Do Cash Incentives Bias RDD Studies? Examination of Results from a National and Statewide Survey

With the ever increasing difficulty of conducting random digit dial (RDD) interviews in the United States, it is necessary to evaluate the effectiveness of offering cash incentives to participants. The main objective in offering cash incentives is to increase response and cooperation rates. However, independent from cooperation we must also consider if offering incentives changes the composition of the sampling units (individuals), characterized by demographic variables and survey responses. We examined the results of offering incentives in two omnibus RDD studies (the Cornell National Social Survey and the Empire State Poll) in order to determine the impact of incentives on the demographic characteristics of respondents and actual responses to substantive questions. We use a logistic regression model to evaluate differences between the subgroups and find that offering an incentive does not bias the sample composition as a whole. There is some evidence, however, that incentives are more effective for lower income and highly educated groups. In addition to offering incentives to participants, the statewide study also offered incentives to the interviewers. We find a significant increase in response rates, but not completion rates, during the period the incentive was offered.