Leveraging Nces As a Source of Big Data for Prevention Research: How Is School Security Related to Student Behavior and Achievement?

Introduction: The National Center for Education Statistics (NCES) provides a central and trusted repository of large education datasets that are available to prevention scientists either as public- or restricted-use datasets. These existing databases offer a unique opportunity to researchers willing to access, harmonize, and compare data from various studies and datasets in the interest of developing and testing new theories related to prevention science. This presentation will discuss lessons learned from a recent IES-funded project that integrated data from multiple NCES datasets with the aim of understanding how school security activities are related to school safety and violence. The presentation will begin with a brief overview of the project and then discuss the unique benefits and challenges associated with these types of integrative data projects.

Methods: The integrative data project synthesized multiple waves of cross-sectional data from two large national surveys – the School Crime Supplement (SCS) to the National Crime Victimization Survey and the School Survey on Crime & Safety (SSOCS). The harmonized and pooled data included survey responses from 38,707 students from the SCS (51% male, 77% White, mean age = 14.72) and 10,340 school administrator responses from the SSOCS (average school composition of 50% male, 57% White). Statistical analyses were estimated separately for the SCS and SSOCS pooled samples, but using parallel analytic strategies based on construct typologies.

Results: The results from the integrative data project indicated that there was no significant association between school security activities and students’ academic, behavioral or perceived safety outcomes; however, the pattern of results was inconsistent across the two datasets. The presentation will provide a brief overview of those empirical findings, but will focus primarily on several key lessons learned during the project: (1) identifying common data elements at the study planning phase, (2) harmonizing data effectively across diverse sources, (3) considering statistical modeling approaches appropriate for pooled or integrative data analysis, and (4) remaining vigilant about privacy, confidentiality, and data security.

Conclusions: NCES currently houses numerous datasets that could be pooled and synthesized to answer critical questions in the field of prevention science using a big data approach. Despite some of the procedural, technical, and methodological complications inherent in this type of work, the benefits of increased precision in predictive modeling will often outweigh those costs. We conclude by discussing examples of the types of innovative and groundbreaking prevention science questions that could be addressed by secondary analyses of data collected and distributed by NCES.