Measuring learning outcomes on core skills is key to building evidence on which teaching and learning interventions are most effective, and ultimately informing policy that prioritises these skills, amongst early learners. However, developing national or systems-level assessments pose considerable challenges, including implementing an effective translation process from centrally developed tools. In developing nations, limited resources compounded the challenges. This paper investigated the effect of test, classroom, and home (or mother-tongue, MoT) language correspondence on the relationship between numeracy and literacy skills amongst early-grade learners (N = 4032) in high-poverty regions in the Philippines. Results show that language correspondence does not in itself have an effect on numeracy skill but positively moderates (i.e. strengthens) the relationship between literacy and numeracy skills. The findings suggest that targeted translation can effectively remove the effect of test language on the construct being measured, language match still exerts an indirect effect on the strength of relationship amongst the skills being tested together. The finding that language factors have no direct effect is dependent on the effectiveness of the targeted translation process, but also underscores its importance especially in linguistically diverse test settings. Finally, the impact of test language and implications for systems-level testing are discussed.