Protest surveys are increasingly used to tackle questions related to participation in social movements. However, it is unclear whether they generate useful and valid data. This study puts the protest survey design to the test by relying on data of 51 demonstrations (2009–2011) in seven European countries. We use data on 15,000 protest participants combined with screener questionnaires and extensive debriefing records of the interviewer teams. We account for noncontact (fieldwork problems), immediate and delayed refusal, and refusal bias. Results show that fieldwork problems are frequent, that immediate refusal is low, and that delayed refusal is considerable. Systematic refusal bias is only found for age and education. Differences between countries and protest issues are small but the issue determines the composition of an event which, in turn, leads to higher or lower refusal. Researchers should be cautious when using protest survey data to compare protest events across issues. The paper pleads for standardization and constant monitoring of the data-gathering process.