Modelling the Process of Market Emergence
Essential trading patterns within the economy are determined by a linear input-output flow matrix. Three types of products are produced in the economy: iron, spades and corn. Spades are required for the production of iron and corn, iron is required for the production of spades and corn is required for corn production (as seed) and is also the only consumption good. Households supply labour and demand corn.
Table 1: Input-output table
We thus have three production sectors. Each sector contains 15 enterprises. Each enterprise demands labour from one household so that there are 45 households in the economy. Units are chosen so that one unit of employment is required to produce one unit of any product.
The enterprises are intelligent in the sense that they learn about their environments and seek to modify their behaviour in the light of experience. The process of learning is represented as a process of model specification, testing and revision. This will be described in more detail presently.
Each date in a simulation experiment concludes with the execution of decisions taken within that date. There are three phases within each date. The first is for data analysis ó including the evaluation of the performance of models defined in previous periods. The second is for high-level planning. The enterprises translate the results of the analysis phase into new models and rules which those models might imply. During the final phase ó the execution phase ó the agents execute the best rule setting the values of each of their decision variables.
The models developed by agents in the simulation experiments reported here develop relationships among the directions of changes in the variables which are observable. Agents also invent their own variables in order to find simpler models that are better guides to action. The variables invented by agents combine existing variables in ways that reduce their dimensionality. If, for example, there are a variety of inputs, each purchased at its own price, then the agent might try to reduce the dimensionality of each price-input combination my multiplying the two since the dimensions of price are (money xb8 good) and the dimensions of the input is (good xb8 time) so that the mathematical product of the two has the dimensions (money xb8 time). As the example indicates, this is a natural operation that we undertake with respect to a wide range of variables in formal modelling. A further example of mapping inputs into money is that all input costs can be summed. Such a simple procedure leads naturally to variables used by economists and business analysts.
Generated with CERN WebMaker