October 4, 2022

Evans School Research Team Tackles Common Challenge in Administrative Data

Scholars and policymakers increasingly utilize administrative data from public program systems to understand program trends, implementation, and impact. Often times, however, administrative data resources lack key pieces of information or are not linked in a manner that allows researchers to examine questions about interplay between different types of programs. 

For the last several years, a team of scholars at the Evans School and the UW School of Social Work have been building a large linked administrative data set that overcomes some of the common limitations. UW Social Work and Evans adjunct faculty Jennie Romich, along with Evans School faculty Mark Long, Heather Hill, and Scott Allard and doctoral students Callie Freitag and Elizabeth Pelletier, have worked to create the Washington Merged Longitudinal Administrative Data (WMLAD), which can be analyzed help answer a host of questions about employment, safety net program participation, and well-being. WMLAD links data from 2010 to 2017 across a number of Washington State agencies: Employment Security Department (ESD); Department of Social and Health Services (DSHS); Health Care Authority (HCA); Department of Health (DOH); Secretary of State (SOS); Department of Licensing (DOL); Washington State Patrol (WSP). 

Another common challenge confronting administrative data involves identifying household spatial location at regular intervals, while protecting the confidentiality and anonymity of households in the data. Not all administrative data contain address information and addresses may not be consistently updated over time.   

Recently, WMLAD team members Mark Long, Elizabeth Pelletier, and Jennie Romich developed an analytic strategy to construct regular spatial location information in instances where WMLAD has sporadic information about spatial location. This important technical work was featured in a Population Studies article entitled, “Constructing Monthly Residential Locations of Adults Using Merged State Administrative Data.”  The authors develop a simple, but powerful, algorithm for predicting monthly residential location when information about location may occur less often in the data. Of particular importance is developing a rigorous way to impute the timing of a residential move for an individual, when long spells exist between two different address entries and it is not clear when the move may have occurred. Not only is this work critical to generating meaningful insights about the interplay of work, social programs, and workplace regulation, but it serves as a model for other scholars to follow in similar administrative data settings.