An Evaluation of Nonresponse and Coverage Errors in a Prerecruited Probability Web Panel Survey

This study examines nonresponse and coverage errors separately in a probability Web panel survey by applying traditional postsurvey adjustments. This was done by using variables whose estimates were obtainable at both the survey respondent and the full survey sample levels and whose values were known for both the full survey sample and the target population. Nonresponse error measured by the differences between the estimates from the respondents and the known full sample values was not found to be large, implying that nonresponse error in this Web survey data may not be critical. However, coverage properties of the full survey sample show some problems, and traditional postsurvey adjustments were limited in alleviating the unequal coverage of the survey sample. This coverage problem was more evident for the subpopulation-level estimates.