According to the findings of Programme for International Student Assessment (PISA) 2012 mathematics study, organized by Organization for Economic and Co-operation Development (OECD), Macao-China has the highest percentage (~42%) of grade repeaters in its 15-year-old student population amongst the 65 participating economies. Three mediation mechanisms, taken together explaining 0.7 grade level of progression in schooling, have been identified to explain the effects of grade repetition on student mathematics performance (Sit, et al., 2015). The three mechanisms are: (1) Insufficient opportunity to learn; (2) Inadequate self-regulation of students; and (3) Inappropriate teacher guidance and management. Analyses of the Macao’s PISA 2012 data reveal that a large number of the grade repeaters are situated at the low end of the mathematics proficiency continuum (i.e. level 2 or below). Grade repeaters are therefore likely suffering from inadequate self-regulation of mental processes, and face problems with their perceived responsibility for failure in mathematics. In PISA 2012, examples of perceived responsibilities measured on a 4-point Likert response scale (i.e. very likely, likely, slightly likely, and not at all likely) are: (1) Not very good at solving mathematics problems; (2) Teacher does not explain the mathematics concepts well; (3) Make bad guesses on the quiz; (4) Course material is too hard; (5) Teacher does not get students interested in the materials; and (6) Just unlucky (OECD, 2013). PISA 2012 has developed an index (i.e. FAILMAT) for secondary analyses by the researchers, and higher values (positive) of the index correspond to the higher level of external attributions of failure such as bad guesses or the teacher; whereas lower values (negative) correspond to the less external attributions to failure such as mediocre abilities or the difficult course contents (OECD, 2013). Drawing data from the PISA 2012, this study seeks to examine the similarities and differences between Macao’s grade repeaters and non-repeaters concerning their locus of control for their academic failure in mathematics. The following hypothesis is postulated for statistical significance testing: Grade repeaters have a higher level of external attributions of failure in mathematics, such as bad guesses or the teacher, whereas the non-repeaters have a higher level of the less external attributions of failure in mathematics, such as mediocre abilities or the difficult course contents. Although the independent t-test fails to establish statistically significant difference between Macao’s grade repeaters and non-repeaters in FAILMAT (t =1.936, p =.053), statistically significant differences (p <.01) are established for two of the six perceived responsibility for failure in mathematics. Compared with the non-repeaters, grade repeaters are found more prone to attribute failures to “make bad guesses” (p <.001) and “not very good at solving problems” (p <.01). Also, grade repeaters have a slightly greater tendency than the non-repeaters to attribute failures to the teacher, i.e. that they cannot get the students interested in the learning materials (p =.056). Implications of the study are proposed to render help to Macao’s grade repeaters for the betterment of self-regulated learning in mathematics.