Different scales, different results? How is survey responses influenced by the employed scales
The paper analyses the results of a series of split ballot experiments included in three representative surveys fielded in 2016 and 2017. The focus of the experiments is to compare alternative versions of various scales to test the validity of the intensity hypothesis and to compare fully labelled scales with end labelled ones in a Slovak language questionnaire. We find strong support in favour of the intensity hypothesis observing less answer to the more intensely formulated answer categories in the tested rating scales. It seems, however, that to identify the effect on a five categories agree – disagree scale reliably, a sample larger than N = 1200 is necessary. Thought relatively small, the effect becomes more notable if respondents tend to choose answers mostly towards the scale endpoints. We demonstrate how multivariate analysis results can be substantially influenced by minor scale modifications. Our experiment comparing a fully labelled scale to an end labelled scale supports the hypothesis that end labelled scales make the endpoints more salient and are more often selected by the respondents.
Sociologický ústav SAV, Klemensova 19, Bratislava, Slovak Republic
Publication order reference