The unrealistic assumptions that economic models are based on have rendered many models insufficient and impractical in application in the real world. Aishwarya Ketkar offers Agent Based Computation as an answer for those who want to simulate models capturing actual human behaviour and experiences….
Economy is an interactive complex system reflecting the interaction between various agents residing within it. The attempts at classifying Economics into a perfect science are transforming into varied perspectives being used to predict consumer behavior. There is a clear indication of evolution of a novel mathematical ideology into economics taking aid of the massive computing power made feasible by technology.
Leon Walras was the first person to classify Economics as a science. To Walras and the Physicists of his time, economy was a closed system following simple mathematics to gain equilibrium. The general notion pointed towards external shocks, followed by temporary disequilibrium adjustment and new equilibrium. Physics was at a very rudimentary level when Physicists first applied the concept of equilibrium in Economics. The first law of Thermodynamics (Energy conservation) was the only discovery present in Physics (of Thermodynamics). The establishment of second law of Thermodynamics (endless increasing entropy) itself proved that there was no one simple mathematical perspective to handle Economic systems. Now that a lot of principles in Physics have evolved, the perspective to look at Economic systems has changed tremendously. What is baffling today, is aptness of applying non evolutionary mathematical algorithms to define and predict an economic phenomenon.
ACE amalgamates the elements and views from Economics, Computer Science and Social Sciences to study modern Economic systems. Excellent surveys of extant literature in this field indicate that the conceptualization of an economy as a system of interacting agents is not a new concept; the simple issue is that there is no stable evidence to suggest the convergence of economic processes towards a stable equilibrium. Additionally, the assumption of perfect rationality advocated by traditional Economics is increasingly being abandoned by social scientists. It is assumed that an individual gets information, processes it, analyses it and then makes decisions. Theoretically, outcomes predicted by these systems are misleading and differ from their outcomes in the present world. The assumption of perfect rationality is a restrictive assumption. ACE corrects this caveat by enabling more realistic assumptions which lead to more definite outcomes.
The idea of ACE is simple: Agents (decision makers) interact with each other and the environment they are placed in through a predefined set of ‘simple’ rules. They achieve locally optimal results that may or may not result in global optima. These rules can be designed to be perfectly rational or not depending on the phenomenon under research. The rules can evolve over time to simulate effects of experience and culture. Thus, the economy can be modeled as a ‘complex adaptive system’ which it ideally should be.
One of the most significant studies in Agent Based Computational Economics is the work done by Joshua Epstein and Robert Axtell that resulted in development of the Sugarscape model. The Sugarscape model envisions the behavior of artificial people namely the agents on a landscape growing generalized resources namely, Sugar. A certain amount of Sugar is established on the landscape which keeps growing at a uniform rate as a fixed period passes. The agents are born on the landscape of the sugar with a fixed vision, metabolism and genetic attributes which are randomly chosen from a uniformly distributed population. The field of Agent Based Modeling in evolutionary biology extensively discussed by Axelrod introduces further evolutionary changes in the programming of agents on Sugarscape based on the genes inherited through the agents. The behavior of randomly chosen agents is governed by fairly simple rules that try to maximize their utility as the economic theory envisages. Their movement, for example, is determined by their sight, brain power and metabolic rate. They just have to look as far as they can, locate their sugar source and spend some energy towards moving and acquiring it. When agents exhaust their energy (determined by metabolic rate), they die. emerge as a result of these simple rules. For instance, as soon as the model is executed, agents who are born on the sugarscape with little sugar die off quickly before they are able to spot the sugar. This leads to huddling of agents around the more fertile part of the Sugarscape. Epstein and Axtell then introduce a variety of other factors into the system namely seasonal variations, Tribes, and sexual reproduction.
This study is based on the economic tradeoff chosen by the consumers between two goods in order to attain the maximum level of utility given the constraint that is reaching the highest indifference curve. It results in the development of an economic market where agents trade Sugar and Spice based on their metabolisms for the two resources and their present endowments. With the introduction of Price, the agents can be asked to choose their tradeoff between Sugar and Spice at that particular Price point. Summing these values across agents and prices gives us a demand and supply schedule that is generated from the ‘bottom –up’. Epstein and Axtell also determined the average actual quantity and price at which trades took place. This point can then be compared to intersection of the supply and demand curves to determine the disparity (if any) between the market clearing price and the expected clearing price (assuming the existence of a Walrasian auctioneer).
Sugarscape has had a tremendous impact on the research community in ACE and economics in general. Almost all contemporary research refers back to their model and its framework. There are a lot of factors that impact the game under consideration. Using ACE there are local interactions which are modeled under a set of user defined conditions. The output from ACE models, with less restrictive assumptions, is still a sub optimal outcome; however in cases suchh as those observed by Mark in his model, there is a rare emergence of globally optimal set of strategies. Furthermore, it can also be seen that the level of rationality, whether perfect or bounded, has an impact on the outcome of the model. In this way agent-based simulations cann help us to close the gap between theory and the real world.
Aishwarya Ketkar is pursuing MS Mathematical Finance at Boston University.