Sesgo de no-respuesta y modelos de superpoblación en encuestas electorales

Abstract

Nonresponse bias (and, to a lesser extent, measurement error) has become
the main source of error for electoral forecasts in Spain. Although
the post-stratifi cation techniques and ratio estimators currently used in
the polling industry reduce deviations, they do not show enough capacity
to mend the biases introduced when collecting data. This research
reveals how a more effi cient use of the electoral information available
outside the sample could help to signifi cantly improve the accuracy
of predictions, and uses simulation techniques to show that this may
be accompanied by less expensive sampling designs. The analysis,
nevertheless, also concludes that the proposed specifi cation is not a
panacea and affi rms that there is still scope for reducing nonresponse
bias, pointing to several issues for future research.