The impact of higher education institution-firm knowledge links on firm-level productivity in Britain

Abstract

This article estimates whether knowledge links with universities impact on establishment-level Total Factor Productivity (TFP). Using propensity score matching, the results show a positive and statistically significant impact although there are differences across production and nonproduction industries and domestically and foreign-owned firms.

Acknowledgements

The authors acknowledge the support from the Economic and Social Research Council (Grant No. RES-171-25-0032), the Scottish Funding Council (SFC), the Department for Education and Learning (DEL) in Northern Ireland, the Higher Education Funding Council for England (HEFCE) and the Higher Education Funding Council for Wales (HEFCW). This work contains statistical data from Office for National Statistics (ONS) which is Crown copyright and reproduced with the permission of the controller of HMSO and Queen's Printer for Scotland. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research data sets that may not exactly reproduce National Statistics aggregates.

Notes

¹The actual questions used to define Higher Education Institutions (HEI) collaboration are Q.16 and Q.18 (see http://www.bis.gov.uk/policies/science/science-innovation-analysis/cis/cis4_questionnaire for details).

²There are two main channels through which collaborating with HEIs may have an impact on Total Factor Productivity (TFP): indirectly through product and process innovation which improves efficiency and/or involves the use of better technology. There will also be a direct impact on TFP which will arise if collaboration with HEIs gives rise to knowledge spillovers from the HEI to the establishment.

³Some industrial sectors are omitted from the Annual Respondents Database (ARD) – such as agriculture and much of financial services.

⁴We calculated the weights at the two-digit industry level, split into five employment size bands.

⁵Our model is essentially a reduced form of the model employed by Arvanitis et al. (2006), which is itself based upon the standard CDM model (Crepon et al., 1998).

⁶More detail on these variables is available in an unpublished appendix (see http://www.gla.ac.uk/t4/economics/repec/gla/unpublished%20appendix%20for%20AEL.pdf).

⁷The results from the (weighted) probit models are available in the unpublished appendix. Different models were estimated depending on sector.