ABSTRACT
A latent class model is applied to allow entrepreneurial ecosystems (EEs) to influence the effect of entrepreneurial activity on growth in European Union regions. Using this methodology, clusters of regions that differ significantly in their relationship between entrepreneurial activity and growth are identified. This is consistent with the hypothesis that EEs affect this relationship. Subsequently, cluster membership is related to regional characteristics representing a range of components of EEs and marked differences in a variety of these regional characteristics are found. Taken together, the results support the notion that EEs help shape the impact of entrepreneurial activity on growth.
DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.
Notes
1. Some argue that EEs are only present in regions where productive entrepreneurship exists. In contrast, we take the view that EEs can exist in all regions, but that they differ in quality, creating performance differences.
2. See Reynolds et al. (Citation2005) and Bosma (Citation2013) for a detailed explanation of the methodology underlying the GEM survey.
3. For example, see Bosma and Sternberg (Citation2014) and Content, Frenken, and Jordaan (Citation2019), who use a similar approach to calculate indicators of regional entrepreneurship in the EU. The limitation of using this approach is that it results in cross-sectional indicators averaged for the period, preventing us from adding a time dimension to our analysis.
4. Figure C1 in Appendix C in the supplemental data online shows the distribution of the prevalence rates of these different types of entrepreneurship and what NUTS level is used for which country.
5. By taking this two-step procedure, we implicitly impose on the model that the estimated parameters of the growth equation are equal across regions. In a sample that includes developing countries, this assumption would be too restrictive, but in a sample of European regions, this should be acceptable. Moreover, as we are primarily interested in EE, this procedure allows only the marginal effect of entrepreneurial activity to differ across latent classes.
6. Factors to distinguish between groups of regions include, for example, income (Hessels & van Stel, Citation2011), institutions (Hall & Gingerich, Citation2009) or geographical location (Redding & Venables, Citation2004).
7. Bosma, Content, Sanders, and Stam (Citation2018) use a different approach by positing entrepreneurial activity as a mediator of the effect of institutions on economic growth. However, such an approach still assumes that the effect of ecosystem characteristics is the same across units and it only examines one element of the underlying EE.
8. At the 5% significance level, a Hausman specification test of ordinary least squares (OLS) versus random effects rejects OLS in favour of random effects (1327.78, d.f. = 7). A random effects specification is not rejected in favour of fixed effects specification (8.48, d.f. = 7).
9. One would expect this to happen first to the more specific measures of entrepreneurial activity, with, on the one hand, extreme entrepreneurship strictly defined as, for example, only those activities that contribute to GRP growth and, on the other, inclusive proxies that include all kinds of ‘entrepreneurial’ activity such as self-employment or new firm formation. As explained by Bosma et al. (Citation2018) and also discussed above, TEA and OPP are more inclusive and noisier than JOB.