364 – Nonprobability/Web Sampling and Data Analysis
The Efficacy of Non-Probability Online Samples
Alan Roshwalb
Ipsos Public Affairs
Zachary Lewis
Ipsos Public Affairs
Robert Petrin
Ipsos Public Affairs
The use of non-probability online opt-in samples is very common in commercial marketing and survey research practice. The use of non-probability online samples in experimental studies is well understood and without concern. When their use is for population estimation, it is here that classical survey statistical theory has concerns in their use. Given the cost, speed and flexibility of online surveys, non-probability samples are being used more and more for the measurement and tracking of attitudes and behaviors. In this study, we examine two non-probability online sample surveys for the purpose of exploring strategies to adjust and calibrate non-probability samples. Post-survey weighting is examined including standard raking ratio, propensity, and Generalized Regression weighting. The different auxiliary variables used in the adjustments include standard demographics, technology adoption, and components of an attitudinal-behavioral consistency strategic model. The attitudinal-behavioral consistency model motivates the discussion and provides guidance for measures to include in adjustments. It is under this framework that we examine the efficacy of non-probability online samples.