The present study challenges the assumption that equipping students with positive learning attitudes and beliefs can compensate for socioeconomic status (SES) effects on students' academic achievement. It unravels the association between SES and students' achievement by examining direct and indirect SES influences (via students' science attitudes and beliefs such as science epistemological beliefs, interest, and self-efficacy) on students' science achievement using an analysis of secondary data involving 5,355 15-year-old students (from 138 schools) and their parents from Hong Kong who participated in the Programme for International Student Assessment (PISA) 2015. Two-level structural equation modelling (SEM) results showed that (a) the total standardised SES effect (direct and indirect) on students' science achievement was 42.46% more when compared to the case where only direct SES effects were accounted for; and (b) the total standardised effect of science classroom variables was much smaller than that associated with SES. These results suggest that the strength of the association between SES and students' achievement is understated if we focus only on direct effects. Relatedly, students' attitudes and beliefs are not insulated from familial SES influences, so they are not as efficacious in circumventing social structures as they are sometimes portrayed in the literature.