Detection of validity-related faults in the modern electoral studies

Abstract

Amidst a number of high-profile errors in election forecasts, the problems of election forecasting and quality of data obtained from voters have recently attracted much attention. As the investigation of polling errors suggests, professional and academic communities are mainly focused on the accuracy of measurement, whereas the validity problems are generally underestimated. The present study is the first attempt to systematically describe the sources of erosion of validity of pre-election survey estimates. The description is based on theoretical framework and the analysis of fundamental (imperfection of the theoretical model itself), methodological (incomplete and incorrect practical use of the model) and methodical (mismatch between the measurement tools and the model requirements) faults. The author’s analysis helps to define a «task» aimed to reduce the faults and to draw up the key directions for improvement to be made in election polling.