Pew Education Frame Consultation
Problem
Sampling college students for research studies is complex and costly.
College students pose a unique problem in research due to their relatively low share of the population and transient nature. Consequently, identifying and finding students for research studies increases the complexity of sampling methods and locating efforts. Moreover, traditional source files, such as the United States Postal Service’s Computerized Delivery Sequency File (CDS), are limited in their ability to identifying housing occupied by undergraduate college students even when enhanced with market research data. While being an important part of American society, students are thus difficult to include in survey-based data collection studies without specialized targeting.
Solution
NORC and Pew created an improved sampling frame for future data collection efforts targeting undergraduate college students.
During the fall of 2024, NORC and the Pew Research Center undertook a study to address accurate coverage of undergraduate college students. Our study evaluated the degree to which the United States Postal Service’s Computerized Delivery Sequency File (CDS) includes or “covers” college students who live in dormitories. We also examined the potential of creating a model to identify off-campus locations more likely to contain college students.
Drawing on the Integrated Postsecondary Education Data System (IPEDS) data file and data from the American Community Survey Public Use Micro Data Area Sample, we created a frame of postsecondary institutions, with a full list of all possible on-campus residential mail delivery points as covered by the CDS. In addition, we tested various models to predict the incidence of college students at the small area level.
Result
Our modeling demonstrated promise in identifying Census blocks with higher incidence of college students.
Our study demonstrated that college student coverage may be haphazard enough to limit the usefulness of solely depending on data sets such as the CDS for targeting. However, it was able to successfully develop a promising model for identifying Census blocks with higher incidence of college students. While a more robust validation approach is needed to understand limitations of this approach, our study illustrated that, given additional time and resources, models considering type of institution, nature of the state and region, and other contextual variables are beneficial to stratification at the Census block level and more accurately identifying the target population.
Project Leads
-
Ned English
Associate DirectorProject Director -
Katie Archambeau
Senior Data Scientist IILead Methodologist
