The Effects of Features of Survey Measurement on Self-Rated Health: Response Option Order and Scale Orientation

Abstract

Self-rated health (SRH) is widely used to study health across a range of disciplines. However, relatively little research examines how features of its measurement in surveys influence respondents’ answers and the overall quality of the resulting measurement. Manipulations of response option order and scale orientation are particularly relevant to assess for SRH given the increasing prominence of web-based survey data collection and since these factors are often outside of the control of the researcher who is analyzing data collected by other investigators. We examine how the interplay of two features of SRH influence respondents’ answers in a 2-by-3 factorial experiment that varies (1) the order in which the response options are presented (“excellent” to “poor” or “poor” to “excellent”) and (2) the orientation of the response option scale (vertical, horizontal, or banked). The experiment was conducted online using workers from Amazon Mechanical Turk (N = 2945). We find no main effects of response scale orientation and no interaction between response option order and scale orientation. However, we find main effects of response option order: mean SRH and the proportion in “excellent” or “very good” health are higher (better) and the proportion in “fair” or “poor” health lower when the response options are ordered from “excellent” to “poor” compared to “poor” to “excellent.” We also see heterogeneous treatment effects of response option ordering across respondents’ characteristics associated with ability. Overall, the implications for the validity and cross-survey comparability of SRH are likely considerable for response option ordering and minimal for scale orientation.