Is Economic Complexity Spatially Dependent? A spatial analysis of interactions of economic complexity between municipalities in Brazil

Even though the world economy has developed over the past century, divergence has simultaneously led to an increasingly unequal dispersion of wealth. Economists have been fascinated by the underlying mechanisms and determinants; many potential sources of economic development have been scrutinised. Recently, a new theory of economic development has emerged: Hausmann and Hidalgo's economic complexity theory. Economic complexity emphasises the importance of the productive structure and disentangles aggregated measures of economic development such as GDP. Whilst it has been applied extensively in explaining growth at the national level with significant results, it is has only been sparsely used to explain interregional differences and subnational development. This paper assesses whether there are spatial interactions between economic complexity of Brazilian municipalities in 2010 and aims to shed some light on the drivers of economic complexity at the regional level. This paper applies spatial econometrics to assess whether economic complexity is spatially dependent between Brazilian municipalities in 2010 and evaluates what implications this may have for regional industrial policies. As such it is part of both the diversification versus specialisation and of the place-neutral and place-based policy debates. This paper empirically assessed the possible existence of spatial dependence of economic complexity at the municipality level in Brazil in 2010. It found strong and robust evidence in favour of the hypothesised positive effect of spillovers of complexity, especially in the models that include a spatial lagged dependent variable (SAR, SAC, and SDM). The spatial lagged dependent variable estimators are strongly significant in all models, implying that spatial spillovers of capabilities and knowledge do seem to affect economic complexity at the Brazilian municipal level in 2010. These results have important policy implications and can be interpreted as support for the place-based policy approach. Furthermore, the existence of strong path dependency with respect to economic complexity is confirmed by the strong economical and statistical significance of the time lagged Economic Complexity Index variable in all regressions, which provides further evidence in favour of the favourability of place-based policy approaches. Local challenges often differ – this is reflected by the major (spatial) deviations with respect to most variables including economic complexity between Brazilian municipalities in the sample – and these differences appear to endure over time which is convincingly illustrated by the Product Spaces of Brazil. Municipalities should not only take their own bottlenecks and potential solutions of these challenges into account when designing (industrial) policies, but also the circumstances in their neighbourhood or region. Regional policies should take this into account as well and regional cooperation might in certain cases serve as a means to ensure the success of municipal industrial and development policies. Moreover and more generally, multiple control variables have been found to be strongly spatially correlated using univariate Moran's I tests.