effects of respondent and survey characteristics on the response quality of an open ended attitude question in web surveys

• DOI: 10.12758/mda.2019.05
• Publication Date: 2019-07-15
• Authors: Schmidt, Katharina; Gummer, Tobias; Roßmann, Joss
• Keywords: open-ended questions; response quality; web surveys; multilevel modeling; satisficing; GLES 2009-2015; Erhebungstechniken und Analysetechniken der Sozialwissenschaften; cognition; Sozialwissenschaften, Soziologie; questionnaire; 05 social sciences; Online-Befragung; Fragebogen; Umfrageforschung; Antwortverhalten; 0506 political science; Methods and Techniques of Data Collection and Data Analysis, Statistical Methods, Computer Methods; survey research; öffentliche Meinung; public opinion; ddc:300; Messung; online survey; response behavior; measurement; Social sciences, sociology, anthropology; Open-ended questions, response quality, web surveys, multilevel modeling, satisficing; Kognition

Open-ended questions have a great potential for analyses, but answering them often imposes a great burden on respondents. Relying on satisficing theory as an overarching theoretical framework, we derived several hypotheses about how respondent and survey level characteristics, and their interactions, might affect the quality of the responses to an open-ended attitude question in self-administered surveys. By applying multilevel analyses to data from 29 web surveys, we examined the effects of respondent and survey level characteristics on three indicators of response quality: response length, response latency, and the interpretability of the answers. With respect to all three indicators, we found that more educated and more motivated respondents provided answers of significantly better quality compared to other respondents. However, the present study provides evidence that analyzing response quality exclusively with process-generated measures of quality may produce a misleading picture. Therefore, the addition of content-related indicators, such as the interpretability of responses, provides a more informative result. We found that the further the open-ended question was located towards the end of the questionnaire, the fewer interpretable answers were given. Our results also indicated that if the survey was carried out in close proximity to a federal election, responses were more likely to be interpretable. Overall, our study suggests that the characteristics at the respondent and survey levels influence the response quality of open-ended attitude questions and that these characteristics interact to a small degree.<br/>methods, data, analyses, Online First<br/>