Description: In Mexico, an independent electoral commission, the Instituto Nacional Electoral (INE), organizes the Presidential elections. The INE is also in charge of the elections for the House of Representatives and the Senate, as well as an increasing number of local elections. As part of an extensive list of duties and responsibilities, the INE decided, some time ago, to produce a quick count (conteo rápido) to provide preliminary (and reliable) results before midnight on Election Day.
In the first part of this talk, I will present the Bayesian analysis of a simple model that has proved to the successful in several elections (for President and Governor). In the second part, I discuss some more general ideas regarding Bayesian inferences for finite populations, when a stratified sampling scheme is considered.