Conventional survey methodology wisdom holds that demographic questions should be asked at the end of surveys to minimize attrition. However, this reasoning may limit a growing body of research drawing from non-probability-sample web-based surveys that use social media for recruitment. These surveys often suffer from particularly high attrition rates but also rely on demographics to perform post-stratification and weighting.
In this report, we investigate the potential trade-offs between asking demographics early to obtain demographic information for a larger share of the sample and asking demographics last to reduce overall attrition. We experimentally and observationally test the effect of demographic item placement on survey attrition and examine its analytical implications in large-scale, web surveys administered by the Shift Project through Facebook.
We find no effect of demographic item placement on survey attrition but a substantial decrease in item non-response on demographic items placed at the beginning of the survey, providing more information of the sample’s demographic composition and increasing statistical power.