Drawing a conclusion from recent insights in evolutionary game theory, we show that a so-called spite e!ect implies that there is an essential di!erence between individual and social learning. We illustrate its consequences for the choice of computational tools in economics and social settings in general by analyzing two variants of a Genetic Algorithm.

Abstract:Schelling [1969, 1971a, 1971b, 1978] presented a microeconomic model showing how an integrated city could unravel to a rather segregated city, notwithstanding relatively mild assumptions concerning the individual agents' preferences, i.e., no agent preferring the resulting segregation. We examine the robustness of Schelling's model, focusing in particular on its driving force: the individual preferences. We show that even if all individual agents have a strict preference for perfect integration, bestresponse dynamics will lead to segregation. What is more, we argue that the onedimensional and two-dimensional versions of Schelling's spatial proximity model are in fact two qualitatively very different models of segregation.J.E.L. classification codes: C72, C73, D62.

We present an agent-based computational economics (ACE) model of the wholesale "sh market in Marseille. Two of the stylized facts of that market are high loyalty of buyers to sellers, and persistent price dispersion, although it is every day the same population of sellers and buyers that meets in the same market hall. In our ACE model, sellers decide on quantities to supply, prices to ask, and how to treat loyal customers, while buyers decide which sellers to visit, and which prices to accept. Learning takes place through reinforcement. The model explains both stylized facts price dispersion and high loyalty. In a coevolutionary process, buyers learn to become loyal as sellers learn to o!er higher utility to loyal buyers, while these sellers, in turn, learn to o!er higher utility to loyal buyers as they happen to realize higher gross revenues from loyal buyers. The model also explains the e!ect of heterogeneity of the buyers. We analyze how this leads to subtle di!erences in the shopping patterns of the di!erent types of buyers, and how this is For example, according to Royal Mail (1999)`more than 25 percent of all people do not consider themselves loyal to any businessa, implying that about 75% is loyal to some business. related to the behavior of the sellers in the market.

In an experimental standard Cournot oligopoly we test the importance of models of behaviour characterised by imitation of successful behaviour, in particular when the environment becomes more complex. We find that the players do not rely more on imitation in more demanding environments. We explain that the different pattern of output decisions in such environments seems predominantly related to a general disorientation of the players, and more specifically to a significant decrease of best-responses.Imitation has played an important role in the recent literature on learning and adaptive behaviour in economics.1 Imitation seems easy and straightforward, it does not use particularly many cognitive skills and it does not require much information. Imitation certainly qualifies as a 'fast and frugal heuristics for making decisions ' (Gigerenzer and Goldstein, 1996). Not surprisingly, conventional wisdom asserts that imitation of the successful behaviour of others is likely in situations with little information or understanding. Yet imitation, although deemed easy, need not be a realistic mode of behaviour. To assess the extent of imitative behaviour, we set up an experimental symmetric Cournot market under different information treatments, some of them very conducive to imitation. The choice of a symmetric Cournot market is not accidental. First, recent theoretical results, e.g., Vega-Redondo (1997) underpin the basic qualitative insight dating back to Schaffer (1988aSchaffer ( , 1988b) that in Cournot markets in which firms (loosely) imitate the more successful firms, there is a tendency for output levels to drift to the Walrasian competitive equilibrium. This implies that we can use the market outcomes as a first proxy for the possible presence of imitation. Second, any * We wish to thank Xavier Herrero for his indispensable assistance in organising the experiments, and

Schelling (Schelling, T., 1969. Models of segregation. American Economic Review 59, 488–493; Schelling, T., 1971a. Dynamic models of segregation. Journal of Mathematical Sociology 1, 143–186; Schelling, T., 1971b. On the ecology of micromotives. The Public Interest 25, 61–98; Schelling, T., 1978. Micromotives and Macrobehavior. W.W. Norton and Company, New York) considered a model with individual agents who only care about the types of people living in their own local neighborhood. The spatial structure was represented by a one- or two-dimensional lattice. Schelling showed that an integrated society will generally unravel into a rather segregated one even though no individual agent strictly prefers this. We generalize this spatial proximity model to a proximity model of segregation, examining models with individual agents who interact 'locally' in a range of more general social network structures. The levels of segregation attained are in line with those reached in the lattice-based spatial proximity model