Computer Models in Economics
Traditional mathematical Economics is built on a framework of powerful assumptions that together enable the complexities of human trading behaviour to become analytically tractable. With these assumptions in place, it becomes possible to prove strong mathematical theorems, which have been tremendously influential in world politics and economic development (for example, in arguing the case for economic liberalism, free trade, and globalisation). The status of "mathematical proof" gives an aura of solid objectivity to these results, though in fact the assumptions from which they are derived are acknowledged to be unrealistic, for example:
- Pure self-interest
People act exclusively in their own self-interest, as assessed by the "utility" that they expect (i.e. the things they value, and how much – this allows that people can put a value on others' benefit as well as their own). The relevant utility function is, in principle, fixed, and people can reliably be expected to do whatever is in their power to maximise the value of that function.
- Perfect rationality
In deciding how to act, people calculate perfectly without any cognitive biases, and (at least in most standard models) take all relevant factors into account, including probabilities of events and future economic conditions, rational predictions of others' behaviour based on perfect current information, outcomes of relevant game-theoretic calculations etc.
- Pure market
People interact exclusively through market mechanisms, notably "auctions" in which the price of goods is settled through a precise balance of supply and demand amongst all relevant participants.
- Closed equilibrium system
The economy can be modelled as a closed system from which radical novelty or external intrusions are excluded, and in which economic behaviour can be seen as tending towards an equilibrium condition. (Such modelling treats the economic system on the lines of 19th century Physics, which is indeed where the relevant assumptions drew their original inspiration.)
Other less fundamental but equally unrealistic simplifying assumptions that are often made include:
- No transaction costs (e.g. taxes, legal restrictions, or costs of time, travel, or research);
- Pure commodities sold only on price (so no brand or quality distinctions, and no misperceptions based on advertising or marketing);
- Availability of purchasable insurance for any possible outcome.
All of these are quite obviously fantastical, but the presumption is that they can nevertheless provide an ideal model whose results will approximate to reality overall, in much the same way as the behaviour of planets can be modelled very well on the assumption that they are perfect spheres (even though they patently are not). In Economics it is controversial how far this presumption of approximate reliability is justified, something that will no doubt vary considerably between fields of application. Assessing such things is also far from straightforward, not least because so much of the world's commerce has been implicitly organised according to traditional economic models of precisely this kind. When, for example, share prices rise following some relevant event, it can be hard to disentangle the direct impact of the event itself, from its indirect impact due to the results of economic models that are used by analysts to anticipate the likely effects (and who of course react accordingly, triggering actual effects).
Traditional Economics presumably took this direction because of the absence of any rigorous alternative: without making such extreme simplifying assumptions, there was no way of deriving the sorts of solid theoretical results to which Economists aspired. Some alternative approaches have been tried, notably Evolutionary Economics, which substitutes quasi-biological evolution of institutions in place of individual rational planning as the driver of economic explanation (but cannot aspire to the rigorous theory achievable in Biology itself through Population Genetics). As in many other disciplines, the arrival of agent-based computer modelling has at last provided a rival methodology, which holds the prospect of delivering equally solid – but more realistic – new insights into economic behaviour. This new approach has been growing fast, inspired by work from the Santa Fe Institute, and has been particularly popularised by Eric Beinhocker under the name Complexity Economics. For further information and up-to-date resources of many kinds (including course materials), see Leigh Tesfatsion's excellent site on Agent-Based Computational Economics.
The following are simple examples of agent-based computer models with economic relevance: