Westat Explores Use of Machine Learning to Improve Survey Response at GASP 2025

Westat researchers will present groundbreaking work on integrating machine learning (ML) to enhance the efficiency of survey research at the 2025 Government Advances in Statistical Programming (GASP) virtual conference, taking place June 24-27, 2025, from 12–5 pm (ET) each day. GASP is an informal, collaborative event that highlights the contributions of government data producers and users, with a focus on practical demonstrations and knowledge sharing to support federal missions. This event is sponsored by the Data Science for Federal Statistics (DSFS), an Interest Group of the Federal Committee on Statistical Methodology (FCSM).

Westat’s Rashi Saluja, a data scientist and the lead author, will present “Predicting ‘Yes’: ML & Diverse Data to Boost Respondent Cooperation.” Coauthors include Westat colleagues Ryan Hubbard, Jill Carle, Hanyu Sun, Gizem Korkmaz, and Brad Edwards. Westat’s Benjamin Schneider will chair the session.

The paper introduces a novel predictive modeling approach designed to increase response rates in household surveys. By leveraging diverse data sources—including demographics, call records, and early interactions with sampled households—the model forecasts the likelihood of survey participation. These predictions can guide more personalized outreach strategies and inform the most effective contact methods, ultimately supporting more efficient and targeted fieldwork.

“This study demonstrates the potential of machine learning to strengthen survey performance,” notes Saluja, “and offers a replicable framework for integrating varied data sources to improve engagement and data quality across survey research.”