Measuring cognitive proximity

To measure the cognitive proximity, international patent classification (IPC) nomenclature is used. The methodology proposed here have its inspiration in recent proximity studies, for instance by Angue, Ayerbe and Mitkova in 2013.

First step was to prepare the data into desirable format. Another join in SQL has been created between the both groups of patents (university group and the other group) and the table “List of IPC classes and selected dates” (OECD, REGPAT database, July 2019). Each patent application number has been paired with its classifications (IPC). IPC divides the patents into over 68000 distinct sub-domains and has a hierarchic structure. Each patent has its own main code along with sub-codes that refine the description. For the purposes of this paper, the classification was simplified into eight main classes – (A) human necessities, (B) performing operations and transporting, (C) chemistry and metallurgy, (D) textiles and paper,

(E) fixed constructions, (F) mechanical engineering, lighting, heating, weapons, blasting, (G) physics and (H) electricity.

The principle of measuring proximity of patents involves associating each region with a vector summing up its technologies and comparing with a vector of the other region (or group). Each region is divided into two – university and the other. Thus, 42 vectors had been constructed – 21 vectors for regions assigned to university group and 21 vectors assigned to other group (21^st region summarizes patents that hadn’t contain information on inventor’s address or NUTS code). The vector breaks down region’s patents into the 8 IPC sections (there are eight coordinates for each vector, each coordinate summarizes the number of times that region’s patents had been classified into particular IPC class). It then follows that two regions have much closer patent portfolios when their vectors show the same proportions of occurrences in IPC classes. Thus, it is not the magnitude of the vectors that we are interested in, but rather their relative direction. If vectors are co-linear, that means they present the same relative proportions of patents for each of the class. The overall number of patents is not crucial – if two vectors show same proportions, direction of vectors is identical. The overall number of patents (or number of IPC occurrences) changes only the magnitude of the vector, not its direction and thus has no effect on proximity thereafter measured. In other words, region may possess identical technological skills between the university and the other component, owning simply β times more patents (always in the same domain). To measure the proximities, cosines of the angle between the two vectors are considered, as originally suggested by Jaffe in 1986. Since the number of patents held per level is always positive or nil, the cosines calculated are either 0 or 1:

 “0” corresponds to two completely different patent portfolios;

 “1” corresponds to co-linear vectors and thus to identical patent portfolios.

The calculation comprises two stages: (1) to fill in the vectors of the p IPC codes for n regions we want to compare (matrix 𝑀_𝑛𝑝), (2) to calculate the proximity of patent portfolios owned by regions i and j (matrix 𝑃_𝑛). The proximity P is calculated as follows:

𝑃(𝑖, 𝑗) = 𝑀(𝑖) ∗ 𝑀(𝑗)

‖𝑀(𝑖)‖ ∗ ‖𝑀(𝑗)‖

where M(i) is the vector of ith line of M. Matrix M has eight columns (IPC) and 42 rows (regions). The matrix P gives all the proximities measured between the regions taken two by two and thus, calculation yields a symmetrical matrix of 42 columns and 42 rows.

Figure 1. Measure of proximity between patent portfolios

4. Results

This part briefly presents the results obtained using the method outlined above.

Resulting matrix P is in the form of identity matrix. In fact, values in bottom left triangle are mirrored in the upper right triangle. Whatsoever, the matrix is simply too large to display and is partially outlined in Table 1. It contains the proximities between the universities (rows) and their regions (columns). Thus, movement on the diagonal line shows the intra-regional proximities between the universities and its regions. These diagonal values are also displayed on Figure 2. Universities seem to be cognitively proximate and well connected with its regional environ-ment in Budapest, Veszprém, Baranya, Csongrád, Hajdú-Bihar and Szabolcs-Szatmár-Bereg.

Figure 2. Cognitive proximities between universities and their regions

Table 1. Proximity matrix O_HU101O_HU102O_HU211O_HU212O_HU213O_HU221O_HU222O_HU223O_HU231O_HU232O_HU233O_HU311O_HU312O_HU313O_HU321O_HU322O_HU323O_HU331O_HU332O_HU333O_HUZZZ U_HU1010.95210.96810.91260.90550.94840.87400.89010.81440.92630.89980.95560.72010.95670.44950.89820.98100.92310.55720.87110.91640.9323 U_HU1020.75510.74220.56550.86660.88510.71090.61980.49830.61460.71300.76810.59020.72020.17840.69710.85790.83610.29430.52860.72380.9388 U_HU2110.73530.72920.57120.88810.88280.73220.62710.51660.60330.71830.76780.63950.71240.19470.67310.85740.81550.33340.51570.70120.9267 U_HU2120.64040.61470.41460.79920.80710.61220.48990.35980.47630.61470.65900.51480.59160.10100.58870.76000.75370.18520.38260.61620.8720 U_HU2130.72950.70010.50600.84580.87110.66570.56990.42670.57890.68320.74340.56180.67210.15510.68530.82680.83000.23270.48360.70990.9224 U_HU2210.07490.13700.30630.41790.20890.39310.37750.41660.16380.24590.24870.74930.18860.37500.05430.21780.04780.68120.17710.05110.0902 U_HU2220.00140.00160.04640.01220.00000.05100.00000.00000.00000.08940.01240.02430.00000.00000.00000.00000.00410.00000.04960.00890.0000 U_HU2230.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.00000.0000 U_HU2310.92700.91800.89910.65940.78860.68250.81300.71590.97000.76970.89020.51710.87960.42100.92950.82420.84740.44250.92340.92740.7328 U_HU2320.62650.71180.77650.32760.41280.50600.69640.73420.69100.48720.53770.31290.71990.30940.56790.52530.45670.58470.78670.56580.4294 U_HU2330.16090.31470.40400.10520.06370.31400.35560.53780.18440.15990.10560.11690.36810.05480.02060.21000.00860.47130.30740.04200.1580 U_HU3110.95090.93020.90420.81510.89690.77700.84520.72270.97400.85490.96520.68540.89310.45720.95490.90530.91510.48540.90160.95800.8235 U_HU3120.44420.47840.49380.81380.65140.68180.59590.54920.41680.56550.59430.90840.50590.36060.39660.63030.49050.65610.37130.41060.5953 U_HU3130.87910.87110.86540.88780.89480.80750.88150.75020.89890.82610.94040.88340.85430.52190.88380.88500.86170.67310.84690.88080.8044 U_HU3210.96740.94380.90420.85950.93090.79560.88430.74190.97880.86700.98370.75580.91200.50160.97600.92780.94280.55350.91460.97480.8568 U_HU3220.61610.62710.52960.82970.78150.70310.54230.50120.53130.65910.69110.60770.61940.18610.53910.77940.67700.28750.40530.57470.7947 U_HU3230.86920.83230.65480.89150.95330.73210.69490.52970.74920.78090.87000.61820.79520.24870.84290.91490.94140.30690.65670.86020.9780 U_HU3310.83530.86460.72990.84380.88100.77850.75310.67730.72530.75700.81110.59100.85320.22500.75120.90330.84600.45090.69280.77690.9512 U_HU3320.85870.90170.86550.91820.90570.87910.85180.81300.83170.83520.90020.77320.89910.37690.78100.94590.83200.60460.76990.80470.8970 U_HU3330.99350.98440.90510.88730.96720.82560.88630.75660.96620.87560.98830.70870.95240.41580.97300.97510.97470.52110.90430.98060.9405 U_HUZZZ0.21960.27280.44170.48450.31120.46850.49450.50880.32340.35070.37920.80140.31440.44690.21130.32280.17530.73300.33350.20380.1769

Table 2. Regions of Hungary

5. Conclusion

This paper offers detailed view on cognitive proximities between the universities and the regions in Hungary. At first, it has explained the methodology for identifying and collecting the data for university owned and invented patents. The tools such developed may be used without any extensive effort to collect the data for other European countries as well.

It is reasonable to expect, that regions with higher values of proximity should perform well in learning and innovation. Our calculations identified regions in which, universities conduct research in same areas and proportions, as rest of its local economy. Without the doubt, this is desirable output for any innovation system since cognitive proximity facilitates the cooperation and interaction of its own components. Regions that have no or weak cognitive connection with its university component have been identified as well. Whether we aim for well or poorly performing regions, further work is a must before we derive conclusions.

First, the outcomes of our calculations should be paired with the other regional statistics on innovation and macroeconomics. Second, the relationship between the cognitive proximity and the innovation is, we assume, not linear. Thus, for some regions it might be beneficial to lose some of the cognitive proximity.

Having combined both propositions, then desired outcome of further work could be the estimate of the optimal levels of cognitive proximity in different types of regions.

Acknowledgements

This work was supported by the Slovak Research and Development Agency under the contract No. VEGA V-19-147-00

NUTS 3 Region NUTS 3 Region NUTS 3 Region

HU101 Budapest HU223 Zala HU321 Hajdú-Bihar

HU102 Pest HU231 Baranya HU322 Jász-Nagykun-Szolnok

HU211 Fejér HU232 Somogy HU323 Szabolcs-Szatmár-Bereg

HU212 Komárom-Esztergom HU233 Tolna HU331 Bács-Kiskun

HU213 Veszprém HU311 Borsod-Abaúj-Zemplén HU332 Békés

HU221 Gyor-Moson-Sopron HU312 Heves HU333 Csongrád

HU222 Vas HU313 Nógrád HUZZZ Hungary - not regionalised

References

ANGUE, K., AYERBE, C., MITKOVA, L. 2013. A method using two dimensions of the patent classification for measuring the technological proximity: an application in identifying a potential R&D partner in biotechnology. Springer Science + Business Media New York.

BOSCHMA, R. 2003. Proximity and Innovation: A Critical Assessment. In: Regional Studies. Vol.

39. No. 1, pp. 61–74.

COOKE, P., URANGA, M. G., ETXEBARRIA, G. 1998. Regional Systems of Innovation: An Evolutionary Perspective. In: Environment and planning A: Economy and Space.

DOLOREUX, D. 2002. What we should know about regional systems of innovation. In: Technology in Society. 24, No. 3, pp. 243–263.

DORNBUSCH, F., SCHMOCH, U., SCHULZE, N., NADINE, B. 2012. Identification of university-based patents: a new large-scale approach. Fraunhofer Institute for Systems and Innovation Research ISI.

ETZKOWITZ, H., LEYDESDORFF, L. 1997. Introduction to special issue on science policy dimensions of the Triple Helix of university-industry-government relations. In: Science and Public Policy. Vol. 24, No. 1, pp. 2–5.

FAGERBERG, J., MOWERY, C., NELSON, R. R. 2005. The Oxford Handbook of Innovation.

Oxford University Press, ISBN 978-0-19-926455-1

JAFFE, A. B. 1986. Technological opportunity and spillovers of R&D: Evidence from firms’

patents, profits and market value. In: American Economic Review. Vol. 76, No. 5, pp. 984–1001.

NELSON, R. 1959. The Simple Economics of Basic Scientific Research. In: Journal of Political Economy. Volume 67.

SORENSON, O., RIVKIN, J. W. FLEMING, L. 2006. Complexity, networks and knowledge flow.

In: Research policy. Vol. 35, No. 7, pp. 994–1017.

in Western Hungary