Performance on a range of spatial and mathematics tasks was measured in a sample of 1592 students in kindergarten, third grade, and sixth grade. In a previously published analysis of these data, performance was analyzed by grade only. In the present analyses, we examined whether the relations between spatial skill and mathematics skill differed across socio-economic levels, for boys versus girls, or both. Our first aim was to test for group differences in spatial skill and mathematics skill. We found that children from higher income families showed significantly better performance on both spatial and mathematics measures, and boys outperformed girls on spatial measures in all three grades, but only outperformed girls on mathematics measures in kindergarten. Further, comparisons using factor analysis indicated that the income-related gap in mathematics performance increased across the grade levels, while the income-related gap in spatial performance remained constant. Our second aim was to test whether spatial skill mediated any of these effects, and we found that it did, either partially or fully, in all four cases. Our third aim was to test whether the "separate but correlated" two-factor latent structure previously reported for spatial skill and mathematics skill (Mix et al., 2016) was replicated across grade, SES, and sex. Multi-group confirmatory factor analyses conducted for each of these subgroups indicated that the same latent structure was present, despite differences in overall performance. These findings replicate and extend prior work on SES and sex differences related to spatial and mathematics skill, but provide evidence that the relations between the domains are stable and consistent across subgroups. [This paper was published in "Cognition" v218 Article 104918 2022.]