A pragmatic measure of student-reported school climate: How low can you go?

There continues to be great interest in efficient and valid methods for assessing school climate for both research and practice purposes. Consistent with this special issue on pragmatic measurement, this paper distills a previously validated 30-item measure (Bradshaw et al., 2014; Bradshaw, Waasdorp, Debnam, & Lindstrom Johnson, 2014; Lindstrom Johnson, Reichenberg, Shukla, Waasdorp, & Bradshaw, 2019) into a simplified version of the three core domains of Safety, Engagement, and Environment, consistent with the U.S. Department of Education model. Data come from a sample of 69,513 students within 114 middle and high schools. We used machine learning (ML) and item response theory (IRT) approaches to derive a set of two pragmatic measures based on the longer Maryland School Climate Survey, resulting in a 9-item (i.e., 3 items per domain), and an even briefer 3-item measure (1 item per domain). Results indicated that the 9-item measure maintained high concurrent validity with the full scale (latent trait range: 0.83-0.93) and the combined IRT-ML approach yielded superior model fit and content coverage compared to IRT alone. The 3-item indicators also showed strong correlations with original domain scores (ρ range: 0.75-0.83). Together, the results suggest that the 9-item measure may be most useful for research and school-district assessment, whereas the 3-item measure may be potentially more useful as a collection of single-item indicators of Safety, Engagement, and the School Environment.