Convergencia en America Latina: una nueva metodologia de analisis

BRIDA, G, LONDON, S, RISSO, W

U.F.F, Rio de Janeiro

Workshop; International Workshop on Economic Growth; 2008

UFF

Abstract The aim of this paper is to apply a new method to study the evolution of real per capita GDP and growth rates of the American countries. The method combines the tools of Symbolic Time Series Analysis [1] with the nearest neighbour single linkage clustering algorithm [2]. Data symbolization allows to obtain a metric distance between two different countries that is used to construct an ultrametric distance. By analyzing the data of our sample, we derive a hierarchical organization, constructing minimal-spanning and hierarchical trees. From these trees we can detect different clusters of countries (according to their proximity) that can be interpreted as convergence clubs. Specifically, taken data for American countries, we select a double partition based on two variables, GDP and growth rate. We define four regions where the thresholds are the corresponding global average for all countries. Therefore a symbolic time series is defined for each country and the distance are computed. One obtained the distance matrix we proceed to rank the distances selecting the most relevant connections. We connect all the countries in a unique tree where the loops are forbidden. As second step we computed the global distance as a measure of distance among the countries, it could be consider as a measure of integration in the sense that the closes the global distance the more integrated are the countries. In this way, we can obtained a measure of convergence across countries in a dynamical terms, with an economic interpretation far away different of the traditional concept of convergence [3].