Econometric Modeling of Business Telecommunication Demand Using Retina and Finite Mixtures

We estimate the business telecommunications demand for local, intra-LATA and inter-LATA services, using US telecommunications data from a Bill Harvesting survey carried out during 1997. Available information does not allow specifying demand as function of prices; hence we use flexible modelling estimation tools in order to find a useful specification with improved predictive ability over alternative models. We use an automated selection method called RElevant Transformations of the Inputs Network Approach (RETINA) due to PerezAmaral, Gallo and White [9] to approximate the demand function for each service. In addition, we use finite mixtures as a convenient framework to model heterogeneity due to the different usages of telecommunications services in business activity. Combining these tools in an appropriate way we obtain models with improved predictive ability over benchmark models estimated without using any variable selection strategy.