Impact of agglomeration on the regional growth of Latin American countries

Theoretical approaches have been developed to examine the effect of agglomeration on growth. However, the understanding of the mechanisms of agglomeration in developing countries remains unaddressed. This paper aims to give empirical evidence of the role of agglomeration on the growth of Latin American regions. The study of the subcontinent is crucial because of the evidence of a rapid pace of urbanization process. Using a database with information of 162 regions of 8 Latin American countries (Argentina, Bolivia, Chile, Colombia, Ecuador, Mexico, Panama and Peru) during the period 2000-2009, we estimate the effect of agglomeration on regional growth in three periods. The measures of agglomeration are urbanization rate and population density. The data is based on the information provided by the National Institutes of Statistics and the Central Banks of each country. The geographical coordinates were obtained from the GeoHack system of Wikitech. The control variables are: educated labor force, public investment and sectoral specialization. After proving that our variables of agglomeration instrumented with altitude are not endogenous and given the finding of spatial correlation between Latin American regions, a spatial autoregressive panel model with fixed effects is estimated by Maximum Likelihood. For the sake of this estimation, we apply three spatial weight matrices. The first one is the k=1nearest neighbors weight matrix which is interpreted as the configuration of low integration. The second one is the weight matrix based on Gabriel method which is interpreted as the configuration of moderate integration. And the third one is a distance weight matrix where all regions are connected. It is interpreted as the configuration of quasi-complete integration. Our findings suggest that agglomeration is vital for the regional growth. However, the effect of agglomeration is not the same everywhere. In this line, we test Williamson?s (1965) hypothesis of more pronounced agglomeration effects at early stages of development than at later stages of development. Using the statistical theory proposed by Hansen (2000), the threshold value of level of development at which the effect of urbanization changes was 5700 dollars of per capita income. Low-developed regions experience larger positive effects of urbanization on their economic growth than high-developed regions. In our sample, the positive effects vanish at 10,500 dollars of per capita income. Finally, the degree of spatial autocorrelation increases with the level of integration that the spatial weight matrix reflects.