Weitere Kapitel dieses Buchs durch Wischen aufrufen
The illuminating prose of Friedrich von Hayek illustrates some of the main issues that are at the core of the bottom-up approach to macroeconomics we offer in this book. The inhabitants of the realistic economies we aim to model form expectations and take actions building on the asymmetric and incomplete information they acquire by exploring limited portions of space and time, while their dispersed market transactions generate aggregate outcomes whose welfare properties are unknowable in principle, at least if one pretends to measure them against some hypothetical Walrasian general equilibrium. The quotation has also a second value added in that it helps us to stress once again that this vision, though considered as heretical by the mainstream, does not represent anything particularly new from a theoretical viewpoint. On the contrary, it is part and parcel of a well-honored but guiltily disregarded tradition in the history of economic thought, one that considers the economic agent as a proper human being instead of a computer-like automaton. We argue that it is time not only to bring this tradition back to life, but also to revamp it by means of new insights from other behavioral sciences, like cognitive psychology and social biology.
Bitte loggen Sie sich ein, um Zugang zu diesem Inhalt zu erhalten
Sie möchten Zugang zu diesem Inhalt erhalten? Dann informieren Sie sich jetzt über unsere Produkte:
The authors of this book have all learnt to drive a car in a continental European country. The outcome would be dramatically different if at least one of the two drivers in our example would come from the UK
This is of course the simplest conceivable setting in the simplest case of a closed economy without public sector. Also in this setting, there are richer models which consider also other classes of agents, such as banks, investors on financial markets, etc. In any case, a relatively small (very small) number of classes are considered. Usually, the number and type of classes is a prior of the modeler.
See McFazdean and Tesfatsion (1999) for an introduction to object-oriented programming into economics, and an application to the endogenous formation of trading networks.
Of course there can be many agents in any generation, but they are ultimately uniform. Each young (old) is a clone of any other young (old). Therefore, despite the appearance of a very large number of agent, the model boils down to only two of them.
In what follows, underscores denote vectors.
For a complete analysis of these issues, see Evans and Honkapoija (2004).
Notice that on this particular modeling choice the Keynesian and Neoclassical frames of thought are indistinguishable. The differences are the laws supposed to underlie the behavior of the means: in fact, according to Keynesians it is sufficient to study a “working” relationship between, for example, aggregate consumption and aggregate income; while, according to the Neoclassical reductionist approach, the mean outcomes have to be derived as if they were the optimal responses by a representative consumer or firm.
One must be honest: the self-averaging property is only a sufficient condition in order to rely upon mean values and, consequently, non-self-averaging does not imply their discard. However, non-self-averaging is sufficient to strongly distrust the neoclassical justification for the use of mean values and to open the path to AB models.
By within and between class interactions we always mean non-mean-field, one-to-one interactions.
Notice, however, that more sophisticated techniques for the choice of parameter values are also available ( Bianchi et al., 2008).
Validation of ABMs is becoming one of the major points in the agenda of agent-based researchers. In her website, Leigh Tesfatsion maintains an entire section dedicated to this topic (http://www.econ.iastate.edu/tesfatsi/empvalid.htm).
See footnote 10.
Admittedly, the logical order between the validation and the calibration stages we propose is not universally accepted among the agent-based community. Richiardi et al. (2006), for instance, argue for putting calibration before validation, as they consider validation as the final step of a well-calibrated simulation model.
- The Making of the BAM Model
Domenico Delli Gatti
- Springer Milan
- Chapter 2
Neuer Inhalt/© Stellmach, Neuer Inhalt/© Maturus, Pluta Logo/© Pluta, Frankfurt School