Effects of number of response options in web surveys: The role of verbal labels
Relevance and research question: Fully labelled agree/disagree rating scales are often used to obtain data pertaining to subjective phenomena in web surveys. Although the link between the number of response options and the quality of information obtained with rating scales is well established, the role of the verbal labels attached to the options is less well studied. The verbal lables define the length and the fineness of a fully labelled rating scale and adding response options may therefore either lengthen the scale and/or make it finer. The aim of the present work it to examine the impact of altering the number and verbal labels of response options on the quality of the information obtained. Fully labelled rating scales with five, seven and nine response options are examined. Verbal labels are either used to lengthen the scales or to make them finer.
Method and data: University students in Germany and Iceland were presented with six questions measuring attitude towards university education in a web survey. Respondents were randomly assigned to six experimental conditions. 1) Five response options with the labels strongly, somewhat and neither/nor, 2) five response options with the lables fully, somewhat and neither nor, 3) seven options with the labels fully, strongly, somewhat and neither nor, 4) seven response options with the labels strongly, somewhat, slightly and neither/nor, 5) seven response options with the labels fully, somewhat, slightly and neither/nor and finally 6) nine response options with the labels fully, strongly, somewhat, slightly and neither/nor.
Results: The results show that the effect of the number of response options on the quality of the information obtained depends on whehter the added response options lengthen the scale or make it finer. A finer scale seems to reduce non-differentiation and extreme response style.
Added value: The findings show that the verbal labels attached to the response options must be taken into account in empirical research on the optimal number of response options.