I use a combination of blind and non-blind test scores to show that middle school teachers favor girls in their evaluations. This favoritism, estimated as individual teacher effects, has long-term consequences: as measured by their national evaluations three years later, male students make less progress than their female counterparts. On the other hand, girls who benefit from gender bias in math are more likely to select a science track in high school. Without teachers’ bias in favor of girls, the gender gap in choosing a science track would be 12.5% larger in favor of boys.