Rată de participare, supraraportare și predicții electorale

Abstract

The paper proposes a model for correcting errors for the voting predictions based on social survey data. The model considers the relation between the rate of participation in voting (at aggregate level and by category of population), for the previous elections, and the current voting intention. The sociodemographic profile of the voters is measured using exit poll data. The vote intention is given by current public opinion surveys. The basic idea is to use for estimating the actual participation in voting behavioural variables at the ecologic level, instead of using declarative variables. The aim is to set up a predictive model for voting, at an aggregate level. The 2000 elections’ data partially support the model, while the predictions for 2004 are consonant, no matter which data set is considered.