Abstract

In this paper, the strengths and weaknesses of two Chilean political polls and the National Socioeconomic Characterisation Survey are analyzed from a statistical modelling point of view. The rationale of the analytical strategy is based on a distinction between identified parameters and parameters of interest. This is equivalent to make a distinction between what we can learn from the data provided by a survey and what we want to learn from those data. Using partial identification techniques, each survey is analyzed at different levels according to specific subpopulations. Based on these analyses, we emphasize not only the way in which the results should be reported, but also the necessity to make explicit the uncertainty induced by the non-response rates at the survey report.