Participation Decisions and Measurement Error in Web Surveys

Survey Methodology;Web Surveys;Survey Nonresponse;Measurement Error;Breakoff;Number of Questions Per Page;Survey Methodology

Peytchev, Andrey A.Schwarz, Norbert W. ;

University of Michigan

关键词: Survey Methodology;    Web Surveys;    Survey Nonresponse;    Measurement Error;    Breakoff;    Number of Questions Per Page;    Survey Methodology;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/96497/Peytchev_dissertation.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This dissertation examines participation and question answering decisions in web surveys, in a unified framework. These decisions include whether the respondent will choose to start the survey, whether to continue or break off, whether to provide an answer to each survey question, and how much effort to exert in answering the questions. First, nonresponse is found to be highly correlated within a person; individuals who do not participate in one web survey are far more likely to also not participate in another web survey. Several survey-related factors are identified that further influence nonresponse. More surprisingly, independent survey requests are found to contribute to nonresponse—asking someone to complete a survey makes them less likely to agree to complete a later survey, unrelated to the first. Breakoff, however, seems to be more dependent on the survey than specific to a person. Breakoff is found to be higher on pages asking questions on sensitive behaviors and complex layout, particularly the use of grids. Interestingly, breakoff is found to be also higher on pages merely displaying section introductions, much like in telephone interviews. Finally, a series of experiments investigate the impact of placing multiple questions per page, and in grids. Greater measurement error is found when multiple questions are placed in a page, and even greater when they are also displayed in a grid. Evidence is found in support of a perceived burden hypothesis due to layout.