
the system will change
dramatically (Shiell, Hawe and Gold, 2008).
However, other systems may be stuck in a state they cannot escape from due to
choices
made early on in the evolution of the state. is is a concept known as path
dependence (Brown
et al., 2005a). Path dependence means that the current possibilities of the system
are in some
sense constrained by the past choices that were made. For instance,
urban development
often features path dependent eects, since residents tend to
move toward where services
are available in cities, and then cities and businesses
tend to place services where there are
lots of residents, meaning that early on when a few residents or services make a few choices
they can dramatically alter the future
development of the city (Brown et al., 2004).
A special case of path dependence is sensitivity to initial conditions (
Mitchell, 2009
).
is
property, which is also a hallmark of chaotic systems, states that every starting point
of
the system is very close to another starting point with a vastly dierent future.
is is sometimes referred to as the “buttery eect”, i.e., as Edward Lorenz put it,
does the ap of a butterys wings in Brazil set o a tornado in Texas?” (Lorenz, 1972).
In
other words, the exact conditions of a system must be known in order to understand
how that system will develop in the future, and, unfortunately, from a predictability
standpoint, knowing close to the exact conditions does not help you very much in predict-
ing the future. is is a very strong claim about a system, and in general
many complex
systems do feature some sensitivity to initial conditions, but do not
have exactly this
property. However, a weak version of this property might just state
that where you start
matters signicantly, which does seem to aect most complex systems. In other words,
many complex systems may be greatly aected by
their starting conditions even though
the resulting states of the systems may not be
completely divergent from similar starting
conditions. However, there can easily be
regions of starting conditions and it may be
Modeling Complex Systems for Public Policies42
|
possible that altering one parameter of
the system can move a system from one region into
another. is is again similar to
the concept of tipping points, but now stated in terms of
the initial conditions of the
system rather than the ongoing state of the system.
Sensitivity to initial conditions and tipping points are some of the many properties
that arise in complex systems that are nonlinear. Nonlinearity was also discussed
in the rst
chapter, but a nonlinear system is one where the inputs do not necessarily