Filter: Show German content (if available)
Business cycle research: Iterative forecasts within the national accounts framework (2005-present)
Outline:Monitoring and forecasting short-term developments in the economy on the basis of national account concepts (2-year-forecast horizon) and flash-estimates; methodological enhancements for iterative, knowledge-based forecasting (development of the agent-based software package DIWAX for data management and manipulation within a consistent national accounting scheme).
Material:Presentations, Publications, Software
Knowledge-based decision support systems for market workability diagnostics (1998-present)
Outline:This project focuses on developping an agent-based decision support system (DSS) designed to assist human experts in analysing the workability of market processes and serves as a framework for electronic market studies. The building blocks of a market study are represented as agents that can interact with the user and with each other allowing for cooperation between human experts and the machine. The aim of the system is not to replace the user but to give assistance in many ways (e.g. by means of completeness and consistency checks). The system does not resolve contradictory facts but it makes them tranparent and shows the consequences of alternative assumptions. The core contents of a market study are stored in an SQL database which does not only allow for quick and easy retrieval but also for machine learning based on the growing experience that evolves when the system is used over time. Traditionally, experts deliver a heap of paper or a text file when they are asked to give their opinion. These papers are interconnected very poorly (via unidirectional footnotes only), they are difficult to compare and inaccessible for machines and - when the number of studies increases - for human beeings alike. Therefore this kind of knowledge representation is considered inadequate. Structured electronic market studies can overcome these problems. This approach is not limited to a specific view on how markets may work. However, following the German ordolibertarian tradition, one branch of the system's agents tree is developed for market analyses on the basis of the Coordination Failure Diagnosis (CFD) concept as suggested by GROSSEKETTLER. This concept combines empirical evidence supported by theoretical reasoning about whether or not market processes show a sound process pattern or not. Starting from the null hypothesis of reasonably satisfactory market performance the system cooperates with the user in order to find evidence for the opposite (existence of coordination failures); if it does not succeed the null hypothesis prevails, otherwise political measures to improve the coodination capacity of the market should be taken into consideration (depending on the kind of the detected failure). Therefore, the system can be used for systematic market screenings and serve as a warning system for undesirable trends in the market system. The implementation of the CFD concept is done for demonstration purposes. If - for one reason or another - this is not what you want, different agents can be added that help to cover other concepts. Therefore, this approach supports a new view on how academic progress and policy consulting should be linked: economists should not only think of new theories but they should also develop the tools for applying them. As a side effect, this assures that economics keep track of real-life problems; furthermore implementing economic theories within objects is a very valuable test for completeness, consistency and applicability for the theories themselves.
Material:Presentations, Publications, Software
Macroeconomics/business cycles: Soft computing for expectation formation modelling and forecasting (1996-present)
Outline:This research project aims at modelling expectation formation processes that are more realistic than the traditional autorgeressive or rational approaches by means of soft systems that allow for (1) vagueness/uncertainty, (2) hypothesizing (mental models) and (3) a bounded learning capacity when it comes to model the forecasting behavior of economic agents. Key soft computing technologies for this purpose are fuzzy logic, neural networks, genetic algorithms and gene expression programming. Results of this research are also applied to real-world forecasting tasks.
Material:Presentations, Publications, Software
Economics of software markets
Outline:Workability of software markets, forms of competition and regulation design in platform industries
Material:Presentations, Publications