By Michela Carlana

In most countries, girls are lagging behind in math performance compared to boys, with negative implications for their readiness for STEM (Science, Technology, Engineering and Math) universities and occupations (OECD,2014, Card and Payne, 2017). The underrepresentation of women in these highly profitable fields may be fueled by social conditioning and gender stereotyping (Nosek et al., 2007, Guiso et al., 2008). In my job market paper, I study whether exposure to teachers’ gender bias affects math achievements, high-school track choice, and self-confidence of boys and girls.

**How do I measure teachers’ gender stereotypes?**

A crucial challenge of this research is the measurement of teachers’ gender stereotypes. In my paper, I focus on implicit bias, as measured by the Gender-Science Implicit Association Test (IAT). This test is computer-based tool developed by social psychologists, which exploits the reaction time to associations between fields of study (Scientific and Humanistic) and gender (male and female names) (Greenwald et al, 1995). The idea underlying the test is that the easier the mental task, the faster the response production and the fewer the errors made in the process.13 The IAT requires the categorization of words to the left or to the right of a computer or tablet screen and it provides a measurement of the strength of the association between two concepts – specifically in the Gender-Science IAT, gender and scientific/humanistic fields. Recently, IAT scores have been used also by economists when studying gender and race discrimination: despite being noisy measures, IAT scores have been shown to predict relevant choices and behaviors in lab experiments and in real-world interactions (Rooth, 2010, Reuben et al., 2014, Burns et al.,2016, Glover et al., 2017). Implicit stereotypes can operate without awareness or intention to harm the stigmatized-group (Bertrand et al., 2005). Even if teachers do not explicitly endorse gender stereotypes, their implicit bias may affect their interaction with pupils.

**What are IAT scores correlated with?**

I have collected AT scores on around 1.400 math and literature teachers in several provinces in Northern Italy. On top of implicit stereotypes, I administered a survey to teachers, which included demographic characteristics (age, place of birth, parents’ education, children age and gender, …), information about their teaching experience and explicit gender views. Interestingly, implicit bias is strongly correlated with own gender, field of study, and with the gender norms in the city of birth, as measured by the World Value Survey and by female labor force participation. Other characteristics, such as experience, gender of own children, teacher “quality”, have a small and statistically insignificant correlation with IAT scores.

I built an original dataset merging teacher surveys with administrative data from students, including standardized test scores, high-school track choice, parents’ education and occupation, and an additional questionnaire I administered on self-confidence in different subjects.

**Do teacher gender stereotypes affect math performance?**

Exploiting quasi-random assignment of students to teachers with different level of implicit bias, I find that exposure to teacher stereotypes substantially affects gender differences in math achievement. As in other countries, in Italy girls are lagging behind in math throughout the educational career (Fryer and Levitt, 2010): the average gender gap in math performance increases of 0.08 standard deviations between grade 6 and grade 8. Classes assigned to math teachers with one standard deviation higher gender bias have a 34% higher gender gap in math improvements during middle school, which corresponds to an increase of 0.03 standard deviations. All characteristics of peers and teachers are capturedy class fixed effects, as classmates attend all lecture together during grade 6 and 8. The results are robust to the inclusion of student and math teacher characteristics, interacted with pupil gender.

The figure below illustrates this result. I consider three groups of teachers according with their IAT score – those with a *“girls-math”* attitude (IAT score £ -0.15), *“no bias”* (-0.15“boys-math” attitude (IAT score³0.15). The gender gap in math performance does not significantly increase between grade 6 and 8 in classes assigned to teachers with a *“girls-math”* attitude, but it gets bigger the stronger the stereotypes of teachers.

**Are girls lagging behind or boys doing better when assigned to more biased teachers?**

I compare students by gender enrolled in the same school and cohort, but assigned to teachers with different level of stereotypes. The Figure below shows that girls lag behind when assigned to more biased teachers: the impact is linear throughout the distribution of teachers’ IAT scores. Boys are not significantly affected by gender stereotypes of their own math teacher.

Girls from disadvantaged backgrounds in terms of initial abilities are more negatively affected by teachers’ gender stereotypes. These results are consistent with the stereotype threat theory: individuals at risk of confirming widely-known negative stereotypes underperform in fields in which their group is ability-stigmatized.

**Teacher Bias affects Girls’ Self-Confidence in Math and Track Choice**

For a subsample of students, I collected information on their own assessment of ability in math. I find that biased math teachers activate negative self-stereotypes and induce females to believe that they are *worse at math* than what would be expected given their achievements. This result is important for three reasons. First, it shows that self-confidence of women in math is affected by social conditioning from teachers. Second, it is an important mechanism in order to to understand the effect of teacher bias on math performance of female students. Third, it may have potentially long-run consequences on girls’ educational choices. Indeed, I find that girls exposed to more biased teachers are more likely to attend less demanding high-school tracks: being assigned to a teacher with one standard deviation higher bias increases girls’ probability of choosing vocational training of 13 percent (which corresponds to 2 percentage point increase).

**Which are the policy implications of my research?**

In my job market paper, I provide evidence that implicit gender bias can form an unintended and often invisible barrier to equal opportunities in education, affecting math performance, self-confidence and track choice of girls.

These findings raise the question of which kind of policies should be implemented in order to alleviate the impact of gender stereotypes. The implicit bias measured by IAT scores should not be used to make decisions about others, as hiring or firing decisions. IAT scores are educational tools to develop awareness of implicit preferences and stereotypes. Hence, one set of potential policies may be aimed at informing people about own bias or training them in order to assure equal behavior toward individuals of ability-stigmatized groups and others. An alternative way to fight against the negative consequences of stereotypes is increasing self-confidence of females in math. My research in progress aims to investigate both types of policies.

Michela Carlana is a PhD student at Bocconi University and a LEAP affiliated student. She will start as Assistant Professor at Harvard Kennedy School in July 2018. More information about her research can be found on her personal website.