Dag van de Sociologie edition:8 location:Leuven date:29 mei 2008
As respondents move through a survey, fatigue sets in and motivation levels are likely to drop. Consequently, higher satisficing levels are expected in later parts of a survey than in earlier ones. Since satisficing could be conceived as taking cognitive shortcuts, shorter response times and lower data quality are expected in the presence of satisficing. Given the general hypothesis of increasing satisficing levels as a survey progresses, we thus formulate two hypotheses. Firstly, we expect to find shorter response times at later stages of the survey. Secondly, we predict lower levels of data quality at the end of a survey compared with in the beginning. To test the hypotheses, we will use data from a web survey experiment in which the order in which the questions are presented are manipulated. Collection of client side paradata will allow us to compute exact response times for each question. By doing so, we can compare the response times for each question at different stages in the web survey and thus test the first hypothesis. To test our second hypothesis, we will take a closer look at indicators of lower data quality such as more non-substantial answers (don't know responses), a less differentiated answer pattern on rating scales and more rounding of numeric responses.