Analyses of poverty indicators using PPI methodology

Abstract

Poverty is a negative socio-economical phenomenon which has a destructive influence on not only life quality for people who are caught in a poverty trap but for society in general.  Poverty prevents society from realising its potential and leads to social development regression. The first of the sustainable development goals declared by the United Nations in 2025 as global goals of civilisation development until 2030, is to overcome poverty. The growth of poverty rates is observed for the first time in twenty years, and this requires studying the causes and developing policies to prevent this tendency.  It is well-known poverty has various forms and national and cultural features, that should be taken into account when solving the poverty problem. One of the relevant tasks in studying this problem is the development of adequate methods of measuring poverty and determining the category of people who are considered poor This study aims to develop a clusterization model using the set of socio-economic indicators in order to identify the poor people cluster. The survey includes 8400 respondents from 7 European countries. Developing the model was carried out using machine learning methods in several steps: 1) data processing and statistical analyses; 2) selection of significant indicators by the classification model; 3) clustering by k-mean algorithm; 4) hierarchical clustering; 5) comparing outcomes of modeling and interpretation of results. The selection of indicators was carried out by classification methods using PPI methodology. Data processing and analyses were performed using Python. Using this approach we can split the population into groups with different living standards levels identifying the poor people group with a simple questionary considering national (local) features. This help to increase the effectiveness and timeliness of poor families’ support.