Girls Code It Better
Last registered on April 17, 2019


Trial Information
General Information
Girls Code It Better
Initial registration date
April 15, 2019
Last updated
April 17, 2019 8:28 PM EDT

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Harvard Kennedy School
Other Primary Investigator(s)
PI Affiliation
University of Bologna
Additional Trial Information
On going
Start date
End date
Secondary IDs
In most countries, girls underperform in math and they are less likely to choose a STEM career. Exposure to stereotypes has negative consequences on girls’ self-confidence in math, their performance measured by standardized test scores, and their likelihood to choose advanced STEM tracks (Carlana, 2019). According to the stereotype-threat theory, one potential way to reduce the negative effects of exposure to gender stereotypes is by increasing girls’ self-confidence in math.

I will exploit an already ongoing project, "Girls code it better" (GCIB), a coding course targeting female middle-school students, to study the relationship between gender stereotypes and education. GCIB has been running since 2015 in Italy and I will focus on 14 school that have taken part to this project in this school year (2018-2019).

The purpose of my research is twofold: firstly, I aim at evaluating whether GCIB could be an effective channel to reduce gender stereotypes; and secondly, I would like to establish whether a role model intervention can positively affect girls' willingness to enrol in a coding course. I will exploit two separate RCTs to answer to these two questions.

To evaluate the impact of GCIB on students, I will measure certain outcomes such as their self-confidence, perception of gender stereotypes, choice of the field of study and performance in school. I will exploit three sources of information: a survey, the results of an Implicit Association Test concerning school subjects and gender stereotypes, and administrative data provided by the Ministry of Education on pupils' school grades and high school track choices.
I would like to understand whether girls become less vulnerable to the stereotype threat after attending the project GCIB. To achieve this, I will compare the outcomes of girls who attended GCIB during the 2018-19 school year (treatment group) with those of the girls who applied and were not admitted (control group). It is important to underline that the girls that applied to the project have been randomly selected in the treatment or in the control group.
Furthermore, I will also analyze the impact of GCIB on the peers of treated and control students in terms of gender norms and educational choices, to assess the presence of eventual spill-overs.

To tackle the second point, I will evaluate the impact of two different role models on the girls' willingness to enrol in coding activities or courses. The treatment is randomly assigned at the classroom level: one group of classes will be presented with the story of a successful female professor, a second group will read information about a peer who did last year GCIB, while a third group will serve as a control. By comparing the willingness to take a coding course across the three groups, I will identify the most effective intervention.

External Link(s)
Registration Citation
Carlana, Michela and Margherita Fort. 2019. "Girls Code It Better." AEA RCT Registry. April 17.
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
There are two interventions with different aims.

The first intervention concerns only the female students who applied to participate to the coding class "Girls Code It Better" (GCIB) and that were randomly selected in the treatment (20 individuals) or in the control group. The treatment is attending the course itself in the school year 2018-2019.

The second intervention is at the class level and concerns all the students. During the administration of the questionnaire, some classes will be randomly exposed to two different role models (one each). The aim of the intervention is to understand whether telling the students the successful stories of women who engaged in STEM-related activities encourages girls' to participate in coding courses.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Gender-STEM Implicit Association Test; Questions on explicit stereotypes; willingness to participate in a coding class; chosen high school track
Primary Outcomes (explanation)
The Gender-STEM IAT considers the milliseconds it takes to individuals to associate a world to a certain group. The words to associate are male and female names and scientific or humanistic school subjects and they have to be associated to as fast as possible to a group of two words. In one case the groups to choose between are "female + humanistic" and " male + scientific", while in the other case the groups are "against" the stereotype and invert the above categories: "female + scientific" and "male + humanistic". By comparing the time it takes to associate the words in the two cases, it is possible to obtain a measure of the impact of gender stereotypes.

Questions on explicit stereotypes ask whether: – a) there are innate differences in math abilities between boys and girls b) engineering is not a suitable job for a woman c) girls are as likely as boys to perform well in scientific high-schools d) nursing is a job suitable for men, etc
Secondary Outcomes
Secondary Outcomes (end points)
Network data; educational choices and preferences
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
There are two main experiments, one involving only the female students who applied to participate in the course "Girls Code It Better" (GCIB) and one concerning all the students interviewed. They are both RCT.

In the first case, female students who apply to participate in GCIB are randomly accepted or rejected. Up to 20 students per school are enrolled in the course, while the others serve as a control. The course started at the beginning of the school year and will finish by the end of April. The evaluation of its impact on the above-mentioned outcomes will take place right after.

In the second case, the treatment is randomly assigned at the classroom level. During the completion of a survey to evaluate the impact of GCIB, students will be presented with one of the two stories or nothing, according to the group the class is selected into. The survey concludes by asking students whether they would be interested in undertaking a coding course, which is the main outcome for this second RCT.
Experimental Design Details
Not available
Randomization Method
The randomization aimed at selecting the students into the course was done in the office by a computer: a number was randomly assigned to each student and those with the associated smallest number were selected.

The randomization concerning the role model treatment is performed by Qualtrics, the platform that we use to administer the survey. We randomly associated a treatment status to each cohort and each section. When the students select their year and class, they are inserted in one of the three groups.
Randomization Unit
The randomization of the first RCT, the one concerning the girls' participation in GCIB, was done at the individual level.

The randomization unit of the second one, involving the role model, was done at the class level.
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
For the first RCT, the treatment was clustered by schools and there are a total of 14 schools.

For the second RCT, the clusters are the classes and there are a total of 221 classes.
Sample size: planned number of observations
For the first treatment, the main sample is composed of the girls who applied to participate in GCIB. There a total of 357 female students. We also consider some outcomes concerning the totality of the students in each school, a total of 5015 pupils. For what concerns the second experiment, we consider all the students in the schools, so 5015 individuals.
Sample size (or number of clusters) by treatment arms
For the first RCT, 280 students are treated and 77 are among the controls (Intention to treat).

For the second RCT, on average 74 (about 1675 students) classes will be exposed to treatment A; 74 classes (about 1675 students) to treatment B and 74 (about 1675 students) classes to the control group - about a third per group.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
University of Bologna
IRB Approval Date
IRB Approval Number