Specialization of Regions and Universities: The New Versus the Old

Abstract

This paper analyzes whether there is a correspondence between a university's research specialization and industrial specialization in the region hosting the university, and to what extent universities influence regional productivity. Moreover, the analysis seeks to answer if a difference can be detected between the influences of old and new universities on regional performance. To achieve this end we utilize a unique data set on spatially disaggregated data for Sweden in the period 1975–99. A two‐step Heckman regression analysis is implemented to examine whether universities' research specialization matches regional specialization in production as compared to the average region. The results suggest a correspondence in specialization, as well as positive productivity effects. However, there are also considerable differences across regions, albeit primarily unrelated to the age of the universities.

Keywords:

• Universities
• norms
• regional specialization
• policies

Acknowledgements

Helpful comments have been provided by participants at the 2nd Workshop on the Process of Reform of University Systems in Venice, 2006, the Schumpeter 2006 meeting and two anonymous referees. The author would also like to thank the Marianne and Marcus Wallenberg Foundation for generous financial support.

Notes

1. Indication of intensified regional restructuring towards knowledge‐intensive industries is illustrated in increased branding activities (Medicon Valley, Silicon Glen, Sophia Antipolis, Telecom Valley, etc.) and the establishment of regional investment organizations.

2. Positive effects are however reported for the USA (Link, 1996; Caloghirou et al., 2001), for Norway (Gulbrandsen and Smeby, 2002) and Belgium (Ranga, 2003) in Europe. Other studies claim that the transfer of IPRs had little to do with the increase in commercialization (Mowery and Sampat, 2001; Mowery and Ziedonis, 2001, 2002; Mowery et al., 2001; Nelson, 2001, 2002).

3. The extent of, and channels through which, knowledge is sourced from universities appears to vary with firm and industry characteristics. See, for instance, Zucker and Darby (1996), Meyer‐Krahmer and Schmoch (1998), Hall et al. (2000), Fontana et al. (2002), Mohnen and Hoareau (2002), Arundel and Geuna (2004) and Geuna and Nesta (2006).

4. Anselin et al. (1997) stress the differences across industries. See Feldman (1999) for a survey of the spillover literature. The role of proximity for innovation is addressed by Kline and Rosenberg (1987), Acs et al. (1992) and Arundel and Geuna (2004).

5. Note that Franklin et al. (2001) report the opposite results: older universities were found to be more entrepreneurial and more open.

6. See Amin and Thrift (1994), Lundvall and Johnson (1994), Armstrong et al. (1997), Autio and Yli‐Renko (1998), Maskell and Törnquist (1999), Thanki (1999) and Karlsson and Zhang (2001).

7. In particular, the endogenous growth models failed to model the way in which knowledge is converted (spills over) into commercially viable products. Hence, the main mechanism to promote growth—knowledge spillovers—remains exogenous. This shortcoming was to some extent remedied in the subsequent neo‐Schumpeterian models (Aghion and Howitt, 1992, 1998; Aghion and Griffith, 2005). Still, these models are constrained to a very particular form of entry and spillover mechanism (R&D races).

8. Altogether Sweden has three private and 14 public universities.

9. For a more detailed description of this process, see Acs and Braunerhjelm (2005).

10. Including funding for research and PhD students (Högskoleverket, 2006).

11. The universities in Stockholm and Gothenburg, together with the four universities we are focusing on, comprise the major part of university research in Sweden.

12. The Tobit method is a conceivable candidate. However, the estimates reflect both changes in the probability of being above the limit, and changes in the value of the dependent variable if already above the limit. A decomposition of the effects is possible (McDonald and Moffitt, 1980), but the problem is that the two separate effects will always have the same sign and significance. In the present case it may well be that the probability effects and the marginal effects differ. Alternatively, since the location choice of firms is multinomial by nature, one way of accounting for this would be to estimate a multinomial logit or probit. However, the multinomial logit relies on a very strong assumption, the independence of irrelative alternatives, and the multinomial probit involves the evaluation of multiple integrals, something that is not feasible if the choice alternatives exceed three or four. Given these limitations, we believe that the best model to use is Heckman's two‐stage estimation technique.

13. It should be noted that the probit and corrected Heckman OLS equations include the same explanatory variables in vector Z. A possible practical problem is then multicollinearity between Z and λ. There is no theoretical basis that such problems must arise, however, since the latter variable is a non‐linear combination of Z, while OLS is a linear estimation technique.

14. Standard properties are assumed to prevail with regard to the error term.