Desktop versionMobile version

The role of German universities in a system of joint knowledge generation and innovation

 | 
Mirja Meyborg

6. The Concept of the Spatial Analysis

Full text

1As already discussed in subchapter 2.4, it is arguable whether geographical proximity still matters in an ever more globalised economy. The growth of multinationals and global markets as well as the rapid diffusion of the ICT may lead to the conclusion that geography matters much less than did years before. There are many new opportunities for enterprises and institutions which make it possible to nearly completely neglect geographical proximity. In order to provide another or even new input on the topic of joint knowledge generation, the used publication data firstly need to get further prepared. Hence, in order to be able to provide appropriate answers on the topic of geographical proximity, it is necessary to be aware of all distances, measured in kilometres, which are to be covered between the German universities and their individual cooperation partners. Thereby, the ascertainment of the exact geographical position of each co-author has been conducted as follows.

  • 35 A listing is existent in which each German post-code appears with its corresponding geographical c (...)
  • 36 See KIT – Impuls geber für Karlsruhe und die Technologie Region, Kowalski und Schaffer (2012).
  • 37 Of course, there is a lot of university interaction on longer distances which will be considered f (...)

2First of all, the detailed information for each German university and each German university co-author on its address could be used, as it was possible to filter out each individual postal code. By means of another macro, the postal codes could be further assigned to their geographical coordinates, namely its latitude and longitude35 If some post-codes were missing, another search strategy had to be chosen. Thus, frequently occurring cities need to be firstly assigned to their individual post-code, after assigning them to their individual geographical coordinates. Besides, as all German universities are considered, and as some of them are located close to the national border, not only German co-authors but also all co-authors from adjacent countries who are located within a particular radius are considered in this regard as well. As there was no listing available which covers all possible foreign post-codes and their geographical coordinates, frequently occurring cities of each foreign adjacent country were again firstly chosen. Afterwards, they got assigned to their individual geographical coordinates, too. Of course, this procedure was time-consuming, but, as it is aimed to find out whether proximity still matters, a geographical radius has been chosen that narrows down the actors that have to be assigned to their individual geographical coordinates. Finally, a radius of 1,000 kilometer has been adopted due to two reasons; first of all, it can be said that the farthest distance that has to be covered between any two German universities is about 1,000 kilometer (e.g. from Flensburg to Munich or from Greifswald to Freiburg). Secondly, it is assumed that any collaboration that is farer away than 1,000 km is not considered as a regional one anymore; thus, distance of more than 1,000 km is not dependent anymore on the cooperation activity of the German universities.36, 37

  • 38 See appendix for further information on this macro.
  • 39 Of course, there are empirical approaches in the area of transportation modeling which consider ro (...)

3Finally, as each co-author as well as each German university got its individual latitude and longitude, all distances could be calculated and assigned by means of a further self-written macro38. The calculation itself is done by the Haversine theorem39 (Gellert et al. 1989):

d= distance between the two points.
r= earth’radius.
Φ1= latitude of point 1.
Φ2= latitude of point 2.
λ1= longitude of point 1
λ2= longitude of point 2.

  • 40 By means of SPSS a kind of cluster analysis has been conducted which has grouped all German univer (...)
  • 41 See Bellgardt (2004) for further information on applying SPSS.

4Thus, the set of publication data can be separated into two parts. On the one hand, all cooperation partners within the 1,000 km radius can be taken in order to show first results regarding proximity patterns. Second, the geographical distribution of the German university coauthors within the 1,000 km radius has been taken for a subsequent cluster analysis. Generally, cluster analysis groups sets of any objects in such a way that those who are in the same group are more similar to each other than to those in other groups. In this context, by means of SPSS statistics40, 41 each co-author which resides in t h e above mentioned 1,000 km radius could be classified as either being a regional partner or a supra-regional one. The analysis has been conducted for four time periods and has delivered the following cut-off points for the data set with all co-authors that are within t h e mentioned radius:

Table 9: Results of the Cluster Analysis, 2000-2009 (own illustration).

  • 42 The average cut-off point is approximately 468 km and will be used for further analyses.
  • 43 Supra-regional refers to all German university co-authors that are beyond the average cut-off poin (...)

5According to the table, in 2000, all co-authors that have been located in any German university within a radius of 508 km are recorded as regionally located, for 2003 any that were located within a radius of 427 km, for 2006, the cut-off point was at 537 km and finally for 2009, it has again decreased to 402 km.42 Thus, by means of the cluster analysis, it is secondly looked at the overall spatial distribution of all German university co-authors either being regionally located or supra-regionally43.

6Up to this point, it is known where about the German university coauthors are located, either regional or supra-regional based on the results of the cluster analysis; thus, there are up to now two decisions to make in this regard. Therefore, a second step is based upon the geographical distribution of the German university co-authors depending on where they are coming from; hence, it is a country-based analysis. In doing so, it is ensured to also pay attention on different groups of countries. Due to the fact that the German university coauthors come from around 140 different countries, five different groups of countries have been developed in this regard. They are as follows:

  • EU15, including Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherland, Austria, Portugal, Sweden, Spain and United Kingdom (all calculations without Germany plus Switzerland and Norway).
  • EU12, including Bulgaria, Estonia, Lithuania, Latvia, Malta, Poland, Romania, Slovakia, Slovenia, Czech Republic, Hungary and Cyprus.
  • North America, including the three largest partners, namely the United States (US), Canada and Mexico.
  • BRIC, including Brazil, Russia, India and China, and
  • JANZ, including Japan, Australia and New Zealand.

7The hypotheses and the empirical results for both steps are illustrated in the following chapters according to the spatial analysis.

Notes

35 A listing is existent in which each German post-code appears with its corresponding geographical coordinate, so that it was possible to develop a macro which could search for the post-code, simultaneously recording its latitude and longitude in a new column.

36 See KIT – Impuls geber für Karlsruhe und die Technologie Region, Kowalski und Schaffer (2012).

37 Of course, there is a lot of university interaction on longer distances which will be considered for further analyses and is displayed in the following chapters.

38 See appendix for further information on this macro.

39 Of course, there are empirical approaches in the area of transportation modeling which consider road networks, railway networks or the air space in order to measure distances between any two actors using travel time. This PhD thesis has used the Haversine theorem which measures distances in km and not in hours. Using a complex transportation model would not have been productive for the conducted analyses. A transportation model which considers not only the three possibilities of using roads, railways or the air space but also high-stressed traffic networks would definitely gain too much weight in this regard.

40 By means of SPSS a kind of cluster analysis has been conducted which has grouped all German university co-authors according to its particular distances to any German university that has been subject of study. The SPSS TwoStep Clustering Component is a scalable cluster analysis algorithm designed to handle very large datasets. Capable of handling both continuous and categorical variables and attributes, it requires only one data pass in the procedure. In this context, the BIRCH algorithm has been used and the Bayesian Information Criterion.

41 See Bellgardt (2004) for further information on applying SPSS.

42 The average cut-off point is approximately 468 km and will be used for further analyses.

43 Supra-regional refers to all German university co-authors that are beyond the average cut-off point, also including those co-authors that are beyond the 1,000 km radius.

List of illustrations

Caption d= distance between the two points.r= earth’radius.Φ1= latitude of point 1.Φ2= latitude of point 2.λ1= longitude of point 1λ2= longitude of point 2.
URL http://books.openedition.org/ksp/docannexe/image/236/img-1.jpg
File image/jpeg, 24k
Caption Table 9: Results of the Cluster Analysis, 2000-2009 (own illustration).
URL http://books.openedition.org/ksp/docannexe/image/236/img-2.jpg
File image/jpeg, 33k

CC-BY-SA-4.0

The text only may be used under licence CC BY-SA 4.0. All other elements (illustrations, imported files) are “All rights reserved”, unless otherwise stated.

Search OpenEdition Search

You will be redirected to OpenEdition Search