Background: This study was part of a large-scale quasi-experiment on the causal relations among feedback, self-directed learning, and mathematics achievement of primary students in Hong Kong. The larger study was sponsored by a General Research Grant awarded by Research Grants Council of UGC, Hong Kong (Project Number 844011). Aims and Keywords: The study aimed to develop and validate, using a Rasch approach, of an instrument for measuring mathematics achievement of Primary five students. Sample: The sample comprised 1,368 Primary 5 students from 16 Hong Kong schools who participated voluntarily. Methods: A pre-test/post-test design was used for the larger study. Data for this study were collected at the pre-test in form of a mathematics test completed during class-time. A 35-item mathematics test was developed after careful analysis of the local curriculum, and in consultation with teachers on the suitability of the test for students by the end of Semester One of Primary 5. Rasch analysis with the Winsteps software (version 3.70) was then used to validate the instrument. Results: The results showed that: (1) item-fit was between 0.5 and 1.5 for all items except one; (2) Eigen value of the first contrast of the Rasch PCA of Residuals was below 2.0; (3) Rasch person and item reliabilities were 0.80 and 1.00 respectively; (4) there was good alignment between item difficulty and student ability; (5) no gender-DIF was found in the items. Conclusions: Measurement validity is one of the most crucial factors for the overall validity of any research study. This study developed a valid and reliable instrument for measuring mathematics achievement of Primary 5 students.