The heterogenous relationship between migration and innovation: Evidence from Italy

ABSTRACT

This paper offers a novel take on the relationship between migration and regional innovation by analysing the impact of both international and internal migration flows across Italian provinces, by skill level, and on three types of intellectual property rights (IPRs), namely patents, trademarks and design rights. Allowing us to capture innovation beyond technology and high-tech manufacturing, our results shed light on the relationship between different types of migrant human capital and this array of innovative outcomes. Focusing on Italian provinces in the period 2003–2012, our empirical analysis reveals that internal migration is more significantly related to innovation than international migration. Moreover, medium- and high-skilled migrants are positively associated with all three types of IPRs, while low-skilled migration has a negative association. There are also significant differences across provinces, with a clear distinction between the more economically developed Northern provinces and the rest of Italy.

KEYWORDS:

• International migration
• internal migration
• innovation
• intellectual property rights
• Italy

JEL CLASSIFICATION:

• O15
• R23
• O34, F22

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/13662716.2022.2138279.

Notes

¹ The Regional Innovation Scoreboard (RIS) is the regional extension of the European Innovation Scoreboard (EIS). RIS facilitates a comparative assessment of innovation performance of EU Member States at the regional level (European Commission 2019).

² The micro-data dataset on migration was obtained after a specific request to ISTAT, in compliance with legislation on statistical confidentiality and personal data protection.

³ To retain the zeros, we followed a quite common procedure used by Bratti and Conti (2018) and added 0.001 before taking the logarithm.

⁴ Since we have a short time interval (${\delta }_t)$and differences in migration flows between provinces are quite persistent, NUTS3 fixed effects could not be included. This problem was also emphasised by Niebuhr (2010). We opt for using fixed NUTS2 $\left({\delta }_j\right)$effects, capturing regional differences, instead of provincial ones (Bratti and Conti 2018; Bratti, De Benedictis, and Santoni 2014; Wagner, Head, and Ries 2002). Bratti and Conti (2018) considered this intermediate approach to be particularly effective for the case of Italy.

⁵ Foreign workers, mainly operating in industry (particularly construction), the accommodation sector and family services, have an altogether modest presence in sectors in which Italians are widely employed such as: IT, research and development, and business services (1^st Report on Immigrants in Italy, 2007).

⁶ The pairwise correlations between the instruments and the migrations are above 0.83.

⁷ The Cragg-Donald F statistic (and its robust analog, i.e. rk Wald statistic) is compared to tabulated Stock and Yogo (2005) critical values.

⁸ For the models explaining design intensity, endogeneity tests fail to reject the null hypothesis. Demko (2012) pointed out that, even in the case of strong instruments, weak correlations between the instrument and the error term are not always rejected. To cope with this, we also provide estimations using a Limited Information Maximum Likelihood estimator (or LIML) for designs. LIML estimation is more reliable than the IV estimator when instruments are many or weak in correlations with the error term (Bascle 2008; Murray 2017). The LIML model confirms the results obtained with the 2SLS (see Appendix A.9).

⁹ For medium- and high-skilled international migration it’s safe to said that one standard deviation comes with 19.3% decrease in patent intensity.