Purpose: Collective teacher efficacy (CTE) is a promising construct for understanding how schools can foster student achievement. Although much of the early research on CTE took place in North America, researchers from other parts of the world are now delving into this topic. The current study explores whether these powerful collective beliefs function similarly across diverse cultural and linguistic groups: Arab and Jewish teachers in Israel, and teachers in Turkey and the USA. Design/methodology/approach: Participants included 4,216 teachers from Israel, Turkey and the USA, representing four cultures--Arab, Jewish, Turkish and American. We tested configural invariance using multigroup confirmatory factor analysis (AMOS) and alignment optimization (Mplus) to identify the groups in which specific parameters are noninvariant, and to compare the latent factor means. Findings: Configural invariance showed adequate fit of the model structure across the four groups. Based on invariance tests, using the alignment optimization method, CTE scales held different meanings for specific items across the four cultures, where the USA and Arab cultures were the sources of these differences. Furthermore, in comparing the two-dimensional CTE belief scale across the four groups, latent means revealed the highest mean ranking for the USA and the lowest for Turkey. Originality/value: This research makes a significant theoretical contribution by examining and comparing the concept of teachers' collective efficacy in multiple cultures. This comparison can also contribute to instructional teaching practices worldwide.