Big Data Meets Survey Science

18Feedback Loop: Using Surveys to Build and Assess Registration‐Based Sample Religious Flags for Survey Research

David Dutwin

NORC at the University of Chicago, Chicago, IL, 60603, USA

18.1 Introduction

Survey research has long utilized data appended to sample sources to effectively identify and survey specific populations of interest. Telephone sample providers have built partnerships with credit history companies to link telephone numbers to information about the person or household associated with that telephone number. Is she Hispanic? Young? Affluent? Not only have marketers utilized such appended data to target audiences, but survey researchers have used such data to more effectively survey populations of interest.

Although some appended data comes from direct linkages, i.e. a sample file of phone numbers matched to a concurrent data base of credit histories via phone number, other data is model‐based. Examples are companies that aggregate voter registration lists¹ and provide registration‐based samples (RBSs) to survey researchers, marketers, and of course, political campaign officials and pollsters. The typical RBS datafile contains data both direct and modeled. Voter file data are appended with listed telephone numbers, credit information, and other available data available for linkage directly by name and address. But other data is modeled as well: some states report party registration of individuals, others do not, and for such states, party identification is a modeled ...