Assessing the Measurement Equivalence of School Climate for LGBTQ and non-LGBTQ Students Using Population-Based Data

• In Event: 1-101 - Measurement Validation and Equivalence in LGBTQ Youth Research: Examples and Recommendations

Integrative Statement

School climate plays a critical role in students’ well-being and academic success (Cohen et al., 2009). Due to the widespread prejudice and discrimination, there is evidence that lesbian, gay, bisexual, transgender and questioning (LGBTQ) students are more likely to experience hostile school climates than their non-LGBTQ counterparts (Kosciw et al., 2016). Therefore, there is a reason to believe that LGBTQ students may experience their school environments as qualitatively different compared to their non-LGBTQ peers. Yet, no study has taken into account whether school climate scales – often used in survey research – are equivalent for LGBTQ and non-LGBTQ students. We tested the measurement equivalence of the school climate scale (Hanson, 2012) included in the California Healthy Kids Survey (CHKS), a population-based survey that has been widely used to establish LGBTQ youth population health disparities (Day et al., 2018; Perez-Brumer et al., 2017). The scale conceptualizes school climate as a second-order latent construct defined by four first-order constructs: caring relationships with adults (6 items), perceived school safety (2 items), meaningful participation (3 items), and school connectedness (4 items).

Data are from more than 2,000 California middle and high schools that administered the CHKS from 2013 to 2015. Using robust maximum likelihood estimation, we evaluated whether the basic model structure (configural), factor loadings (metric), and item intercepts (scalar) of the school climate scale were invariant (i.e., equivalent) for LGBTQ (n = 94,038; 11.5%) and non- LGBTQ (n = 720,895; 88.5%) students. Given the large sample size, measurement invariance was evaluated using -.002 change in comparative fit index (CFI) for nested models as a criterion (Mead et al., 2008).

First, a preliminary confirmatory factor analysis validated the second-order measurement structure of school climate as conceptualized in the CHKS (Figure 1). After configural invariance was demonstrated (Model 1 [M1], Table 1), we found empirical support for metric invariance of both the first- (M2) and second-order factors (M3). Likewise, we found invariance across groups on item (M4) and first-order factor intercepts (M5). Given the support for measurement invariance, group comparisons between LGBTQ and non-LGBTQ students on factor means or manifest variable averages of school climate scores, including firs- and second-order scores, are defensible. We found significant differences on the school climate factor (z = -0.18, p < .001, d = 0.11) and on the factors that constitute it (school connectedness: z = -0.26, p < .001, d = 0.10; caring relationships with adults: z = -0.15, p < .001, d =0.06; perceived school safety: z = -0.25, p < .001, d = 0.09; meaningful participation: z = -0.08, p < .001, d = 0.03) with LGBTQ students reporting lower scores compared to non-LGBTQ students.

The school climate measure has the same general factor structure and the same meaning for LGBTQ and non-LGBTQ students. Group comparisons supported the evidence that LGBTQ students report more negative experiences of their school climate and of the several aspects that form part of it. The discussion will consider implications for research and education.

Authors

• Salvatore Ioverno, University of Texas at Austin
  Presenting Author

• Melissa A Rodgers, University of Texas
  Non-Presenting Author