The impact of KIBS’ location on their innovation behaviour

ABSTRACT

Knowledge-intensive business services (KIBS) are widely perceived as being important drivers of technological progress and innovation. They generally depend on knowledge exchanges and, therefore, geographical proximity to markets, customers and suppliers would be expected to be a critical factor in their performance. This paper investigates how the innovation performance of German KIBS firms is related to their distance and size from the nearest city. The analysis largely conforms to a textbook type of spatial urban hierarchy and, indeed, finds that there are very strong distance-decay and city size effects, and these also vary according to the innovation type.

KEYWORDS:

• distance decay
• innovation
• knowledge-intensive business services (KIBS)
• urban structure

JEL:

• D22
• L84
• O31
• R12

ACKNOWLEDGEMENTS

The authors acknowledge the constructive and valuable comments made by the participants at the 3rd Geography of Innovation conference in Toulouse, France, 2016; as well as the journal's anonymous referees.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. For example, see the seminal studies by Acs, Audretsch, and Feldman (1992) and Jaffe, Trajtenberg, and Henderson (1993).

2. NACE = Statistical Classification of Economic Activities in the European Community. For the years before and after the validity of this classification, we recoded the classifications to NACE Rev. 1.1.

3. For a detailed description of the combined database and the construction of variables, see also Brunow and Blien (2015).

4. For further information on the IAB establishment panel (EP), see Fischer, Janik, Müller, and Schmucker (2008) and Ellguth, Kohaut, and Möller (2014).

5. $d=\text{acos}(\text{sin}\varphi 1\cdot \text{sin}\varphi 2+\text{cos}\varphi 1\cdot \text{cos}\varphi 2\cdot \text{cos}\mathrm{\Delta }\lambda )\cdot R$, where d is distance; ϕ is latitude; λ is longitude; and R is radius.

6. NUTS = Nomenclature of Territorial Units for Statistics.

7. We also tested incremental distance to the closest city types of higher order, but could find significant patterns and, therefore, such variables are omitted from the analysis.

8. We also tested other ring sizes starting with 1 km and gradually increased ring size and added rings ranging from, for example, 1 km to 3 km, 3 km to 6 km etc. up to 100 km. The finer scale of rings did not bring further results and we therefore decided to take only one ring with a radius of 17 km to warrant better comparison with existing studies on regions. The average distance of the centroid from one NUTS-3 region to another is 34 km. Choosing 17 km draws a circle around each firm that is comparable with the NUTS-3 diameter. Within such a ring, we did not weigh down more remote postcode areas because the analysis of smaller rings did not show significant differences in parameters.

9. Several other variables relating to urbanization and location economies were tested, but they turned out to be insignificant: the absolute number of distinct industries and the number of firms in all other sectors; the proportion of human capital employed in all KIBS firms and in all other sectors; the distance to the closest university; the distance to the closest R&D institution to capture other sources of potential human capital spillover; and the distance to the local state government and to the federal government located in Berlin.

10. The slope of the probability functions depicted in Figure 2 relates to the average marginal effect (AME) of distance on innovation probabilities.

11. The threshold of a 25 km minimum distance is chosen as it covers 90% of all KIBS establishments. Differentiating according to the closest city type, it covers more than 95% of all KIBS establishments located closest to a metropolis, 60% of those closest to a large city and slightly more than 75% of those closest to a small city.

12. The distance patterns might be misleading and biased if, for instance, the closest city type is a small city, but the next larger city is also nearby. Such cases become especially relevant in dense areas, such as the Rhein-Ruhr area. Because the incremental distance measure turned out to be insignificant, we conclude that such problems are of minor importance.

13. In robustness checks we therefore introduced distances to federal state capitals and the capital city Berlin. These measures always turned out to be insignificant, while the results so far remained. This also implies that the distance measure does not serve as a proxy for governmental institutions.