ABSTRACT
The paper investigates whether the tertiarization and rapid urbanization faced by developing countries favour agglomeration economies. Focusing on Ecuadorian cantons, a productivity equation is estimated using the generalized method of moments model with instruments controlling for endogeneity. The varying impact of industrial concentration, diversity, competition and density across industries is investigated and, for the first time, the implication of the level of urbanization on agglomeration externalities is studied. Stronger effects are found for services. The threshold of urbanization at which diversity, density and competition agglomeration externalities all generate positive effects was 33%, while they seem challenged by congestion in highly urbanized cantons.
DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.
ORCID
Grace Carolina Guevara-Rosero http://orcid.org/0000-0001-7605-1443
Notes
1. Another strategy consists in using firm-level data. However, some variables at the firm level, which are needed to explain the individual variability, are generally missing. Thus, only one variable (lifetime) is available to explain the variability at the firm level. Consequently, such a strategy is not adequate for econometric reasons.
2. Previous versions of the paper focus on labour productivity. They are available from the authors upon request.
3. Diversity externalities can also be derived from so-called related variety, which is the interaction between sub-industries within one industry (Frenken, van Oort, & Verburg, Citation2007). The present study provides evidence on the diversity externalities that originated in the interchange between different industries. However, estimations using related variety measures were also conducted. Two measures were tested: (1) related variety at the cantonal and sectoral level based on Frenken et al. (Citation2007); and (2) related variety at the cantonal and sectoral level based on the standard measure. Owing to the sensitivity of the results to the measure used (Frenken versus standard), they are not presented here, but are available from the authors upon request.
4. Ordinary least squares (OLS) and 2SLS estimation methods were also used. By comparison, the 2SLS and GMM estimations show quite similar results in terms of coefficients signs and magnitudes, while the OLS estimation differs regarding the sign of the competition effect.
5. Enough evidence is found to reject the null hypothesis of the Pagan–Hall general test. The estimates are then presented with standard errors which are robust to arbitrary heteroskedasticity.
6. The classification of industries is shown in Appendix A in the supplemental data online.
7. The differences in the coefficient estimates between both industries are statistically significant according to the Chow test. The F-statistic of the test is 59.54 and p = 0.000.
8. According to the Population and Housing Census of 2010, urban settlements are defined by provincial and cantonal capitals under the political and administrative division of the country. Rural areas include parish capitals, other towns, peripheries and sparsely populated settlements.