de Mendonça, Mário Jorge Cardoso da Silva, José Jaime
Year of Publication:
Texto para Discussão 2444
The ICT Household Survey aims to collect data on information and communication technologies (ICT) access and use in Brazil. We use this research to elaborate two exercises. The first one applies logistic regression to determine the factors responsible for connecting a domicile to the Internet. The second one estimates a demand model for the Internet applying a multinomial model of discrete choice. The model contemplates four types of connection: DSL with fixed telephone line, TV cable or fiber optic and mobile via modem or 3G and 4G chip. In the case of logistic regression, we verified that socioeconomic variables such as schooling, income and social class are factors that influence whether the household is connected or not. We verified that the higher the level of schooling or income, the greater the effect on the probability of the household be connecting to the Internet. We observe that there is no defined pattern concern to the effect of a variable on the probability of choosing a type of connection. Thus, each variable responds differently depending on each type of connection. This study is the first attempt with disaggregated data of estimation of Internet demand for Brazil. Although it contains certain weaknesses due to the limitations of the database, it is still a valid exercise insofar as it suggests important indications for the construction of a more reliable database for the supporting studies of Internet demand in Brazil.
broadband internet demand logistic regression type of connection multinomial model