Researchers developed a machine learning (ML) model to help predict household response behavior in the Medical Expenditure Panel Survey (MEPS), a longitudinal survey that follows households for 2.5 years. The model combines information from multiple sources, including American Community Survey (ACS) data, Advance Call Records (ACR), and early survey contact paradata, to predict the likelihood that households will respond to upcoming contacts and complete their first-round interview.
By identifying households that may need additional support and helping researchers determine the most effective contact approach, the ML model can help survey teams use field resources more efficiently and support efforts to sustain high response rates. The study also demonstrates how combining ML with diverse data sources can improve survey field operations.
Learn More
Predicting “Yes”: Machine Learning and Diverse Data to Boost Respondent Cooperation
Authors: Rashi Saluja, Hanyu Sun, Gizem Korkmaz, Jill Carle, Ryan Hubbard, Brad Edwards, Rick Dulaney