This is the study I believe they are referring to: http://cogdevlab.uchicago.edu/sites/cogdevlab.uchicago.edu/files/uploads/Maloney%20et%20al%20(2015)%20Interge

This is the study I believe they are referring to: http://cogdevlab.uchicago.edu/sites/cogdevlab.uchicago.edu/files/uploads/Maloney%20et%20al%20(2015)%20Intergenerational%20effects.pdf http://cogdevlab.uchicago.edu/sites/cogdevlab.uchicago.edu/f... (PDF) > we tested the interaction between parents’ math anxiety and the frequency of parents’ homework help while controlling for students’ grade, gender, beginning-of-year math achievement, and beginning-of-year math anxiety So the study controls for baseline performance. It's definitely possible that some selection bias remains, but it looks like a reasonably well-executed study on a large sample (438 children) and I wouldn't discount the results out of hand.