Technological and organisational innovation Firms' learning behavior Institutional frameworks fsQCA SUR model a b s t r a c tThe innovation systems approach, which has taken a prominent position in the academic literature, has also influenced policy-makers around the globe. Most research analyses innovation systems taking a national, regional or sectoral perspective, following a 'technological imperative'. Yet changes in institutional conditions and the importance of non-technological innovation question the accuracy and the relevance of the existing boundaries of innovation systems. These developments ask for a better understanding of how innovation systems integrate within and across different levels. Drawing on a novel combination of configurational and econometric analysis, we analyse 384 Swiss firms and identify five co-existing innovation systems: two generic innovation systems, the autarkic and the knowledgeinternalisation; one regional innovation system, the protected hierarchy; and two sectoral innovation systems, the public sciences and the organised learning. The generic innovation systems entail the 'Science, Technology and Innovation' (STI) and the 'Doing, Interacting and Using' (DUI) learning modes. These systems are structurally distinct and do not integrate. In contrast, all regional and sectoral innovation systems integrate the learning modes of the generic innovation systems and complement them with idiosyncratic elements. The perspective on co-existing innovation systems that we develop here indicates the existence of two layers of innovation systems: a 'central' layer that hosts generic innovation systems and that constitutes the foundation for a second 'surface' layer that hosts regional and sectoral innovation systems. We discuss the implications of layers of co-existing innovation systems for policy-makers and future research.

provides helpful comments on earlier drafts. We also would like to thank Natalia Abrosimova and Jan Hagen for excellent research assistance and Natalie Reid for editorial support. Furthermore we would like to thank the Swiss Economic Institute for data provision. Funding This study is partly funded by the Swiss State Secretariat for Education, Research and Innovation (SERI) through its Leading House on the Economics of Education, Firm Behaviour and Training Policies.

and the participants of the annual meetings of the Canadian Economics Association, Swiss Society for Economics and Statistics, the Bildungsökonomischen Ausschusses im Verein für Socialpolitik, the Spring Meeting of Young Economists, the Colloquium on Personnel Economics, the DRUID Society Conference and the research seminars at the University of Zurich for helpful comments and suggestions. We acknowledge and thank Natalia Abrosimova and Jan Hagen for research assistance and Natalie Reid for editorial support. Furthermore we thank the Swiss Economic Institute (KOF) for data provision. Funding This study is partly funded by the Swiss State Secretariat for Education, Research and Innovation through its Leading House on the Economics of Education, Firm Behaviour and Training Policies.

Systematically combining quantitative and qualitative research approaches offers the potential for a more comprehensive and nuanced understanding of social scientific phenomena. With their strong opportunities for building, qualifying, and testing social scientific theories, methodological integrations thus enable researchers to make substantive contributions that would not have been possible with one method alone. In this article we demonstrate how the integration of Qualitative Comparative Analysis (QCA) and conventional statistical analysis offers researchers new opportunities for contributing to the social sciences. Whereas statistical analysis is variable-oriented and relies on correlational analysis to make comparisons across cases, QCA is based on set theory, is case oriented, and relies on Boolean algebra to make comparisons between cases. Drawing on the literature on the interdependency between theoretical contribution and methodology, we review studies that integrate QCA and statistical analysis to explain how the specific combination of these two approaches allows researchers to strengthen the theoretical contribution of their research. From our review we identify common challenges and provide solutions for integrating QCA and statistical analysis.Johannes Meuer and Christian Rupietta have contributed equally to this work.