1 A Co-evolutionary Simulation of Multi-Branch Enterprises Edmund Chattoe Department of Sociology University of Oxford 3 George Street Mews Oxford OX1 2AA, UK edmund.chattoe@sociology.oxford.ac.uk http://www.sociology.ox.ac.uk/chattoe.html 25th January 1999, revised 13th December 2000 Abstract One significant argument against the application of evolutionary models to the behaviour of firms has been the absence of an obvious “reproduction” mechanism, by which successful enterprises can increase the proportion of their behavioural genotypes in the population solely to their own benefit. Interestingly, there is class of enterprises, relatively new in the history of industrial capitalism, which fits the traditional evolutionary metaphor rather effectively. (It could be argued that this form of organisation has itself evolved in the same way that sexual reproduction is believed to be an evolved mechanism which speeds up genetic mixing.) These are firms with multiple branches (including franchises) that have a single brand image, set of working practices, range of products and so on. If such firms establish a successful mix of attributes, the resulting profits allow them to construct new branches and, in some cases, buy out less successful chains, as Burger King were able to buy out Wimpy. Thus, in a real sense, they are replicating their behavioural genotypes in the population of competing branches. Furthermore, there is a direct link between profit (defined as cash surplus above costs) and “reproductive fitness” which is often neglected in evolutionary models. The most obvious examples of multi-branch enterprises are fast food chains. This paper presents a simulation in which firms with spatially located branches compete for customers. Firms receive feedback about the performance (fitness) of individual branches, and decide what working practices, products and other features of each branch to retain and which to modify. Matters are complicated by the existence of scale economies - which provide an incentive for firms to standardise practices across branches - and variations in consumer preferences (including spatial variations) - which provide an incentive not to standardise unduly. In addition, there are significant costs connected with changing some aspects of branches, such as decor. The simulation is co-evolutionary in two senses. Firstly, each firm may modify its behaviour not only in the light of feedback from its own strategies, but from the strategies of other firms. Secondly, customers are sensitive not only to the attributes of individual branches, but to the possibilities of the market as a whole. Thus, if only one chain offers a vegetarian option, it will capture not only the vegetarian market, but also those who have a casual preference for vegetarian food. However, if no chains offer a vegetarian option, consumers will lack any expectation of such products and only the most devout vegetarians will avoid the market altogether. Different learning mechanisms are compared, not all of which are evolutionary. For example, one simple strategy would be to take the most profitable branch and enforce its strategy in all branches. Clearly, this would not necessarily work well if there was significant variation in the preferences of consumers. It would also suppress the ability to experiment with the market. The purpose of the paper is to specify a simulation in which evolutionary and non-evolutionary mechanisms of learning can receive a “fair comparison”. Because consumers are modelled independently, measures of consumer surplus, costs and profits can be used to compare different adaptation mechanisms in a consistent way.