Model-Based Sampling Using an Online Probability Panel

• Activity Number: 210 - Contributed Poster Presentations: Survey Research Methods Section
• Type: Contributed
• Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Survey Research Methods Section
• Author(s): Meimeizi Zhu* and Angela Fontes and Justine Bulgar-Medina
• Companies: NORC at University of Chicago and NORC at University of Chicago and NORC at University of Chicago
• Keywords: sampling; modeling

Abstract

To help researchers better understand individual financial wellbeing and management, a pilot study was launched in September 2019. The study invited selected AmeriSpeak panelists to enroll in an online application to link their financial institutions/accounts. This allows researchers to passively track all account transactions in real-time.

In this present work, we conducted a model-based sampling method to select samples for the full launch of a financial transaction study in 2020. A model has been built to predict response rate, consent rate, and the enrollment rate for the rest of the panelists by analyzing the pilot data from the pilot study. The predictors hypothesized to have primary influence include the panelist’s demographic information, incentive levels, and financial profile information. Ultimately, panelists who have a higher predicted response rate, consent rate or enrollment rate were oversampled and invited to participate to maximize BAA enrollments. The ultimate success and accuracy of this approach will be presented.