[Next] [Previous] [Up] [Top] [Contents]

3 The effects of modelling by agents on complexity

3.2 Including the process of modelling by economic agents


The other case is where the conditions described above do not hold, i.e. where the process whereby the agent models its environment itself has significant effects. That such cases may exist in real life is surely not in doubt (e.g. in emerging markets, during periods of great volatility in the stock market or the behaviour of economists themselves). What is debateable is:

  1. the extent of these situations;

  2. the tractability of them;

  3. whether such an area of study can be said to be part of the tradition of economics as it has developed.

Although point (1) is the most important, to be decided by empirical evidence, I do not have either the knowledge or the time to consider this. I will only comment that since the assumption that one does not loose anything important by conflating the agents model with the true model is a very strong one, the burden of evidence should weigh more heavily on it as opposed to its converse.

Point (2) is a valid point, and one which I am highlighting here in this paper. The introduction of including processes of modelling by agents in our models, greatly increases the behavioural complexity of the situation. In fact such an introduction seems to introduce a qualitative jump in the whole enterprise of such modelling*1. On the other hand it does seem, that at least some aspects of such situations are meaningfully formalisable and analysable (for example both [1] and [5] exhibit credible models of economic systems which exhibit both rational expectations and more complex dynamics in clearly defined circumstances).

Point (3) is in many ways trivial. The qualitative leap highlighted above may justify categorising such modelling under a separate heading, but unless such models turn out to be significantly less successful than more traditional ones, this would not justify this not being a substantial and legitimate line of research.


From Complexity to Agent Modelling and Back Again - Bruce Edmonds - 15 MAY 97
[Next] [Previous] [Up] [Top] [Contents]

Generated with CERN WebMaker