Knowledge externalities, innovation and growth in European countries: the role of institutions

ABSTRACT

This paper provides an empirical analysis of the linkages between institutions and economic growth in the European context and highlights innovation as the intermediate variable that drives this interplay. Building on the literature in the evolutionary approach to the economics of innovation and in the economic growth theory with a political economic perspective, we assume that knowledge externalities can fully take place where institutions guarantee a level playing field in the access to knowledge. We estimate the effects of a set of relevant institutional variables on the growth rate of technological knowledge and per capita GDP for a sample of European countries. The empirical analysis confirms that institutions that tend to equalise opportunities to innovate significantly amplify the impact of an exogenous increase in the knowledge base on the growth rate of per capita GDP.

KEYWORDS:

• Technological change
• innovation diffusion
• knowledge externalities
• economic growth
• institutions

JEL CLASSIFICATION:

• O30
• O41
• O43

Acknowledgments

The authors wish to acknowledge anonymous referees and the editor of this journal for their helpful and insightful comments and would like to thank Francesco Crespi for valuable suggestions. The usual disclaimers apply.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Giorgio d'Agostino http://orcid.org/0000-0003-1287-7603

Margherita Scarlato http://orcid.org/0000-0002-7080-3664

Notes

1. Innovation and diffusion refer to the economically motivated efforts aimed at incorporating technological advances into economically exploitable activities (Dosi and Nelson 2010).

2. Investigating the origin of institutions is beyond the scope of this paper. For a discussion of this issue, see Carmignani (2009) and Dari–Mattiacci and Guerriero (2011). The present paper does not consider the role of informal institutions and culture in shaping innovation and economic growth. For an overview of this specific aspect, see Ahlerup, Olsson, and Yanagizawa (2009), Algan and Cahuc (2014), Kaasa and Vada (2010) and Tabellini (2008).

3. The interquartile mean is used to reduce the bias of extreme values on the estimates of the elasticity measures.

4. In line with d'Agostino, Dunne, and Pieroni (2016a,b), Equation (6(6) $e_{\gamma }=\underset{\mathrm{Direct}\mathrm{effect}}{\underbrace{d_{21}\frac{{\overline{\gamma }}^a}{\overline{\gamma }}}}+\underset{\mathrm{Indirect}\mathrm{effect}}{\underbrace{(d_{12}\ast d_{21})\frac{{\overline{\gamma }}^a}{\overline{\gamma }}}},$(6) ) explicitly allows that the link between the covariates (technology and institutions) be asymmetric, hence we cannot use an interaction term directly in Equation (3(3) ${\gamma }_{it}={\alpha }_1{\gamma }_{t-1}+\cdots +{\alpha }_p{\gamma }_{t-p}+{\beta }_1\mathrm{\Delta }{x}_{it}+\cdots +{\beta }_q\mathrm{\Delta }{x}_{it-q}+{\zeta }_1{x}_{it-1}+\cdots +{\zeta }_q{x}_{it-q}+d_t+\mathrm{\Delta }{ϵ}_{it},$(3) ).

5. We do not include an indicator related to the degree of factor and product markets regulation in the empirical analysis because of a lack of data on these dimensions which could be used in a consistent way together with the other institutional variables. On the effects of regulatory practices on innovation in European countries, see the empirical analysis in Barbosa and Faria (2011) and Calcagnini, Giombini, and Travaglini (2016). See also Blind (2012) for an assessment for OECD countries.

6. Note that d'Agostino, Dunne, and Pieroni (2016b) showed that a comparison between three perception indexes of corruption, collected from different sources (World Bank, Transparency International and ICRG), provide statistically equivalent results and, hence, they can be considered as reliable proxies for institutional quality. It is reasonable to extend this finding to the other institutional variables used in our empirical analysis.

7. The second test statistic may be more powerful for heterogeneous panels because it is based on an appropriately standardised average of the individual augmented Dickey–Fuller test and, thus, has a standard normal limiting distribution. Under the null hypothesis, the Levin–Lin–Chu test statistic states that there is no unit root in the series, whereas the Im–Pesaran–Shin test statistic states that there is a unit root.

8. In more detail, the LM Kleibergen–Paap statistic defines a weak instrument when the bias of the IV estimator, relative to the bias of the OLS estimator, exceeds the threshold of 10%. Yogo (2004) provides tables of critical values that are obtained from weak instrument asymptotic distributions and depend on the estimator being used. Following the same strategy, Yogo (2004) provides a tool to estimate the maximum size of the bias, based on the Wald test for each coefficient. As before, instruments are weak if the Wald test based on IV statistics have an actual size that can exceed the threshold of 10%, compared with OLS.