Classical microeconomic theory often relies on the assumption that it is possible to analyze economic relationships between a small number of rational agents - say one buyer and one seller - and then to extend the results to a larger group, of which the subset studied is representative. However many socio-economic phenomena emerge as a consequence of a large number of locally interacting heterogeneous agents. Indeed, opening the black box of the representative agent and explicitly modeling these localized interactions represents a challenging, but active, direction of the current research. The main focus of my thesis is the study of situations involving large number of agents, each owning a piece of information (an opinion, an hard knowledge or a preference) that can be shared with a limited number of their peers in the population. While the dynamical properties of distributed systems are difficult to study analytically, other methods (such as network theory, fractal theory, agent based modeling, and optimization) can give useful insights. Indeed, while the topics addressed in my thesis are diverse, the toolboxes used to investigate them are similar. In presence of decentralized interactions, the networks, where agents are represented by nodes and relationships between them by links, are powerful tools of analysis. In the first two chapters, I use them to show how, with simple but effective theoretical modeling, it is possible to generate outcomes that share some of the characteristics of real world social networks. On the contrary, the second and third chapter, using agent based modelling, focus on the dynamic characteristics of opinion exchanges when heterogeneous agents are given the opportunity to share, with some of their peers, their points of view (opinions and price expectations respectively). Finally, in the last chapter, I deal with a classical problem of information aggregation addressing the relationship between rent extraction and electoral campaign in a probabilistic voting model. More in detail, in the first chapter of the thesis (written with Timoteo Carletti and published on Physica A), I study a particular kind of weighted networks: those builded inside fractals. A fractal is a mathematical object whose Hausdorff dimension exceeds the topological one. Building networks inside these objects generates tree-like structures that share the self similarity property with them. I show how some topological features (average degree of connectivity, average shortest path, node strength distribution and average clustering coefficient) of these networks, can be analytically char- acterized in terms of the parameters typical of a fractal object (the scale factor and the number of copies). Interestingly, I find that the weighted fractal networks exhibit the small world property, typical of many empirical social networks. Indeed, as the graphs grow, their average clustering coefficient do not vanish and the average shortest path grow very slowly. Although this is a theoretical paper, its results could be used to study the dissemination of information in hierarchical networks (such as those that characterize the organizational charts of large, fordists, companies) where the same network structure (e.g., an office man- ager receive information from the higher levels of authority and share part of it with his colleagues) is repeated through different levels with losses of information along the way. Although empirical social networks have been studied for many years, only recently it was shown that they almost universally share some fundamental features. The most important of these are the small world property (most nodes are not directly neighbors of one another, but most nodes can be reached from every other by a small number of steps) and the presence of weak ties (some links are rarely activated, but are critical in the dissemination of information). In the second chapter (written with Duccio Fanelli and Timoteo Carletti, and published on Advances in Complex Systems), I show that these characteristics spontaneously emerge in a social network of interacting agents. In particular, this dynamics is simulated in the framework of a simplified model of opinion diffusion in an evolving social network, where agents are made to interact, and possibly update, their beliefs and modify their social relationships according to the results of past opinions exchanges. This type of model is then applied applied to the study of market interactions. In the third chapter of the thesis (written with Gani Aldashev and Timoteo Carletti, and published on Journal of Economic Behaviour and Organization), I study the informational efficiency of a financial market where a single asset is traded, on which the different agents have noisy private information. The traders expectations are assumed to have an adaptive components (the past price matters in the evolution of the price expectations) and a social interaction component with confirmatory bias (agents tends to disregard information that contradicts what they believe). I show that, taken separately, each of the deviations from rationality worsen the information efficiency of the market. However when the two biases are combined, the degree of informational efficiency of the market (which is measured as the deviation of the long-run market price from the fundamental value of the asset) can be non-monotonic both in the weight of the adaptive component and in the degree of the confirmatory bias. For some ranges of parameters, I find the unexpected result that the two biases tend to mitigate each other’s effect, thus increasing the informational efficiency. Finally political competition, and more generally representative democracy, where policy choices are delegated to political representatives, is considered as an efficient way of aggre- gating the preferences of social groups of large dimensions. In the fourth chapter, which is also my Job Market Paper, motivated by the presence of some stylized facts previously unac- counted for by the literature, I propose a probabilistic voting model with inefficient electoral competition and office motivated politicians. I study the relationship between rent-seeking behavior and electoral campaign in the absence of special interest groups. Candidates have the possibility to influence voters’ behavior through a costly investment (campaigning), to be deduced from their rent. I focus on two channels through which this campaigning may change the result of the elections: on one side, it increases the popularity of the candidate that outspends the opponent and, on the other side, each voter weights the expenditure of each candidate proportionally to his bias toward or against him. In this setting, I derive the equilibrium values of rent and campaign spending. I find that the latter is always positive and increases in size when voters have procedural concerns. I provide comparative statics and results that could be tested empirically. Finally, I study the welfare effects of some poli- cies aimed at reducing the inefficiencies of the electoral competition. On this regard, I show that limitations on campaign expenditures are always welfare decreasing for voters while anti-corruption policies can make the voters better off, even though they face difficulties of implementation in countries with positively skewed income distributions.
|Date of Award||14 Jun 2012|
|Supervisor||Timoteo Carletti (Co-Supervisor), Gani Aldashev (Supervisor), Guillaume Deffuant (Jury), Enrico DIECIQUE (Jury), Marc LICALZI (Jury) & Micael CASTANHEIRA (Jury)|