This research project will advance knowledge about nonresponse error introduced by different modes of data collection for a wide variety of survey measures. The results of this research will contribute to the development of surveys employing multiple modes, thus enabling greater public participation and strengthening survey data quality. Population-based surveys are used to measure everything from unemployment to the health of the nation, guiding policy decision making at national, state, and local levels. But responses to surveys in general, and telephone surveys in particular, have declined significantly over the past two decades. As a result, researchers are exploring multiple modes of survey data collection. This project will provide additional information about the magnitude of nonresponse error among alternative modes of data collection.
This project will estimate nonresponse error directly rather than inferring it from demographic differences between those who respond and do not respond to different survey protocols. Nonresponse error will be measured for two modes of survey administration: computer-assisted telephone interviewing (CATI) and telephonic interactive voice response (IVR). To minimize mode effects, questions for the test surveys will be selected among those least likely to be impacted by mode. These questions will be chosen among questions included in several federal surveys, such as the Health Information National Trends Survey, National Household Travel Survey, the Behavioral Risk Factor Surveillance System survey, the National Crime Victimization Survey, the National Immunization Survey, and the Current Population Survey. Using an address-based sample frame, residents in diverse neighborhoods in the Boston metropolitan area will be randomly assigned to one of two experimental survey modes: CATI or IVR. At the end of survey field period, nonrespondents will be assigned to in-person interviewers. The interviewers will contact household members who did not respond to the survey, inviting them to complete the survey. Monetary incentives will be provided to encourage response. In-person interviews typically have the highest response rates among any survey administration mode. Data from CATI and IVR nonrespondents will be used to assess nonresponse error for key measures included in the survey. The resulting combined samples will provide direct estimates of the variables included in the survey with a minimum of nonresponse error against which to compare estimates from the two survey administration modes. The project is supported by the Methodology, Measurement, and Statistics Program and a consortium of federal statistical agencies as part of a joint activity to support research on survey and statistical methodology.
