Entrepreneurial ecosystems and economic resilience at local level

ABSTRACT

The main aim of this paper is to investigate if and to what extent entrepreneurial ecosystems (EE) have an impact on economic resilience at local level. The paper is based on a quantitative analysis for the Italian provinces (NUT-3 level) and presents two novelties: first, it provides a composite index of EE at local level by capturing the different aspects encompassing political, social, cultural and economic dimensions of an EE; second, it analyzes the role of EE in terms of resistance to and recovery from external shocks. The empirical results show that EE has a relevant role in explaining the resilience of local systems to economic shocks. The positive effect also remains when controlling for the direct impact of new firm formation, thus highlighting that the EE concept has a greater significance for characterizing resilience and entrepreneurial activity at local level than entrepreneurial rates.

KEYWORDS:

• Entrepreneurial ecosystem
• resilience
• regional policy

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For an analysis of the Italian double-dip recession see Locatelli, Monteforte, and Zevi (2019).

2. NUTS (Nomenclature of Territorial Units for Statistics) is the a geocode standard for referencing the subdivision of countries for statistical purposes, adopted by the EU. NUTS-3 delimit areas with a population between 150,000 to 800,000. In Italy NUTS-3 corresponds to the administrative provinces with an average population of about 600,000 people.

3. Data shows that this is the class-size with the highest share of entrepreneurs (Curci and Micozzi 2017). There is also a strong correlation between the composition of the population and aggregate level of entrepreneurship for the US, with a peak for the 25–34 class-size (Shane 1996).

4. In this paper we exclude the demand variable (i.e. provincial VA) from the original EE index, since in the empirical model the resilience indexes, the dependent variables, are measured in terms of VA. However, the exclusion of VA in the EE index does not sensibly change the original EE provincial ranking.

5. For further details about the construction of the Index see Iacobucci and Perugini (2020).

6. To date, the number of Italian provinces is 110. However, during the period 2007–2016, the number and the political and geographical structure of Italian provinces changed. The provinces of Ogliastra, Carbonia-Iglesias, Medio Campidano, and Olbia-Tempio Pausania were formed in 2006, while the provinces of Monza, Brianza, Barletta-Andria-Trani, and Fermo in 2009. As a result, these 7 provinces are not included in the sample, so the number of provinces was reduced to 103.

7. Estimation results for the FCP, which are available from the authors upon request, confirm the result obtained for the overall EE index reported in Table 5, i.e. they have a positive and statistically significant impact only on the recovery phase. Also, the coefficient for the systemic condition shows the highest value.

8. It is common in the regional science literature to test the model for the presence of spatial autocorrelation (for a deeper analysis on spatial models see LeSage 1998). We therefore perform the Lagrange Multiplier tests and accordingly we estimate the model in equation (1)(1) $R_{n,p}={\alpha }_0+{\alpha }_{1}E{E}_p+{\alpha }_k\sum _{k=1}^N{X}_{k,p}+F_p+{\epsilon }_{n,p}$(1) as a Spatial Autoregressive Model (SAR), which takes into account the presence of a spatially lagged dependent variable, and a Spatial Error Model (SER), which measures spatial autocorrelation through the spatially lagged error terms (Anselin 2005). We detect a statistically significant spatial correlation with the 5-nearest neighbours weighting matrix only. Estimation results, which are available from the authors upon request, show that the coefficients on the spatial lag dependent variable, is statistically not significant, suggesting that a positive impact on the resilience of a province will not spread through the provincial system. The spatially lagged coefficient in the SER model is statistically significant and with a negative sign only when the dependent variable is the recovery index. This suggests that there are some local characteristics that cannot be captured by the explanatory variables but that negatively affect the recovery of the neighbour province. For completeness, we also estimated a Spatial Durbin Model (SDM), by adding as explanatory variables the spatial lag of the independent variables. The coefficients on the spatial lag of the EE index are not statistically significant in all cases. Overall, results show only a weak spatial correlation, and this is limited to the selection of a 5-nearest neighbours weighting matrix.

9. This indicator refers to birth of an enterprise that has at least one employee in the birth year and of enterprise that existed before the year in consideration, but were below the threshold of one employee (OECD and of the European Communities 2008).

10. This is defined as number of firms with average growth in the number of employees greater than 20% per year over a three-year period and with ten or more employees at the beginning of the period (OECD and of the European Communities 2008).

11. To limit the proliferation of subscripts in equation (2) and (3) we avoid the subscript t for time. However, in equation (2) the dependent variable is taken as average in the 2005–2007 period, while the independent variables on the right-hand side are taken in the year 2004. In equation (3) the independent variables, except for the fitted EO, are taken in the year 2007.

12. The EE index used in these regressions is computed by excluding the ‘Entrepreneurship Culture’ variable. However, the Spearman rank correlation coefficient with the original EE index is 0.98.