Tom Smith National Opinion Research Center
The workshop will bring together experts and researchers from the United States, Europe, and other countries to discuss various approaches to deal with nonresponse bias, integrate these approaches, and recommend the best practice to the survey research community. Participants will include leading international statisticians and survey methodologists with expertise on the measurement of and adjustment for nonresponse bias; principal staff on major social-science survey projects; and representatives of major governmental and nongovernmental survey research organizations such as the US Census and Statistics Sweden. Workshop participants will have expertise in conducting cutting-edge research on nonresponse bias, directing social-science research programs, and managing major large data collections. Workshop sessions will include presentations summarizing and synthesizing the latest research on nonresponse bias, outlining the state of research on the topic and best prospects for further development, and identifying specific recommendations for both research and application.
Broader Impact
The workshop will produce a research agenda to test and refine the promising techniques that have already been identified to address nonresponse bias. It will also produce recommendations of the optimal ways that current best practices can be utilized to detect and reduce nonresponse bias.
from Sample Frames, Auxiliary Databases, Paradata and Related Sources to Detect and Adjust for Nonresponse Bias in Surveys Tom W. Smith NORC/University of Chicago The International Nonresponse Workshop was sponsored and supported by the National Science Foundation, the World Association for Public Opinion Research, and the International Association of Survey Statisticians. It was held in Chicago on June 2-3, 2011. The agenda of the workshop, including the list of attendees, appears in Appendix 1. This workshop report considers 1) how auxiliary data (AD) can be used to analyze nonresponse bias, 2) other benefits of using AD, 3) recommendations regarding using AD for nonresponse analysis relating to a) sample frames, b) linked databases, c) paradata, d) aggregate data, and e) AD in general, 4) propensity models, 5) resources for nonresponse analysis, and 6) nonresponse bias measures. Nonresponse is a serious problem in surveys and nonresponse bias is an important component in total survey error. Fortunately, multi-level AD from sample frames, linked databases, paradata, and related sources can be utilized to both detect and adjust for nonresponse bias. By fully exploiting such available data at both the case- and aggregate-levels, a better understanding of and adjustment for nonresponse bias can be achieved. Theoretical insight into the causes of nonresponse, empirical mastery of the various AD sources and variables, and statistical and methodological advances in the use of multi-level and multi-source models are needed to achieve optimal improvement. The recommendations of this workshop should facilitate that effort.
