Encuestas controladas por cuotas: una aproximación empírica a su error total

Abstract

The majority of survey data available for public opinion research and electoral forecasting through representative samples, in Spain and Latin America, has been obtained with the final selection of interviewees through quotas, a non-probabilistic method. The reliability of the estimates made from this data is poorly understood and has little theoretical support: we know quite a bit about its biases, but almost nothing about its variance and "margin of error." This research empirically studies the variability of the samples obtained in the periodic barometers of the Center for Sociological Research of Spain (CIS), taking advantage of the length and regularity of some of its series estimating stable parameters. We find that the growth of variance is moderate and is within what is expected in a complex design.