It appears you don't have support to open PDFs in this web browser. To view this file, Open with your PDF reader
Abstract
In this paper, we aim to identify the covariates associated with the proportion of votes of candidates elected in Brazilian municipalities with a population of more than 300,000 inhabitants. We analyzed the vote proportions from the 2018 presidential runoff election using distributions within the Generalized Additive Models for Location, Scale, and Shape (GAMLSS) class. Unit distributions are quite useful for modeling vote proportions due to their flexibility to accommodate asymmetry and heavy tails. Furthermore, they provide adequate representations of the physiological properties and the empirical distribution of the data. We _t the beta, simplex, unit gamma, and unit Lindley regression models, considering random and fixed effects components to verify spatial correlation among the municipalities. The beta regression with fixed components regarding Brazilian regions is superior. The covariates with significant effects are the proportion of evangelicals, monthly household income per capita, the political spectrum of the governors' party elected in 2014 and 2018, and if the municipality is the capital of the state. We note that some Brazilian regions impact the vote proportions' mean and dispersion.