Catálogo de publicaciones - libros

Population learning in dynamic economies with endogenous network formation has been traditionally studied in basic settings where agents face quite simple and predictable strategic situations (e.g. coordination). In this paper, we begin instead to explore economies where the overall payoff landscape is very complicated (rugged). We propose a model where the payoff of any agent changes in an unpredictable way as soon as any small variation in the strategy configuration within her network occurs. We study population learning when agents: (i) are allowed to periodically adjust both the strategy they play in the game and their interaction network; (ii) employ some simple criteria (e.g. statistics such as MIN, MAX, MEAN, etc.) to myopically form expectations about their payoff under alternative strategy and network configurations. Computer simulations show that: (i) allowing for endogenous networks implies higher average payoff as compared to ”static” networks; (ii) populations learn by employing network updating as a ”global learning” device, while strategy updating is used to perform ”fine tuning”; (iii) the statistics employed to evaluate payoffs strongly affect the efficiency of the system, i.e. convergence to a unique (multiple) steady-state(s); (iv) for some class of statistics (e.g. MIN or MAX), the likelihood of efficient population learning strongly depends on whether agents are change-averse or not in discriminating between options delivering the same expected payoff.

In this paper we analyse a growth model where agents have different factor endowments and form coalitions to produce output. Economic growth is the result of the accumulation of human capital. The latter in turn is a by-product of the production activity within a coalition. The maximum rate of growth is obtained when the grand coalition forms. However, if endowments are heterogeneous and the rule governing the division of the coalitional output states an equal sharing among the members of a coalition, agents with better endowments may not be willing to coalesce with poorly endowed agents. Indeed richer agents tend to form coalitions among themselves and the poor ones cannot benefit of the positive externalities of coalescing with the richest agents. This determines both a lower output and a lower long-run growth rate.

We study the statistical properties of the network of shareholding relationships in the Italian stock market (MIB) and in two US stock markets (NYSE and NASDAQ). The networks are found to be scale free a feature which doesn't seem to be in accord with classical theories of portfolio optimization. Several statistical properties differ across markets and allow for a classification. We introduce two quantities, HI and SI, analogous to in-degree and out-degree for weighted graphs. The distribution of HI and SI allow us to characterize from a statistical perspective the individual ownership concentration of stocks and the individual power of holders. They also suggest two different global pictures of the structure of the networks: the MIB seems characterized by medium size holders controlling separate subsets of stocks, while the US markets seem characterized by very large holders sharing control over subsets of stocks.

This paper adds a labor sector to the model of output fluctuations of Aoki (2002, Chapt. 8), and Aoki and Yoshikawa (2003) to produce a new model with a labor sector.

The concept of ultrametrics is introduced to the labor dynamics of this paper to measure ”distances” between clusters of unemployed workers in different geographical locations, with work experiences, and/or human capitals to reflect differences in probabilities of them being rehired from a pool of unemployed workers. We maintain the assumption of our previous model that marginal product of labor does not equalize instantaneously across sectors of our economy.

This model shares the same property with our earlier one that GDP responds to changes in demand patterns among the sectors. In addition, the model of this paper exhibits a relation between unemployment rates and GDP similar to that of the Okuns' law in its business cycle fluctuations. This Okun's coefficient increases as the average GDP increases. Our model also reaches stationary business cycles faster as more demands are put on more productive sectors of the model.

The clustering of agents in the market is a typical problem dealt with by the new approaches to macroeconomic modeling, describing macroscopic variables in terms of the behavior of a large collection of microeconomic entities. Clustering has many economic interpretations [3], that are often described by Ewens' Sampling Formula (ESF). This formula can be traced back to Fisher as “species sampling”, and its main use was restricted to genetics for a long time. Contrary to the usual complex derivations [18], we suggest a finitary characterization of the ESF pointing to real economic processes. Our approach is finitary in the sense that we provide a probabilistic characterisation of a system of individuals considered as a closed system, a population, where individuals can change attributes as time moves on. The intuitive meaning of the probability is the fraction of time the system spends in the considered partition. As ESF represents an equilibrium distribution satisfying detailed balance, some properties which are otherwise difficult to prove are derived in a simple way.

This paper investigates the statistical properties of , defined as the absolute value of the logarithmic price change, for the Nikkei 225 index in the 28-year period from January 4, 1975 to December 30, 2002. We divided the time series of the Nikkei 225 index into two periods, an inflationary period and a deflationary period. We have previously [18] found that the distribution of absolute log-returns can be approximated by the power-law distribution in the inflationary period, while the distribution of absolute log-returns is well described by the exponential distribution in the deflationary period.

To explain these empirical findings, we have introduced a model of stock market dynamics [19,20]. In this model, the stock market is composed of two groups of traders: , who believe that the asset price will return to the fundamental price, and , who can be noise traders. We show through numerical simulation of the model that when the number of interacting traders is greater than the number of fundamentalists, the power-law distribution of absolute log-returns emerges from the interacting traders' herd behavior, and, vice-versa, when the number of fundamentalists is greater than the number of interacting traders, absolute log-returns are characterised by an exponential distribution.

A discrete-time dynamic model of a financial market is developed, where two types of agents, and , allocate their wealth between two risky assets and a safe asset, according to one-period mean-variance maximization. The two groups of agents form different expectations about asset returns and their variance/covariance structure, and this results in different demand functions. At the end of each trading period, agents' demands are aggregated by a , who sets the next period prices as functions of the excess demand. The model results in a high-dimensional nonlinear discrete-time dynamical system, which describes the time evolution of prices and agents' beliefs about expected returns, variances and correlation. It is shown that the unique steady state may become unstable through a Hopf-bifurcation and that an attracting limit cycle, or more complex attractors, exist for particular ranges of the key parameters. In particular, the two risky assets may exhibit “coupled” long-run price fluctuations and time-varying correlation of returns.

The characterisation of agents' preferences by decreasing absolute risk aversion (DARA) and constant relative risk aversion (CRRA) are well documented in the literature and also supported in both empirical and experimental studies. This paper considers a financial market with heterogeneous agents having power utility functions, which are the only utility functions displaying both DARA and CRRA. By introducing a population weighted average wealth measure, we develop an adaptive model to characterise asset price dynamics as well as the evolution of population proportions and wealth dynamics. Some numerical simulations are included to illustrate the evolution of the wealth dynamics, market behaviour and market efficiency within the framework of heterogeneous agents.

In competitive coevolution, the Red Queen principle entails constraints on performance enhancement of all individuals if each is to maintain in relative fitness measured by an index relating to aggregate performance. This is encapsulated in Lewis Caroll's Red Queen who says ”in this place it takes all the running you can do, to keep in the same place”. The substantive focus of this paper is to experimentally generate stock market ecologies reflecting the Red Queen principle for an explanation of the observed highly inegalitarian power law distribution in investor income (measured here as stock holdings) and the emergence of arbitrage free conditions called market efficiency. With speculative investors modelled as using genetic programs (GPs) to evolve successful investment strategies, the analytical statement of our hypothesis on the Red Queen principle can be implemented by constraint enhanced GPs which was seminally developed in [19], [7] and [10].

In this paper, we present an extension of the Genoa artificial stock market (GASM) (Raberto et al., 2001) that includes a limit order book as mechanism for price formation. At every time step an agent is chosen with uniform distribution to issue an order. The order can be a limit order or a market order. If the order is a limit order, it is stored in the book; if the order is a market order, a transaction occurs. Prices are formed at variable time steps, i.e., only when a market order is issued. We investigate how the new asynchronous trading mechanism affects the statistical properties of simulated prices. This computational experiment shows that the fat tails of the returns distribution can be recovered simply as a consequence of the limit order book without any additional assumption on agents' behavior.