Table 7 Preregistered research questions, hypotheses, and findings

RQ1. Are there systematic differences in teachers’ evaluations of student performance (i.e., grading of student work), their evaluations of the effort they assume students put into the work, and their estimations of students’ mathematical ability that could be explained by the students’ gender or race?

H1. We hypothesize that we will not find any systemic differences in teachers’ evaluation of students’ written work (i.e., their ratings for the correctness of students’ solutions) based on the students’ gender or race, as found in prior empirical research (Copur-Gencturk et al., 2020). This prediction was based on theory positing that people tend to attribute the successes of members of stereotyped groups’ to effort and non-stereotyped groups to ability (Graham, 2017; Graham & Williams, 2009), and is consistent with previous findings (e.g., Copur-Gencturk et al., 2020; Fennema et al., 1990; Tiedeman, 2000, 2002). However, we expect to find that teachers rate the students’ ability higher when a White or male student name appears on work in situations where there is ambiguity in the work (i.e., when the students’ work is not completely correct) compared with a non-White (Hispanic/Black) or girl name, consistent with previous findings. We also expect that teachers will rate the effort of students’ work higher for female students’ work

✓ There were on average no differences in correctness ratings for gender (an effect size of 0.01 SD favoring girls, p = 0.59) or for race (an effect size of 0.02 SD favoring White students, p = 0.32)

✓ Teachers gave higher ability ratings to male names for not fully correct solutions (an effect size of 0.03 SD favoring boys, p = 0.03)

✕ No race-based differences in teacher ability ratings were detected (underpowered; an effect size of 0 SD favoring White students, p = 0.78)

✕ No gender-based differences in teacher evaluations of effort (an effect size of 0.01 SD favoring boys, p = .72)

RQ2. To what extent are teachers’ beliefs and dispositions (mathematics anxiety, beliefs about mathematical intelligence, levels of sexist beliefs, perceptions of being underestimated because of race or gender) related to such biases?

[We had no specific preregistered hypotheses for the modern sexism moderator. We added modern sexism to our model to explore whether teachers’ beliefs that gender disparities exist in current society were related to gender bias.]

_ Modern sexism moderated teachers’ gender-biased ability evaluations for not fully correct solutions (an effect size of 0.03 SD favoring boys, p = 0.01)

H2a. Teachers who have higher levels of mathematics anxiety may draw on their biases more often than other teachers when evaluating students’ work because mathematics anxiety can impair a person’s cognitive ability (e.g., Beilock et al., 2010; Ramirez, Shaw, & Maloney, 2018a, 2018b)

✕ Teachers’ gender-biased ability evaluations for not fully correct solutions were moderated by math anxiety but in the opposite direction (an effect size of 0.02 SD favoring girls, p = 0.04)

✓ Math anxiety moderated teachers’ race-biased ability ratings for not fully correct solutions (an effect size of 0.02 SD favoring White students, p = 0.07)

H2b. Teachers who believe that mathematical intelligence is fixed and innate will tend to show more gender bias because prior research suggests that teachers who believe that mathematical ability is fixed and innate also believe that boys, but not girls, have this ability (Copur-Gencturk et al., 2021)

✕ Fixed mindset did not moderate teachers’ gender-biased ability evaluations for not fully correct solutions (an effect size of 0.01 SD favoring girls, p = 0.57)