Examining Measurement Invariance of a School Climate Survey Across Race and Ethnicity

Abstract

Positive school climate is a key determinant factor of students’ psychological well-being, safety, and academic achievement. Although researchers have examined the validity of school climate measures, there is a dearth of research investigating differences in student perceptions of school climate across race and ethnicity. This study evaluated the factor stability of a widely used school climate survey using factor analyses and measurement invariance techniques across racial/ethnic groups. Results of a confirmatory factor analysis indicated a five-factor structure for a school climate survey and weak measurement invariance was found across Hispanic, Black, and White student groups (ΔCFI = .008). According to paired t-tests, significant differences were found among racial/ethnic respondent groups across two factors: teacher and school effectiveness and sense of belonging and care. Validated school climate measures that are culturally and racially responsive to students’ experiences allow for accurate interpretations of school climate data. Discussion and implications are provided.

Systematic and evaluative review of school climate instruments for students, teachers, and parents

“…Acceptable internal consistency and test-retest reliability were obtained with values equal or higher to 0.70 (George & Mallery, 2003) for most of the reviewed measures except for a pair of instruments whose reliability did not reach 0.70 (DeRosier & Newcity, 2005;Whitehouse et al, 2020). Cronbach's alpha was the most popular measure of reliability reported but can be affected by number of items, number of response alternatives, and proportion of test variance (Domínguez-Lara & Merino-Soto, 2015;Watkins, 2017).…”

Longitudinal Measurement Invariance of the Authoritative School Climate Survey

“…Measurement invariance is concerned with evaluating the extent to which a set of items are measuring the same number of constructs with the same degree of accuracy across different conditions. In concurrent evaluations of invariance, these conditions generally take the form of membership in different groups such as biological sex, race, ethnicity, SES, or age (Richardson et al, 2007;Whitehouse et al, 2020). Here, invariance refers to the degree to which various estimates from a latent variable psychometric model are similar across different groups.…”