Co wiemy o respondencie? Rozmiar i źródła błędów w pomiarze cech metryczkowych

Abstract

Questions concerning socio-demographic features of respondent (so-called background variables) are built into each questionnaire schedule. One should assume that asking them we receive information with error. This analysis aims at showing size of this error and its sources. I define error in terms of difference between answers given by the same respondents to the same questions in two interviews. Data comes from control research in national surveys, based on random samples, carried out in 1997-2000. All in all, it includes 7029 randomly selected cases from 83 surveys. The basic conclusion says about widespread scope of error. However, it varies in case of different variables. Errors result from: weaknesses of memory of respondents, tendency to give "socially desirable" answers, and lack of knowledge (e.g. about incomes of the family members). Results of this analysis are discussed against the background of comparable data coming from other research.