Using R-Indicators to Make Case-Level Decisions for GSS 2020
Steven Pedlow
NORC at the University of Chicago
NORC uses R-indicators on projects to monitor overall representativeness and to make operational adjustments to prioritize under-represented groups through future outreach in real-time. R-indicators come from regression modeling on a response variable. They include an overall R-score (overall representativeness), unconditional partial R-scores for each predictor variable subgroup, and a conditional partial R-indicator for each variable. The overall R-indicator score measures the variation in response propensities among all cases. At the same time, the partial R-indicators split the R-indicator score into between (unconditional) and within (conditional) variation. For the 2020 round, the General Social Survey (GSS) applied the case-level sum of the unconditional partial R-indicators to decide whether to target a case with extra effort and higher incentives, continue working a case or stop attempts to recruit a case. We collected a panel sample of 2016 and 2018 GSS respondents in summer 2020 and a new cross-sectional sample during winter 2020-2021. The panel was a unique opportunity to use R-indicators with specific information about every sample member.