Coverage Bias and Sampling Error in a Study Using 1000-Series RDD Sampling

• Activity Number: 327
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): John Hall*+ and Karen Cybulski and Nancy Duda
• Companies: Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc.
• Keywords: Random Digit Dialing ; Sampling ; Bias ; Variance ; 1000 Bank ; 100 Bank

Abstract

Fahimi et al. (2008) concluded that traditional list-assisted RDD samples that use "1 +" (or working) 100 banks include only 80 percent of landline households while using a "1+" 1,000 bank sample would increase coverage of this group to approximately 90 percent. This paper presents results from a survey that used a Genesys RDD sample selected from "1+" 1,000 banks. A random subsample (50 percent) was processed to purge non-household numbers and flags were added identifying sampled numbers that would be included in a "1+" 100 bank frame. We examine the result of the Genesys processing and our calling results, to assess potential coverage bias, and discuss whether it would be more efficient (considering bias, variance, and cost) to rely only on 100 banks, solely on 1,000 bank sampling, or employ a strategy of oversampling the 100 banks.