Estimation of Internet availability at home by a score prediction function

Abstract

The availability of Internet at home is one of the most important steps to enjoy the benefits that come from its use. In this sense, knowing the weight implied by the availability or lack of the network becomes strategic in the design of policies and actions to intensify the socio-territorial penetration of the Internet and the empowerment of its users. The aim of this research was to develop a function for predicting Internet availability in households, based on microdata from the 2020 population and housing census conducted by the National Institute of Statistics and Geography (INEGI). This type of prediction, in addition to clarifying the socio-territorial conditions involved in the physical availability of the Internet in homes contribute to the structuring of an intervention context for the public and private factors of digital disengagement. In methodological terms, the prediction function is based on a random sample of 11 000 census interviewees, divided into two groups of equal consistency: the first has Internet at home and the does not. To generate the prediction, a linear discriminant function was used and valued by a score function. The quality of the prediction was estimated through an internal evaluation with Bootsrap (75% of the sample) and an external one (25% of the sample), both of which were complemented with an analysis of the ROC curve and LIFT graph. The resulting function exhibits a sensitivity of 76.9%, a specificity of 20.5% and a discriminant capacity of 86.7% in a 95% confidence interval. With these efficiency parameters, the found function becomes a possible effective prediction tool for both reflection and action against the digital divide in its initial stage.