Approved for public release, distribution is unlimited / Recent insurgent tactics during Operation Iraqi Freedom (OIF) have demonstrated that coalition logistical convoys are vulnerable targets. This thesis examines the tactics, techniques and procedures (TTPs) used in convoy operations in an attempt to identify the critical factors that lead to mission success. A ground convoy operation scenario is created in the agentbased model (ABM) Map Aware Non-uniform Automata (MANA). The scenario models a generic logistical convoy consisting of security vehicles, logistical vehicles, an unmanned aerial vehicle (UAV) and an enemy ambushing force. The convoy travels along a main supply route (MSR) where it is ambushed by a small insurgent force. We use military experience, judgment and exploratory simulation runs to identify 11 critical factors within the created scenario. The data farming process and Latin Hypercube (LHC) experimental design technique are used to thoroughly examine the 11 factors. Using the 11 factors 516 design points are created and data farmed over to produce 25,800 observations. Additive multiple linear regression is used to fit a model to the 25,800 observations. From the created scenario it is concluded that: convoy mission success may be determined by only a few factors; the actions of logistical vehicles are more critical than those of security vehicles; UAVs provide a statistically significant advantage; and ABMs coupled with LHCs and data farming are valuable tools for understanding complex problems. / Captain, United States Marine Corps

Approved for public release, distribution is unlimited / This thesis examines a 2015 Marine Expeditionary Brigade scheme of maneuver as the baseline scenario for a commercial logistics support software program called SEAWAY. Modifications to this scenario are conducted using a designed experiment in order to explore how the plan characteristics relate to eleven specified input factors. Multiple regression analysis is used to fit models to the resulting data for three different measures of performance: Total Aircraft Sorties, Total Aircraft Sortie Time and Total Aircraft Tons. The results suggest the plan performance is predicted well by a small subset of the factors and their interactions. One implication of this work is a better understanding of which factors are key determinants of the plan characteristics for variations on this specific base scenario. By using these fitted models, the number of SEAWAY runs needed to identify acceptable plans should decrease dramatically. The approach in this thesis provides a blueprint for similar analyses of other scenarios by demonstrating how information gained from models fit during an exploration phase might allow the logistician to quickly determine factor settings that yield an acceptable plan once details of an operation become available. Finally, working with the SEAWAY developers provided them with some new insights. / Lieutenant Commandeer, United States Navy

O presente estudo busca desenvolver um modelo baseado em agentes para estudar em ambiente simulado baseado em sistemas computacionais as propriedades emergentes decorrentes da aplicação e execução da lei penal. Os modelos baseados em agentes têm-se revelado uma alternativa bastante promissora no estudo de sistemas complexos onde a grande quantidade de variáveis dinâmicas e interagentes tornam impossível o uso de técnicas puramente analíticas para a sua abordagem. Especialmente considerando que normas legais penais feitas para regular a conduta das pessoas em sociedade devem durar, a dimensão tempo acaba recebendo uma considerável relevância, o que incentiva ainda mais o uso de sistemas multiagentes, particularmente propícios nestes casos. Buscou-se construir um modelo no qual fosse possível analisar qualquer lei de natureza penal, quaisquer que fossem suas características e a quem estivesse destinada. Chegou-se, para tanto, no construto Cidadão- Estado-Oportunidade-Ambiente, que definem precisamente os contornos de qualquer conduta criminosa e que foram modelados em ambiente mutiagente. Para construir os fundamentos do comportamento dos agentes foram utilizados especialmente os referenciais teóricos de Shavell e Polinsky (2000, 2004), Becker (1968), Cohen e Felson (1979) e Clarke(1995), sem deixar de agregar outras contribuições teóricas, ainda que em escala menor. O construto, no entanto, permite que quaisquer contribuições e análises teóricas de atributos do comportamento dos agentes propostos possam ser utilizadas, uma vez que a sua generalidade permite que qualquer forma de conduta possa ser modelada. Para criar o simulador, foi utilizado o ambiente NetLogo 3.1.4 (WILENSKY,1999), cuja especial capacidade de tratar fenômenos que se desenvolvem ao longo do tempo se revelou de extrema valia e praticidade, demonstrando o potencial surgimento de propriedades imprevisíveis sob outras condições. Os testes realizados para aferir a aderência do modelo, usando distribuições de freqüência onde as médias e os desvios-padrão eram controlados, mostraram que o comportamento dos agentes em ambiente dinâmico segue perfeitamente os conceitos teóricos utilizados como referencial, especialmente no que diz respeito ao comportamento do agente cidadão, modelado segundo os conceitos estabelecidos por Becker (1968). Restou demonstrado pela simulação que a percepção de possibilidade de punição tem impacto considerável no comportamento dos agentes diante da possibilidade de delito, assim como o somatório dos custos envolvidos é grandemente influenciado pelo tipo de delito cometido, ou seja, a dimensão monetária do dano, pela estrutura de custos da justiça e pelos meios de arrecadação destes recursos utilizados pelo estado, conforme afirmam Shavell e Polinsky(2000). Para realizar os testes foram construídas tabelas de valores utilizando a técnica 2k fatorial, descrita em detalhes em Law e Kelton (1982), onde cada um dos fatores de impacto foi desdobrado em valores dinâmicos compatíveis com o sistema real no qual o modelo foi baseado. O teste 2k fatorial é especialmente prático no sentido de aferir a aderência do modelo de simulação aos referenciais teóricos, pois permite grande flexibilidade ao analista na formulação de cenários, pois quaisquer combinações possíveis podem ser testadas, dependendo apenas do foco de interesse da pesquisa. Foram realizados também experimentos de simulação com o objetivo de verificar o comportamento do modelo ao atingir estado estacionário. Os testes de validação e os experimentos de simulação apresentaram resultados semelhantes. / Agent based models are considered a milestone in social sciences research mainly because it´s random behavior allows a variety of analysis that were impossible to make due to complexity constraints. The term complexity usually derives from a large number of interacting and dynamic variables which cannot be considered singularly, in order to keep the whole system attributes intact. For that matter, a systemic approach of analysis is required not only to keep this information within the framework of the system, but also to get the emergent properties that come along with the interaction, not observable singling out one or another factor. Considering these very specific characteristics an agent based model, or multiagent system turns to be a more suitable analysis than analytic counterparts due to its very particular capability to deal simultaneously with a large number of random variables. This research work develops an agent based model to study the emergent properties due to public enforcement of criminal law. In order to build a model of criminal behavior suitable to the project, well known theories about crime were taken in consideration, including Shavell and Polinsky (2000), Shavell (2004), Becker (1968), Cohen and Felson(1979) and Clarke (1995). The construct derived from the analysis is a model based on three interacting agents and the environment, namely: citizen-opportunity-state-environment (offender-victim-public enforcement of law-environment). These interacting agents have the ability to define the boundaries of any given offence, regardless the kind, being capable to cope with from a simple misdemeanor (e.g. a traffic violation) to a felony (e.g. a capital murder). Particularly, the citizen-agent behavior has been modeled according to Becker’s (1968) Crime and Punishment seminal work. These remarks explains why originally the term citizen has been used in first place, and not offender. Rational Choice theory says that anyone can be an offender, depending on certain conditions. The State performance regarding its economic impact in terms of the public enforcement of criminal law is seen through the lenses of Shavell and Polinsky’s theory (2000). The opportunity and environment agents’ behavior are mainly based on Cohen and Felson’s Routine Activity Theory (1979) and Clarke’s Rational Choice Theory (1995). The model was implemented using NetLogo 3.1.4 platform (WILENSKY,1999). The simulation program provides a wide range of dynamic options making it very easy to perform any kind of test in order to assess the behavior of a given criminal rule in its dynamic operation. Validation and experimental tests were performed. The resulting responses were very consistent with the theoretical basis on which de model was based on.

O presente estudo busca desenvolver um modelo baseado em agentes para estudar em ambiente simulado baseado em sistemas computacionais as propriedades emergentes decorrentes da aplicação e execução da lei penal. Os modelos baseados em agentes têm-se revelado uma alternativa bastante promissora no estudo de sistemas complexos onde a grande quantidade de variáveis dinâmicas e interagentes tornam impossível o uso de técnicas puramente analíticas para a sua abordagem. Especialmente considerando que normas legais penais feitas para regular a conduta das pessoas em sociedade devem durar, a dimensão tempo acaba recebendo uma considerável relevância, o que incentiva ainda mais o uso de sistemas multiagentes, particularmente propícios nestes casos. Buscou-se construir um modelo no qual fosse possível analisar qualquer lei de natureza penal, quaisquer que fossem suas características e a quem estivesse destinada. Chegou-se, para tanto, no construto Cidadão- Estado-Oportunidade-Ambiente, que definem precisamente os contornos de qualquer conduta criminosa e que foram modelados em ambiente mutiagente. Para construir os fundamentos do comportamento dos agentes foram utilizados especialmente os referenciais teóricos de Shavell e Polinsky (2000, 2004), Becker (1968), Cohen e Felson (1979) e Clarke(1995), sem deixar de agregar outras contribuições teóricas, ainda que em escala menor. O construto, no entanto, permite que quaisquer contribuições e análises teóricas de atributos do comportamento dos agentes propostos possam ser utilizadas, uma vez que a sua generalidade permite que qualquer forma de conduta possa ser modelada. Para criar o simulador, foi utilizado o ambiente NetLogo 3.1.4 (WILENSKY,1999), cuja especial capacidade de tratar fenômenos que se desenvolvem ao longo do tempo se revelou de extrema valia e praticidade, demonstrando o potencial surgimento de propriedades imprevisíveis sob outras condições. Os testes realizados para aferir a aderência do modelo, usando distribuições de freqüência onde as médias e os desvios-padrão eram controlados, mostraram que o comportamento dos agentes em ambiente dinâmico segue perfeitamente os conceitos teóricos utilizados como referencial, especialmente no que diz respeito ao comportamento do agente cidadão, modelado segundo os conceitos estabelecidos por Becker (1968). Restou demonstrado pela simulação que a percepção de possibilidade de punição tem impacto considerável no comportamento dos agentes diante da possibilidade de delito, assim como o somatório dos custos envolvidos é grandemente influenciado pelo tipo de delito cometido, ou seja, a dimensão monetária do dano, pela estrutura de custos da justiça e pelos meios de arrecadação destes recursos utilizados pelo estado, conforme afirmam Shavell e Polinsky(2000). Para realizar os testes foram construídas tabelas de valores utilizando a técnica 2k fatorial, descrita em detalhes em Law e Kelton (1982), onde cada um dos fatores de impacto foi desdobrado em valores dinâmicos compatíveis com o sistema real no qual o modelo foi baseado. O teste 2k fatorial é especialmente prático no sentido de aferir a aderência do modelo de simulação aos referenciais teóricos, pois permite grande flexibilidade ao analista na formulação de cenários, pois quaisquer combinações possíveis podem ser testadas, dependendo apenas do foco de interesse da pesquisa. Foram realizados também experimentos de simulação com o objetivo de verificar o comportamento do modelo ao atingir estado estacionário. Os testes de validação e os experimentos de simulação apresentaram resultados semelhantes. / Agent based models are considered a milestone in social sciences research mainly because it´s random behavior allows a variety of analysis that were impossible to make due to complexity constraints. The term complexity usually derives from a large number of interacting and dynamic variables which cannot be considered singularly, in order to keep the whole system attributes intact. For that matter, a systemic approach of analysis is required not only to keep this information within the framework of the system, but also to get the emergent properties that come along with the interaction, not observable singling out one or another factor. Considering these very specific characteristics an agent based model, or multiagent system turns to be a more suitable analysis than analytic counterparts due to its very particular capability to deal simultaneously with a large number of random variables. This research work develops an agent based model to study the emergent properties due to public enforcement of criminal law. In order to build a model of criminal behavior suitable to the project, well known theories about crime were taken in consideration, including Shavell and Polinsky (2000), Shavell (2004), Becker (1968), Cohen and Felson(1979) and Clarke (1995). The construct derived from the analysis is a model based on three interacting agents and the environment, namely: citizen-opportunity-state-environment (offender-victim-public enforcement of law-environment). These interacting agents have the ability to define the boundaries of any given offence, regardless the kind, being capable to cope with from a simple misdemeanor (e.g. a traffic violation) to a felony (e.g. a capital murder). Particularly, the citizen-agent behavior has been modeled according to Becker’s (1968) Crime and Punishment seminal work. These remarks explains why originally the term citizen has been used in first place, and not offender. Rational Choice theory says that anyone can be an offender, depending on certain conditions. The State performance regarding its economic impact in terms of the public enforcement of criminal law is seen through the lenses of Shavell and Polinsky’s theory (2000). The opportunity and environment agents’ behavior are mainly based on Cohen and Felson’s Routine Activity Theory (1979) and Clarke’s Rational Choice Theory (1995). The model was implemented using NetLogo 3.1.4 platform (WILENSKY,1999). The simulation program provides a wide range of dynamic options making it very easy to perform any kind of test in order to assess the behavior of a given criminal rule in its dynamic operation. Validation and experimental tests were performed. The resulting responses were very consistent with the theoretical basis on which de model was based on.

O presente estudo busca desenvolver um modelo baseado em agentes para estudar em ambiente simulado baseado em sistemas computacionais as propriedades emergentes decorrentes da aplicação e execução da lei penal. Os modelos baseados em agentes têm-se revelado uma alternativa bastante promissora no estudo de sistemas complexos onde a grande quantidade de variáveis dinâmicas e interagentes tornam impossível o uso de técnicas puramente analíticas para a sua abordagem. Especialmente considerando que normas legais penais feitas para regular a conduta das pessoas em sociedade devem durar, a dimensão tempo acaba recebendo uma considerável relevância, o que incentiva ainda mais o uso de sistemas multiagentes, particularmente propícios nestes casos. Buscou-se construir um modelo no qual fosse possível analisar qualquer lei de natureza penal, quaisquer que fossem suas características e a quem estivesse destinada. Chegou-se, para tanto, no construto Cidadão- Estado-Oportunidade-Ambiente, que definem precisamente os contornos de qualquer conduta criminosa e que foram modelados em ambiente mutiagente. Para construir os fundamentos do comportamento dos agentes foram utilizados especialmente os referenciais teóricos de Shavell e Polinsky (2000, 2004), Becker (1968), Cohen e Felson (1979) e Clarke(1995), sem deixar de agregar outras contribuições teóricas, ainda que em escala menor. O construto, no entanto, permite que quaisquer contribuições e análises teóricas de atributos do comportamento dos agentes propostos possam ser utilizadas, uma vez que a sua generalidade permite que qualquer forma de conduta possa ser modelada. Para criar o simulador, foi utilizado o ambiente NetLogo 3.1.4 (WILENSKY,1999), cuja especial capacidade de tratar fenômenos que se desenvolvem ao longo do tempo se revelou de extrema valia e praticidade, demonstrando o potencial surgimento de propriedades imprevisíveis sob outras condições. Os testes realizados para aferir a aderência do modelo, usando distribuições de freqüência onde as médias e os desvios-padrão eram controlados, mostraram que o comportamento dos agentes em ambiente dinâmico segue perfeitamente os conceitos teóricos utilizados como referencial, especialmente no que diz respeito ao comportamento do agente cidadão, modelado segundo os conceitos estabelecidos por Becker (1968). Restou demonstrado pela simulação que a percepção de possibilidade de punição tem impacto considerável no comportamento dos agentes diante da possibilidade de delito, assim como o somatório dos custos envolvidos é grandemente influenciado pelo tipo de delito cometido, ou seja, a dimensão monetária do dano, pela estrutura de custos da justiça e pelos meios de arrecadação destes recursos utilizados pelo estado, conforme afirmam Shavell e Polinsky(2000). Para realizar os testes foram construídas tabelas de valores utilizando a técnica 2k fatorial, descrita em detalhes em Law e Kelton (1982), onde cada um dos fatores de impacto foi desdobrado em valores dinâmicos compatíveis com o sistema real no qual o modelo foi baseado. O teste 2k fatorial é especialmente prático no sentido de aferir a aderência do modelo de simulação aos referenciais teóricos, pois permite grande flexibilidade ao analista na formulação de cenários, pois quaisquer combinações possíveis podem ser testadas, dependendo apenas do foco de interesse da pesquisa. Foram realizados também experimentos de simulação com o objetivo de verificar o comportamento do modelo ao atingir estado estacionário. Os testes de validação e os experimentos de simulação apresentaram resultados semelhantes. / Agent based models are considered a milestone in social sciences research mainly because it´s random behavior allows a variety of analysis that were impossible to make due to complexity constraints. The term complexity usually derives from a large number of interacting and dynamic variables which cannot be considered singularly, in order to keep the whole system attributes intact. For that matter, a systemic approach of analysis is required not only to keep this information within the framework of the system, but also to get the emergent properties that come along with the interaction, not observable singling out one or another factor. Considering these very specific characteristics an agent based model, or multiagent system turns to be a more suitable analysis than analytic counterparts due to its very particular capability to deal simultaneously with a large number of random variables. This research work develops an agent based model to study the emergent properties due to public enforcement of criminal law. In order to build a model of criminal behavior suitable to the project, well known theories about crime were taken in consideration, including Shavell and Polinsky (2000), Shavell (2004), Becker (1968), Cohen and Felson(1979) and Clarke (1995). The construct derived from the analysis is a model based on three interacting agents and the environment, namely: citizen-opportunity-state-environment (offender-victim-public enforcement of law-environment). These interacting agents have the ability to define the boundaries of any given offence, regardless the kind, being capable to cope with from a simple misdemeanor (e.g. a traffic violation) to a felony (e.g. a capital murder). Particularly, the citizen-agent behavior has been modeled according to Becker’s (1968) Crime and Punishment seminal work. These remarks explains why originally the term citizen has been used in first place, and not offender. Rational Choice theory says that anyone can be an offender, depending on certain conditions. The State performance regarding its economic impact in terms of the public enforcement of criminal law is seen through the lenses of Shavell and Polinsky’s theory (2000). The opportunity and environment agents’ behavior are mainly based on Cohen and Felson’s Routine Activity Theory (1979) and Clarke’s Rational Choice Theory (1995). The model was implemented using NetLogo 3.1.4 platform (WILENSKY,1999). The simulation program provides a wide range of dynamic options making it very easy to perform any kind of test in order to assess the behavior of a given criminal rule in its dynamic operation. Validation and experimental tests were performed. The resulting responses were very consistent with the theoretical basis on which de model was based on.

Cette thèse porte sur l’étude des inégalités dans un cadre macroéconomique, d’un point de vue théorique ainsi qu’empirique. Dans un premier chapitre, j’écris et simule un modèle basé-agents capable de répliquer les distributions fat-tailed des richesses observées empiriquement dans les économies développées. Dans un second chapitre, je prolonge ce modèle théorique pour étudier l’impact économique des discriminations interpersonnelles et institutionnelles. Lorsque les discriminations institutionnelles cessent, l’état final des inégalités dépend de l’économie au moment de la fin de ces discriminations : plus l’économie est organisée, plus le temps nécessaire à une égalisation des revenus et richesses entre ethnies est long, voire infini. Dans un troisième chapitre empirique, j’étudie l’évolution des inégalités de salaire entre Noirs et Blancs aux États-Unis entre 1960 et 2015, en me concentrant sur la période 2000-2015. Je traite les biais de sélection liés à l’asymétrie raciale envers l’emprisonnement, et montre que l’écart - en conditionnant par l'âge et les diplômes - entre salaires médians des Noirs et des Blancs atteint un maximum en 2012. En utilisant la méthode de régression quantiles non-conditionnelles conjointement avec la décomposition de Blinder-Oaxaca, j’établis que la part non-expliquée de cet écart reste stable durant la Grande Récession. Enfin, dans une post-face, en utilisant les métadonnées de RePEC, j’établis que la part des articles scientifiques consacrés à l’étude des crises augmente significativement après 2008 pour 13 des 30 « top journals » en économie. / This PhD dissertation focuses on wealth and wage inequality, and the macro-economy. In a first chapter, I write and run a small macro agent-based model (M-ABM) in which I study the resultant distribution of wealth among households. I show that this model generates fat- tailed distributions of wealth in the household sector, as empirically observed in advanced economies. In a second chapter, I extend this model to study the macroeconomics of interpersonal and institutional discriminations against racial minorities. When discrimination is at work, racial disparities in income and wealth arise. The effect of the abolition of institutional discrimination is path-dependant: the more the economy is organized when this institutional change occurs, the more time it takes to get back to the counter-factual situation where no institutional discrimination was set up in the first place. In a third chapter, I study the evolution of the difference of median log-annual earnings between Blacks and Whites in the US between 1960 and 2015, focusing on the 2008 crisis. I control for selection arising from racial differentials in institutionalised population, and find that the unconditional racial wage gap attains a maximum in 2012. Controlling for age and education, I obtain the same result. Using unconditional quantile regressions, I show that the unexplained part of the unconditional racial wage gap has not increased during the crisis. Finally, in an afterword, I use metadata from RePEC to show that the share of economics papers published in the 13 of the 30 "top" journals containing "crisis" in their titles and/or abstracts has significantly increased in 2008.

L'obbiettivo di questa tesi è di investigare il ruolo della formazione delle aspettative nei modelli macroeconomici basati su agenti e stock-flussi coerenti. Mentre ci sono stati notevoli passi avanti nello sviluppo di tali modelli, la ricerca sulla formazione e sul ruolo delle aspettative in essi rimane ancora poco sviluppato. La tesi è composta da tre articoli, ognuno dei quali si focalizza sulla formazione delle aspettative in un settore economico speci co e presenta una serie di esperimenti riguardanti la variazione dei meccanismi di formazione delle aspetta- tive, dinamiche di opinioni e sentimenti, così come le applicazioni delle politiche economiche. La tesi dimostra l'influenza potenzialmente forte delle aspettative dei agenti sulla volatilità macroeconomica e mostra che, dipendendo dalla loro speci cazione e l'ambiente economico, le aspettative possono essere sia un elemento stabilizzante che un elemento destabilizzante. Inoltre, attraverso l'ampia gamma di esperimenti politici condotti, serve a sottolineare il ruolo importante delle politiche stabilizzanti nei sistemi che esibiscono fluttuazioni endogene, e il capitolo 4 in particolare mette in evidenza la potenziale dipendenza dell'efficacia delle politiche economiche dalle aspettative. Allo stesso tempo, alcuni dei risultati ottenuti avvertono che nei sistemi complessi, gli interventi politici devono essere calibrati attentamente affinché non diventino essi stessi fonte di instabilità. / The purpose of this thesis is to investigate the role of expectations formation in agent-based and stock-flow consistent macroeconomic models. While there have been considerable advances in the development of such models, research on the formation and role of beliefs and expectations within them remains underdeveloped. The thesis consists of three papers, each of which focuses on expectations formation in one particular economic sector and presents a range of experiments concerning the variation of expectations formation mechanisms, belief and sentiment dynamics, as well as policy applications. The thesis demonstrates the potentially strong influence of agents' expectations on macroeconomic volatility and shows that depending on their specification as well as the economic environment, expectations can be both a stabilising and destabilising factor. Moreover, through the wide range of policy experiments conducted, it serves to emphasise the important role of stabilisation policies in systems exhibiting endogenous fluctuations and chapter 4 in particular highlights the potential dependence of policy effectiveness on expectations. At the same time, some of the obtained results caution that in complex systems, policy interventions must be carefully calibrated lest they themselves become a source of instability.

This thesis focuses on the link between financial system and sovereign debt crises through sovereign support to banks on one hand and banks' exposures to weak sovereigns on the other. After illustrating the main relationships on the recent financial crisis, we construct an agent-based network model of an artificial financial system allowing us to analyse the effects of state support on systemic stability and the feedback loops of risk transfer back into the financial system. First, the model is tested with various parameter settings in Monte Carlo simulations and second, it is calibrated to the real world data using a unique dataset put together from various sources. Our analyses yield the following key results: Firstly, in the short term, all the support measures improve the systemic stability. Secondly, in the longer run, the effects of state support depend on several parameters but still there are settings in which it significantly mitigates the systemic crisis. Finally, there are differences among the effects of the different types of support measures.

Die folgende Arbeit befasst sich mit den Untersuchungen von Problemen der Optimierung von Ressourcen in Umgebungen mit unvorhersehbarem Verhalten, wo: (i) nicht alle Informationen verfügbar sind, und (ii) die Umgebung unbekannte zeitliche Veränderungen aufweist. Diese Dissertation ist folgendermaßen gegliedert: Teil I stellt das Investitionsmodell vor. Es wird sowohl eine analytische als auch eine numerische Analyse der Dynamik dieses Modells für feste Investitionsstrategien in verschiedenen zufälligen Umgebungen vorgestellt. In diesem Investitionsmodell hängt die Dynamik des Budgets des Agenten x(t) von der Zufälligkeit der exogenen Rendite r(t) ab, wofür verschiedene Annahmen diskutiert wurden. Die Heavy-tailed Verteilung des Budgets wurde numerisch untersucht und mit theoretischen Vorhersagen verglichen. In Teil II wurde ein Investitionsszenario mit stilisierten exogenen Renditen untersucht, das durch eine periodische Funktion mit verschiedenen Arten und Stärken von Rauschen charakterisiert ist. In diesem Szenario wurden unterschiedliche Strategien, Agenten-Verhalten und Agenten Fähigkeiten zur Vorhersage der zukünftigen r(t) untersucht. Hier wurden Null-intelligenz-Agenten, die über technischen Analysen verfügen, mit Agenten, die über genetischen Algorithmen verfügen, verglichen. Umfangreiche Ergebnisse von Computersimulationen wurden präsentiert, in denen nachgewiesen wurde, dass für exogene Renditen mit Periodizität: (i) das wagemutige das vorsichtige Verhalten überbietet, und (ii) die genetischen Algorithmen in der Lage sind, die optimalen Investitionsstrategien zu finden und deshalb die anderen Strategien überbieten. Obwohl der Schwerpunkt dieser Dissertation im Zusammenhang mit dem Gebiet der Informatik präsentiert wurde, können die hier vorgestellten Ergebnisse auch in Szenarien angewendet werden, in denen der Agent anderere Arten von Ressourcen steuern muss, wie z.B. Energie, Zeitverbrauch, erwartete Lebensdauer, etc. / The main goal of this PhD thesis is to investigate some of the problems related to optimization of resources in environments with unpredictable behavior where: (i) not all information is available and (ii) the environment presents unknown temporal changes. The investigations in this PhD thesis are divided in two parts: Part I presents the investment model and some analytical as well as numerical analysis of the dynamics of this model for fixed investment strategies in different random environments. In this investment model, the dynamics of the investor''s budget x(t) depend on the stochasticity of the exogenous return on investment r(t) for which different model assumptions are discussed. The fat-tail distribution of the budget is investigated numerically and compared with theoretical predictions. Part II investigates an investment scenario with stylized exogenous returns characterized by a periodic function with different types and levels of noise. In this scenario, different strategies, agent''s behaviors and agent''s capacities to predict the future r(t) are investigated. Here, ''zero-intelligent'' agents using technical analysis (such as moving least squares) are compared with agents using genetic algorithms to predict r(t). Results are presented for extensive computer simulations, which shows that for exogenous returns with periodicity: (i) the daring behavior outperforms the cautious behavior and (ii) the genetic algorithm is able to find the optimal investment strategy by itself, thus outperforming the other strategies considered. Finally, the investment model is extended to include the formation of common investment projects between agents. Although the main focus of this PhD thesis is more related to the area of computer science, the results presented here can be also applied to scenarios where the agent has to control other kinds of resources, such as energy, time consumption, expected life time, etc.

The thesis focuses on banking regulation and on the nexus between financial sovereign crises. After illustrating the main mechanisms on the recent financial crisis, we construct several multi-agent network models of a financial system for testing its stability under different parameters. In the first part, we focus on the rationale for banking regulation and we describe its development including the recently introduced Basel III measures. The main conclusion of this part is that regulation is to a large extent influenced by the banks and it does not always secure financial system stability. In the second part, we build an agent-based model which enables us to simulate the impacts of various types of negative shocks given various settings of the banking system and the regulatory environment, including the capital and liquidity measures. Our simulations show firstly that sufficient capital buffers are crucial for systemic stability, secondly that the discretionary measures have little effect once a crisis breaks out and thirdly that liquidity measures are a relevant regulatory tool. In the third part, the model is extended so that it allows for testing effects of state support on systemic stability is tested with various parameter settings in Monte Carlo simulations and for testing of feedback loops in which...