Balancing privacy and utility in synthetic data: Insights from Maryland’s State Longitudinal Data System
Abstract
Synthetic data hold strong potential to increase access to administrative data systems while protecting privacy for individuals. This paper details the approach taken to evaluate the synthesis of Maryland’s State Longitudinal Data System (SLDS) using fully synthetic Classification and Regression Tree (CART) models. Results demonstrate low disclosure risk (near zero) and high research utility, validated through robust evaluations. Practical insights and best practices from this case provide valuable lessons for other organisations seeking balanced synthetic data solutions for administrative data. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.
The full article is available to subscribers to the journal.
Author's Biography
Dr Michael E. Woolley is a Retired Professor of Social Work, University of Maryland, Baltimore.
Mark Lachowicz is a quantitative researcher at the American Institute for Research, where his primary interests are the design and analysis of studies of pre-K through post-secondary educational systems. He served as lead analyst for several Institute of Education Sciences (IES)-funded projects, including the evaluation of the impact of Montessori preschools and a meta-analysis of Social Emotional Learning programmes. Mark has extensive methodological experience in advanced analytic methods including multilevel modelling, structural equation modelling, meta-analysis and longitudinal data analysis. He earned his doctorate in quantitative methods from Vanderbilt University, where in addition to his methodological research he collaborated on several substantive projects such as the evaluation of the Tennessee Universal Pre-K Program. Post-graduation he joined the University of Maryland, College Park as a postdoctoral associate with the Synthetic Data Project, where he developed methods to evaluate a synthetic version of the Maryland Longitudinal Data System database.
Dr Terry V. Shaw’s background and interests focus on leveraging existing administrative data systems to improve state policy and practice related to child and family health, safety and well-being. Terry’s work examines the pathways into and through child serving systems focusing on opportunities for state systems to collaborate, understand service overlaps, improve overall service delivery and address the multiple needs of the children and families involved with these systems. Terry has worked with multiple state child welfare systems to and has experience developing and implementing federally funded research and evaluation projects: Terry was the lead evaluator on two Substance Abuse and Mental Health Services Administration (SAMHSA) funded Systems of Care grants; oversaw disclosure risk assessments for a Federal Institute of Education Sciences (IES) grant in Maryland; as well as multiple state and foundation funded examinations of child welfare engaged children and families.
Dr Daniel B. Bonnéry is a survey statistician, whose main interest is inference under informative selection.
Dr Angela K. Henneberger is a Research Associate Professor at the University of Maryland School of Social Work. Her research is situated at the intersection of education, developmental science and prevention science, with a focus on leveraging administrative data to investigate students’ academic, social and behavioural development within school contexts. She also serves as the Research Director and Principal Investigator of the Maryland Longitudinal Data System (MLDS) Center, where she leads a team of university-based researchers in conducting advanced statistical analyses to inform state and local policy. Angela earned her Ph.D from the University of Virginia as an Institute of Education Sciences (IES) Predoctoral Fellow and completed her postdoctoral training at Pennsylvania State University in the Prevention and Methodology Training (PAMT) programme.
Yi Feng is an Assistant Professor in quantitative psychology at University of California, Los Angeles. Her research is centered on creating and advancing innovative statistical tools that can meet the needs of increasingly more complex research questions in psychology and social sciences.
Tessa L. Johnson’s quantitative research is focused on the development of methodologies used in the social, behavioural and health sciences, especially in the area of population heterogeneity and complex hierarchical structures in longitudinal and multilevel data.
Dr Bess Rose is a Statistician with the University of Maryland School of Social Work and the Maryland Longitudinal Data System Center. Previously, she was a senior study director at Westat, a research and evaluation coordinator at the Maryland State Department of Education, a technical writing advisor at Goucher College and writing instructor at SUNY Buffalo. Bess’s current research interests include understanding the contributions of school and work environments to children’s and adults’ growth trajectories. She received her doctorate from the Johns Hopkins University School of Education, with support of an Institute of Education Sciences (IES) fellowship. In her dissertation she applied cross-classified multiple membership growth models to examine the impact of different types of school moves on academic achievement. Current research includes applying longitudinal data analysis methods to examine poverty as a predictor of wage trajectories.
Laura M. Stapleton is chair of the Department of Human Development and Quantitative Methodology and a professor of Quantitative Methodology: Measurement and Statistics. She previously served as the interim dean of the College of Education and Associate Dean for Research, Innovation, and Partnerships. She joined the faculty of the college in autumn 2011 after being on the faculty in Psychology at the University of Maryland, Baltimore County and in Educational Psychology at the University of Texas, Austin. She served as the Associate Director of the Research Branch of the Maryland State Longitudinal Data System Center from 2013–18. Prior to earning her Ph.D. in Measurement, Statistics and Evaluation, she was an economist at the Bureau of Labor Statistics and, subsequently, conducted educational research at the American Association of State Colleges and Universities and as Associate Director of institutional research at the University of Maryland.
