Statistical Modelling

Firms need to know production functions, factor costs, demand functions and competitors` positions to make rational decisions. How do firms obtain this knowledge, how is organiza-tional knowledge stored, how is it renewed? Classical economic theory does not touch upon these issues - however, it is assumed that the economic agents already possess the relevant knowledge and the issue is the study of equilibria that arise from the decision based thereon. As already noted by famous Austrian economists like Joseph Schumpeter and August von Hayek, this approach neglects innovation as central element of a capitalist economy. There-fore, a number of interesting real-world phenomena like the development of industries and markets, the sustainability of competitive advantage and differences in growth in economic systems cannot be studied with orthodox theory. The goal of SFB10 "Adaptive Information Systems and Modelling in Economics and Man-agement Science" is to study the evolution of such real-world "patterns" through the interac-tion of boundedly rational adaptive economic agents. By combining methods of machine learning developed in artificial intelligence with traditional economic models, the research team of SFB10 could already study a number of such evolutionary phenomena: Average costs decrease as a function of accumulative output due to increase in experience and continuous improvement. Through the simulation of learning processes during prod-uct and process development, we could replicate these phenomena and study how the per-formance of artificial firms composed of agents modelling marketing and production is in-fluenced by the internal organization of the firm. Real-world financial time series show statistical properties that are not captured by classi-cal theory. Through the study of an artificial financial market consisting of investors with heterogeneous adaptive expectations, the emergence of price-time series that resemble real-world analoga could be observed. New technologies and/or changes of consumer behaviour offer chances for market entry via new market segments and can lead to disruption of existing mainstream market lead-ers. Firms thus need capabilities to recognize such chanceshreats and must be organized as groups of learning firms that are tailored to the specific learning environment. Advanced mathematical and statistical methods are necessary to study these issues. Thus, also leading experts in these fields participate in SFB010, and the development of formal methods is fostered through their application in the SFBs context. The applications of SFB010s research are manifold: based on theoretical arguments, we can advise firms how to organize knowledge and quality management, we can consult on market entry strategies and on how firms should value projects in knowledge generation.