Background:
Large-scale studies, such as Trends in International Mathematics and Science Study, provide data to understand cross-national differences and similarities in students’ achievement and characteristics. In this study, we aimed to identify factors predicting mathematics achievement of Turkish students by comparing the achievement of Turkish and Australian students.

Material and methods:
First, construct equivalence and item bias were evaluated to check the comparability. Second, factors predicting mathematics achievement of Turkish and Australian students were identified. Then, propensity score matching was conducted to identify the remaining achievement differences between these countries, after matching students on putatively relevant background variables.

Results:
Results indicated that mathematics skills as measured by the TIMSS mathematics tests were free of construct bias for both groups of students. After removal of some biased items, it was possible to have an item bias free booklet. Additionally, students’ self-confidence and educational resources at their home were significant predictors of achievement. Propensity score analysis indicated that educational resources and self-confidence were indeed effective in explaining achievement differences between these two countries, with educational resources being more effective than self-confidence. Similar results were obtained when using TIMSS 2011 data.

Conclusions:
Implications for cognitive domains of mathematics and classroom activities are discussed.