Complexity Theory, Formal Modeling, and Entrepreneurship in the Production Structure of Ideas
Those who disagree with the reigning paradigm in economics – macro or otherwise – must find alternatives to the system that they challenge or else they may do more harm than good. Those who have strong methodological foundations must find a way to systematically convert their understanding into output: formal mathematical models. To be sure, we cannot do without high theory. Theorizing about economics without formal mathematical models can be a fruitful and invigorating exercise. But research, if it is to be integrated by a broad segment of researchers, must yield something easily interpreted by an audience. Otherwise, it will be received as noise. Last post I juxtaposed Romer’s critique of macroeconomics with the more constructive work of Koppl, Kauffman, Felin, and Longo (2015). I was surprised to see that Romer does not cite any papers with agent-based models. The word “complex(ity)” does not appear in the paper. Again, compare this to Koppl, et al. who cite Beinhocker (2007), among other thinkers within the paradigm of complexity, and who elaborate on the need for complex methods:
As Colander et al. (2004) have chronicled, mainstream economics has been greatly influence by complexity theory (also see Durlauf, 2012). It has become more inductive and open. But mainstream economists still tend to be attached to mathematical methods that are not always well suited to a creative economy, as with DSGE models. (3)
It seems likely that the tools of high theory should continue to evolve in the direction taken by complexity theory. In particular, computability theory seems to give us a set of tools useful for sorting out what is feasible and what is not feasible in both theory and policy. Network theory seems well suited to an economics for a creative world. (27)
Models rooted in complexity theory deviate from the assumptions of the dominant paradigm of micro a la Arrow-Debreu and macro a la Lucas. These include Gode and Sunder (1993) and Axtell (2005). In a recent paper (online, forthcoming in print), Santiago Gangotena provides a framework of this that sharply deviates from the old paradigm: Dynamic Coordinating Non-Equilibrium. In addition to providing a framework, he describes precedent from modelers who create agent-based simulations that suggests this potential:
Holland and Miller (1991) present one of the earliest proposals for the adoption of ABM in economics. They stress the need to conceptualize economies as adaptive systems with adaptive agents capable of producing perpetual novelty. More recent arguments for the adoption of ABM in economics can be found in Vriend (2002); Tesfatsion (2002); Tesfatsion (2003); Tesfatsion (2006); Axtell (2007), and Farmer and Foley (2009). Seagren (2011) argues that ABM is wholly consistent with the generally process driven approach of Austrian Economics.
In the same issue, I have presented a model with similar priors. It is a model of a two-good economy that extends Epstein and Axtell (1996) and that employs ecologically rational agents. Agents in this model have knowledge with complex structure that takes the form of rule governed behavior that is ecologically rational. Agents must attain a sufficient flow of two types of goods in order to survive. Each agent chooses a desired quantity of each good that it would like to hold and adjusts the price at which they are willing to buy and sell the goods as their holdings deviate from the desired levels. Some agents move back at forth at a select rate. Other choose what good they wish to produce according to prices that they have observed. Agents are capable of learning from one another by copying successful agents. Others experiment by selecting new values and strategies (akin to mutation in evolutionary biology). Selection by the system ensures that agents act rationally in the sense that their actions promote survival. Model results show that the system selects for agent-strategies in manner that suggests the equimarginal principle. The average rate of return for each strategy converges upon the same value. Individual optimization presented in neoclassical calculus is a tendency promoted by the system and this is most quickly reached when there exists a fair margin of experimentation.
At this point, there is a shortage of competing approaches to formal non-equilibrium modeling. Some authors are using standard neoclassical agents who maximize utility (i.e., Axtell 2005), which is likely a good starting point for earning respect by those who manage the most prestigious journals. They are taking the production structure of ideas at face value. Influence major journals, and you may gain greater influence over ideas used across the field. This is a significant and necessary approach. Behind these attempts their exist a bastion of possibility to be discovered by entrepreneurs of ideas. Major journals will receive novel ideas when they can be efficiently communicated to the rest of the field. Those who contribute to these journals are given the responsibility of communicating clearly and succinctly innovations of which most researchers are ignorant (i.e., Kirzner 1997). The more intense is the competition within the pool of ideas that these researchers draw from, the more that they will have to offer the top journals. Experimentation is a long-term investment that carries with it high risk for the entrepreneur engaged in it. Those who are most respected have the advantage of drawing from these investments when they appear to be successful. We cannot know ahead of time what the correct formulation will be. The market for academic ideas decides this.