In social surveys, the elder respondents are more susceptible to the survey environment in answering questions, which may impair the quality of survey data and the validity of research results as well as policy decisions. This study examines the impact of bystanders in survey settings on the elderly's performance in answering questions, using data from the 2018 China Family Panel Studies(CFPS). We find that the number of bystanders negatively predicts the severity of item non-response, and positively predicts the severity of self-reported issues. Both effects are marginally diminishing as the number of bystanders increases. Significant group differences are observed in these effects, as the answering performance of female, younger, and less-educated elderly is more susceptible to the influence of bystanders. The "audience effect" in social surveys, as manifested in the elderly population, is statistically insignificant in other age groups. The study also finds that the bystander's identity affects the severity both of item missing and of self-reported issues among the elderly respondents. Overall, the bystander show a double-edged sword effect on the quality of questionnaire survey on the elderly, by enhancing the completeness of data reporting yet undermining the veracity of self-reported information.