The Impact of Providing Incentives to Initial Telephone Survey Refusers on Sample Composition and Data Quality
Currivan, D. (2005, May). The Impact of Providing Incentives to Initial Telephone Survey Refusers on Sample Composition and Data Quality. Presented at American Association for Public Opinion Research 60th Annual Conference, Miami Beach, FL.
Although much research has examined the impact of standard incentive plans on survey response, relatively few studies have evaluated the impact of offering incentives to initial refusers on the final composition of the sample, the distribution of survey responses, or the quality of survey data. As usually intended, offering an incentive payment based on initial refusal may increase the participation of sample members who would otherwise be difficult to include in the survey. Research also suggests that incentives may persuade sample members who have little interest in the survey topic to participate. While greater participation among less interested sample members may reduce unit nonresponse bias, respondents with little interest in the survey topic might also be less motivated to provide complete and accurate responses than other sample members. Combined, these two factors could result in increased participation from some sample members, but poorer quality data from these respondents. To investigate the impact of offering incentives to initial refusers, this study evaluates an incentive protocol currently being implemented in a statewide telephone survey, the New York Adult Tobacco Survey. In this survey, respondents are not initially offered any incentive to complete the interview, but are then offered $20 if they initially refuse to participate. Comparisons of key characteristics and survey responses among sample members who received the incentive versus those who did not indicate how this payment influenced the composition of respondents. Comparisons of response patterns to key survey items among paid and unpaid respondents including analysis of missing data, dont know answers, and other indicators of low respondent effort indicate whether the incentive affected data quality. The implications of these analyses for similar incentive protocols are discussed.