The impact of local and external university knowledge on the creation of knowledge-intensive firms: evidence from the Italian case

Abstract

This paper investigates how far in space university knowledge goes to breed the creation of knowledge-intensive firms (KIFs), depending on the nature (either codified or tacit) and quality of this knowledge. We consider the impact of knowledge codified in academic patents and scientific publications and tacit knowledge embodied in university graduates on KIF creation in Italian provinces in 2010, while distinguishing between local university knowledge created by universities located in the same province and external university knowledge created by universities located outside the province. Our econometric estimates indicate that the positive effects of scientific publications and university graduates are confined within the boundaries of the province in which universities are located. Conversely, the creation of new KIFs in a focal province is positively affected by both local and external university knowledge codified in academic patents, even though the positive effect of this external knowledge rapidly diminishes with geographic distance. Furthermore, the above effects are confined to high-quality universities; low-quality universities have little effect on KIF creation.

Appendices

Appendix 1: Estimation of the decay parameter

To set the distance decay parameter α for each type of university knowledge (patents, publications, and graduates), we evaluated the value that maximizes the log-likelihood function of the econometric model described by Eq. (3).²³ We tested increasing values of α from 0 to 5 in increments of 0.1.

For each type of university knowledge, Fig. 3 shows the likelihood ratio LR = −2[log-likelihood(unrestricted model) − log-likelihood(restricted model)], where the unrestricted model corresponds to Eq. (3), while in the restricted model, we set to zero the coefficients of external university knowledge variables (i.e., xHQ ^external _i and xLQ ^external _i ). More specifically, depending on the value of the decay parameter α considered, the dotted, grey, and dashed lines represent the LR values associated with patents, publications, and graduates, respectively. Finally, the bold line is the critical value of the Chi-square distribution with two degrees of freedom in an LR test at 95 % confidence level (if the LR is higher than this critical value, we reject, with 95 % confidence, the null hypothesis that external university knowledge variables are zero).

Fig. 3

Likelihood ratio tests of external university variables (patents, publications, and graduates), depending on the decay parameter

Figure 3 shows that the values that maximize the log-likelihood (i.e., the LR) are 1.7, 4.6, and 4.4 for patents, publications, and graduates, respectively. Accordingly, these values are used as the decay parameters in evaluating the effect of external university knowledge on new KIF creation. Finally, when considering these maximum values, it is worth noting that we can reject the null hypothesis that external university knowledge variables are zero only for patents (at the 5 % significance level).

Appendix 2: Industry classification

See Table 7.

Table 7 Knowledge-intensive industries

Appendix 3: Publications

For each university, we computed a measure that accounts for the differences in the frequency of publication in each of the 151 research areas listed in the ISI Web of Science. Specifically, for each university u and ISI research area a, we first computed the ratio of the number of ISI publications of university u in research area a to the total number of ISI publications generated by the 80 Italian universities in research area a:

$$ S_{u,a} = \frac{{{\text{ISI Publications}}_{u,a} }}{{\mathop \sum \nolimits_{u = 1}^{80} {\text{ISI Publications}}_{u,a} }}. $$

The ratio S _u,a represents the proportion of ISI publications in research area a of each university u with respect to the total number of ISI publications generated by all Italian universities in research area a.

Then, for each university u, we calculated the arithmetic mean of the ratios S _u,a across the 151 research areas:

$$ {\text{AS}}_{u} = \frac{1}{151}\mathop \sum \limits_{u = 1}^{151} S_{u,a} . $$

For each university u, AS _u represents the average proportion of ISI publications across research areas. Finally, to obtain a count measure, we multiplied AS _u by the total number of ISI publications generated by all Italian universities in the period 2000–2008 (339,737):

$$ {\text{Publications}}_{u}^{*} = 339,737 \cdot {\text{AS}}_{u} . $$

In other words, we used the average proportion of ISI publications produced by each university u across research areas (AS _u ) to attribute to each university u the corresponding fraction of the total number of ISI publications generated by all Italian universities.

Appendix 4: Robustness checks

See Tables 8 and 9.

Table 8 Impact of local and external university knowledge, depending on the nature and quality of knowledge, controlling for other agglomeration effects

Table 9 The spatial range of university knowledge, depending on the nature and quality of knowledge—OLS and 2SLS regressions