This study drew on expectancy-value theory (EVT) to examine the relations between mathematics motivation
(academic self-concept and task values) and student background variables in predicting educational outcomes.
Using latent-variable models with latent interactions, we investigated the multiplicative effect of self-concept
and value, which is central to classic EVT. The mediating role of motivation and gendered patterns was also
explored. Hong Kong's TIMSS dataset for three cohorts (1999, 2003, and 2007) was used over a period where
the education system had experienced considerable changes, providing a strong test of the robustness of these
findings. The results suggested: (a) self-concept is more important for students with lower utility values in
predicting their educational outcomes; (b) while boys and girls had similar levels of math self-concept and
values, girls tended to have higher mathematics achievement and educational aspirations; (c) family socioeconomic
status is more strongly linked to educational aspirations for boys.