Benefícios previdenciários e desenvolvimento socioeconômico: uma análise preditiva

• Ano de defesa: 2020
• Autor(a) principal: Reis, Edmilson Xavier dos
• Orientador(a): Silva, Thiago Christiano
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Católica de Brasília
• Programa de Pós-Graduação: Programa Stricto Sensu em Governança, Tecnologia e Inovação
• Departamento: Escola de Educação, Tecnologia e Comunicação
• País: Brasil
• Palavras-chave em Português: Seguridade social; Previdência social; Desenvolvimento econômico
• Palavras-chave em Inglês: Social security benefits; Supervised learning; Development
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS
• Link de acesso: https://bdtd.ucb.br:8443/jspui/handle/tede/2801

In 1982, the United Nations documented its concerns about the aging of the population, producing recommendations, including the maintenance of social security benefits, to face the challenges that pervade a society with an increasingly advanced middle age. The resources of the social security benefits granted by the General Social Security Regime - RGPS in Brazil work as a guarantee of income source for the elderly population and feed the municipal economy, helping economic development. The debate on social security costs is reopened from time to time. The program is considered to cause increased state spending. This study contributes to the literature on public finances and economic development, documenting which types of social security benefits most predict municipal economic development using non-linear machine learning forecasting methods. We seek to understand whether the RGPS program has any indirect benefits for society. Brazil has a broad income distribution program operated through the RGPS, with various types of social security benefits. The technique of ranking the importance of attributes gives us information about the importance of one attribute in relation to the others to predict the target variable. However, it does not provide us with information about the marginal effect of each attribute. As we are dealing with highly non-linear models, understanding the marginal effect of each variable is a challenge, because there is no conceptual interpretability of the effect of each attribute for the prediction of the target variable. Using ranking techniques of importance of attributes, we found that the variable that most predicts municipal socioeconomic development is the value of retirement by GDP per capita, this gives an idea of representativeness indicating that all retirement value is important for the local economy. The second and third most anticipated benefits are the socioeconomic indicators Firjan education and Firjan employment. To verify the robustness of the results, we used several supervised machine-learning techniques, such as Support Vector Machines, elastic net, extreme gradient boosting trees, random forest, decision tree and linear regression. To find out which algorithm had the best performance, we used the combination of error measurement models MAE, RMSE and Rsquared. The performance of all individual models showed that the SVM technique (RMSE = 0.2057) performed better, but the ensemble (RMSE = 0.2007) is the one that best predicts the relationship between social security benefit and economic development.