That correlation is not invariant to cultural factors. Your own link mentions some of these - namely, "epistemic belief of learning, performance-avoidance goals, and parental education", for which we can easily measure variation across ethnicity (which of course is a lot more visible and more easily operationalized that a more nuanced understanding of culture, including urban vs. rural).