Identifying Clusters within R&D Intensive Industries Using Local Spatial Methods

In recent times, there has been a renewed interest in cluster policies for supporting industrial and regional development. Prominent economics like Porter and Krugman emphasise the role of clusters in regional competition and show in which way clusters can positively affect competition by increasing productivity and innovation. Because of the linkage between growth and innovation, R&D intensive industries play a crucial role in cluster development strategies. Empirical cluster research has to contribute to the understanding the process of cluster formation. In particular for developing profound clusters strategies and assessing the limits cluster policy, knowledge of existing structures and tendencies is necessary. In these strategies, high-tech and research-intensive industries play a crucial role. Audretsch and Feldman argue that industries with high innovation activity tend to cluster for exploiting benefits from tacit knowledge flows. Krugman stresses that information flows and knowledge spillovers may be sensitive to geographic impediments. Since obstacles tend to rise with increasing distance, spatial clusters may be localised. If, however, geographic barriers are less relevant, the reach of tacit knowledge flows may be much larger. For regional policy the geographical level at which clusters occur is of prominent interest. Traditional concentration indices like the Gini coefficient, Theils’s inequalitiy index or the Ellison-Glaeser index are ‘aspatial’ by construction. This means that these indices disregard relevant spatial information on the distribution of a geo-referenced variable. In particular, attribute values of adjacent regions are completely ignored. Moreover, the spatial scale of clustering formation is not taken into account. First experiences with methods of exploratory spatial data analysis (ESDA) like local Moran’s I and Getis-Ord Gi statistics in pattern recognition are available. Le Gallo and Ertur (2003) utilise local indicators of spatial association to analyse the distribution of regional GDP per capita in Europe. Feser et al. (2005), Lafourcade and Mion (2007) and Kies et al. (2009) demonstrate the potential of these ESDA techniques in identifying economic clusters and spatial heterogeneity in geographical space. However, while usually local Moran’s I and Getis-Ord Gi statistics are applied in detecting economic clusters, up to now spatial scan techniques are largely ignored (Kang, 2010). In this paper, advantages and pitfalls of spatial scan tests in identifying R&D clusters are examined. Some essentials in implementing spatial scan techniques in economic clusters research are worked out.