Abstract
Administrative regions do not necessarily correspond to areas that are homogenous in terms of innovation intensity. Although this has been recognized in the literature, quantitative evidence that explicitly considers this problem is rare. Using spatial exploratory analysis on detailed regional data derived from a census of R&D performers in the Czech Republic, we identify local spatial clusters of R&D activities and assess the extent of their (mis)match with administrative borders. Overall, the results support the arguments for regionalization of innovation policy. However, the existing policy units do not appear well suited for this purpose. On one hand, there is a need for policy coordination between multiple administrative regions. On the other hand, however, there are diverse patterns within them. Similar problems are likely to haunt the regionalization process in many other countries, which is alarming, as the regional accent of innovation policies becomes ever more vehement over time.
Acknowledgements
We are grateful to the Czech Statistical Office for providing access to the confidential microdata, in particular the help of Martin Mana with constructing the regional data-set is highly appreciated. Financial support from the Czech Science Foundation (GAČR) project P402/10/2310 on “Innovation, productivity and policy: What can we learn from micro data?” and institutional support RVO 67985998 from the Academy of Sciences of the Czech Republic are gratefully acknowledged. An earlier version of this paper was presented at the 6th International Seminar on Regional Innovation Policies, Lund, 13–14 October 2011 and the GAČR project interim workshop, Prague, 10 January 2012. We thank participants at the these events, in particular Martin Andersson and Jiří Blažek, for comments. All usual caveats apply.
Notes
1. Major organizational changes, i.e. mergers, split-ups, spin-offs, etc. that involve transfers of R&D employment to descendants with new identification numbers can lead to double counting. To avoid this bias, we exclude from the analysis 25 entities, which are newly established during the reference period, and for which the mode of entry is recorded in the Business Register of the Czech Republic as one of the organizational changes mentioned above or for which this information is missing. Their combined R&D employment is 237 only, which is a small fraction of the total and thus the potential impact is negligible.
2. Prague is aggregated to a single POU region, even though formally this is not the case, as the capital city is divided into 15 administrative units and 57 town districts. Because Prague is a compact city, and because other major cities (Brno, Plzeň, Ostrava etc.) constitute single POU units, we do not follow the formal division.
3. More specifically, the (log of) population density accounts for about 38% and 16% of regional differences in the (log of) R&D employment (FTE) per km2 in the business and public sectors, respectively.
4. Ord and Getis (Citation2001) developed augmented statistic that properly test for local spatial autocorrelation in the presence of the global autocorrelation, but unfortunately the data in hand is not extensive enough to allow for computing this measure.
5. However, the difference between Gi and is significant only in data with a small number of neighbours and tends to rapidly decrease with the increasing number of neighbours. It should be noted that in this paper the actual difference is inconsequential.
6. Note that this has been further reinforced by the opening of a brand new campus of Masaryk University in 2011 and the opening of the CEITEC in 2012, although obviously this has not been reflected in the data yet.
7. Note that the reference period of this support programme is 2007–2013; however most of the money is going to be spent in the second half of the period; the first supported projects started operation in 2011, so this has not been reflected in the presented employment figures.