Towards an Automata Approach of (Institutional) Economics

A computational approach towards economics potentially enriches economic science beyond increasing available mathematical techniques. Computational economics (CE) can foster a viable and rich institutional economics that encourages both mathematical rigor and historical relevance while avoiding the mechanical aspects of conventional neoclassical theory. Here we begin such an approach by regarding markets as computational entities or literal automata, where 'automata' refers to the formal notion of a computational device developed in computability theory (a branch of formal logic). We begin by introducing the reader to a literature that draws heavily on computability theory and from which we learn that an approach that is too abstract or context-insensitive is vulnerable to uncomputability results. For example, certain instances of neoclassical theory are possibly uncomputable, i.e., unable to be computed in countably many computations. These results suggest limited practical relevance to certain economic theories. We argue this critique can be avoided by specifying the context in which economic exchange takes place. Recasting the rules that constitute a market onto an automaton ensures that the economic context is explicitly defined, and further, it enables us rigorously to analyze the relevance of different market settings for economical performance. Already there is a large literature -- in particular in experimental economics and finance -- dealing with concepts similar to our suggested automata approach. In the case of the experimental literature, numerous experiments have been conducted that analyze the dependence of economical performance on the market institution. Similarly in the financial literature, as a consequence of the ongoing automation of markets, it has become an issue to analyze the relevance of different market designs. In other words, both situations treat economic performance as dependent on the context given by the market institutions. This paper takes this approach one step further by actually perceiving markets as computational entities. To illustrate this point we provide an example showing how markets can be encoded as automata.