Data Quality in Social Survey Research

Background

The attention to data quality measurement originated in psychometrics, whose roots are in the behaviourism paradigm, born in the early 1900s when the focus of the academic community of psychologists shifted from introspective methods to external environments in order to understand and explain human behaviour. Behaviourism has transmitted its appeal to social survey methodology by attempting to pursue “the comparability of responses” through the invariance of the stimuli and the standardisation of the interview situation (Fideli & Marradi, 1996, p. 74). The mechanical stimulus-response model represents the survey interview as follows: the stimulus (the questionnaire's question) is characterised by a high degree of structuring, and the respondents have to answer the same questions in the form and order designed by the researcher (Gobo, 1997). Fowler and Mangione summarise the core of the behaviourist manifesto in survey research: “the key element of measurement is standardisation. Standardisation aims to expose all respondents to the same question and response alternatives. In this way, differences can be interpreted as actual differences among respondents” (1990, p. 14). The highly standardised interview theorised by behaviourists requires establishing strict rules, especially regarding the interviewer's role, which is conceptualised as one of the leading causes of measurement error in the survey. In this approach, the perfect behaviourist interviewer has high skills to follow standardisation rules, administering stimuli designed by others within the cage imposed by standardisation in order not to introduce bias and errors during the interview (Biemer et al., 1991).