No public surveys, no data? A proposal for income forecast in brazilian municipalities

Abstract

Objective: Due to the lack of regularity from the census in Brazil, the proposal to use alternative indicators is relevant. The population's income, primary census information, is a variable used in studies in different areas such as public policies, forecasting, and planning a new business. However, on average, this information is released every ten years in Brazil; thus, it is necessary to explore frequency variables to estimate the population's income. In this sense, this study proposes a predictive income model based on technological and communication aspects.Method: We choose two variables: internet and cable TV access. Our study included the analysis of the 5570 Brazilian municipalities through linear (OLS) and spatial models (Spatial Auto-Regressive [SAR] and Geographically Weighted Regression [GWR]).Results: The results with the spatial models showed better results. The autoregressive spatial regression (SAR) presented a more significant explained variance (R2 = 0.51) than the linear model (R2 = 0.41) and the GWR model (R2 = 0.44), which indicates a significant spatial dependence and contribution of the geographic perspective in modeling and explaining the phenomenon.Conclusion: The results were found to contribute to the development of public policies in regions with difficult access to information on the population's income and with managers and companies in the technology area that seek to plan the improvement and expansion of the provision of digital communication services through a model updated with secondary public data.