|
Abstract:
|
Nonresponse in surveys is a problem both in the science of survey methodology and the business of survey research. The fundamental difficulty arises from the fact that typically we record and analyze very little about selected sample members unless and until they respond. The process of administering a survey however now typically generates a great deal of ancillary data, paradata, that provide an opportunity to make inferences about the potential behavior of a sample member even when no response has yet been obtained. The talk will present a methodology that has shown promise, in face-to-face surveys in the US, for tackling simultaneously the business issue of predicting nonresponse and the science issue of understanding it. Possible implications for a broader conceptualization of data quality will be discussed.
|
