Socioeconomic Class of Brazilian Cities for Health, Education and Employment & Income IFDM: A Clustering Data Analysis

摘要

The FIRJAN system, through the simple average of three basic aspects of development, Employment & Income, Education and Health, calculates an index, the IFDM, to sort the city into four categories in order to assist the government in public policies. However, the simple average of these three aspects and the classification of the cities in four categories, as previously defined, may not actually represent natural groups that these cities are included. Therefore, by means of an unsupervised classification, such as the clustering data analysis, it is proposed to examine whether there are natural groupings of cities the basic aspects mentioned. For this, we used the hierarchical method WARD and the non-hierarchical k-means method, with the criteria of validation width silhouette (SWC) and sum of squared errors (SSE) to find groups of municipalities in three basic aspects of development. For validated statistically were used Monte Carlo analysis width criterion of silhouette, under the null hypothesis that the data were random. With significance level 5%, was rejected H0, thus indicating there is strong evidence of the existence of natural groups. We identified two as the best number of groups and, after analyzing the percentage of cities in each Brazilian state are in group 1 and 2; it was possible to validate the resulting grouping of prior knowledge of the development of each region of the country. We also identified two subgroups found in each of the groups resulting, therefore, in four representative categories. The subgroups also went through the same analysis that the groups.