On technology emergence and pattern detection in aggregate innovative activity

Theories of General Purpose Technologies (GPT) introduced heterogeneity in the world of technological change in order to explain fluctuations in economic growth, productivity paradoxes, or technological clusters. The first part of this thesis addresses in detail a GPT's emergence, presenting a new theoretical model that focuses on pervasiveness character of a GPT emphasizing knowledge network characteristics. Each new product in the economy is a result of a combination of technologies forming a complex network structure of technological inter-dependencies, where a general knowledge emerges as a result of knowledge spillovers, coordination of innovative efforts among economic agents, dynamics of expected profits, and the structure of knowledge base. The model demonstrates similar network characteristics when compared to empirical networks of products and technologies as well as explains clustering of innovations in time, change of technological paradigms and mechanism leading to a technological lock-in.
It is argued that robotics and especially new developments in service robotics can be considered as a potential GPT. Employing a machine learning technique, namely Support Vector Machine, the second part of this thesis introduces a methodology for identification of service robotics patents within databases. ... mehrThe result is a novel possibility to allocate patents which reduces expert bias regarding vested interests on lexical query methods, avoids problems with citational approaches, and facilitates evolutionary changes. Resulting patents are geographically localized and analyzed, being a proxy for knowledge production in service robotics.
The last part of the thesis focuses on a general detection of emergent patterns in micro data. Firstly, a method for statistical identification of clusters of innovative activity is applied to service robotics patents and all German R&D data. A micro-geographic approach identifies spatial localization or dispersion by comparing observable spatial distance patterns between R&D establishments to counter-factual simulations. Findings demonstrate the localization of the knowledge production in service robotics as well as the share of localized German industries in production being higher than in services. Secondly, employing a new methodology based on Markov chain simulations it is assessed whether the number of sustained superior employment growth performing firms in four European economies is different from what would be expected by chance. A mixed evidence of presence and absence of factors determining firm dynamics is found.