Matching Nonprobability Internet Panel Samples with Probability Samples

• Activity Number: 522
• Type: Topic Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Charles DiSogra* and K.P. Srinath and Andrew Burkey
• Companies: Abt SRBI and Abt SRBI and Abt SRBI
• Keywords: Panel ; Internet ; Matching ; Propensity ; Online ; Non-probability

Abstract

Surveying rare groups in a time-sensitive frame can be cost-effectively done using non-probability online panels. Although an unattractive proposition for a probabilistic researcher, it can be a practical alternative to costly traditional surveys. Abt SRBI has used non-probability sources to track national influenza vaccination rates for two such rare groups: pregnant women and health care personnel. Each sample is weighted to their respective set of available national benchmarks. A concern with the estimates based on these samples is the likely bias due to no known frame selection probabilities and the exclusion of those who do not use the internet or join online panels. Another concern is not being able to assign a measure of precision to the estimate. One approach to addressing these concerns is to match the online sample with a probability sample from the same population groups. The National Health Interview Survey of the general population is the best such available source. We demonstrate a propensity score method of matching based on Internet use. The matched sample estimates may help quantify the bias and lead to a method of assigning a generalizable measure of precision.