Social identity considerations and norms may affect behavior and preferences of disadvantaged groups, perpetuating gaps in economic outcomes. In this research, we want to understand whether there are any pre-existing biases or missing information that preclude women from attempting to apply for training and a career in the high growth technology sector in Peru. These could come from misperceptions on women’s abilities to pursue a career in the tech sector, or from lack of role models or appropriate networks. In order to understand what are the barriers to pursuing a career in the tech sector, we are randomly varying the recruitment message to potential interested applicants to a 5 month “coding” bootcamp and leadership training program, offered only to women. In addition to a control group message with generic information, in a treatment message, we correct misperceptions about women’s abilities to pursue a career in technology, provide role models and highlight the fact that the program is offered solely to women. Our aim is to measure any differences in application rates between the two groups of applicants. These will indicate to what extent the different hypothesized barriers are at work in this setting. In partnership with the training provider, we also will be able to identify differences in the characteristics of applicants (in terms of cognitive and non-cognitive abilities), to shed light on how the barriers operate for different individuals. We are in the process of evaluating the results of the first experiment in Arequipa, while we are currently launching larger experiments in Lima and Mexico.
External Link(s)
Citation
Del Carpio, Lucia and Maria Guadalupe. 2016. "Addressing gender biases and social identity in the technology sector in Peru." AEA RCT Registry. May 03. https://doi.org/10.1257/rct.1176-1.0.
We have partnered with a non-profit organization (NGO) that empowers women youth from low-income backgrounds in Peru, Mexico and Chile with education and employment in the tech sector. Potential candidates apply to participate in the program. A selection process assesses each participant, first through several written exams that evaluate interpersonal characteristics as well as logic and math skills, and then by a personal interview that evaluates traits like perseverance and commitment. Selected candidates then go through an immersive five-month coding bootcamp and are at the end connected with tech companies in need for talented coders.
In order to understand what are the barriers to pursuing a career in the tech sector, our intervention randomly varies the recruitment message to potential interested applicants. In addition to a control group message with generic information, in a treatment message, we correct misperceptions about women’s abilities to pursue a career in technology, provide role models and highlight the fact that the program is solely for women.
Intervention Start Date
2016-02-03
Intervention End Date
2016-05-31
Primary Outcomes (end points)
We measure differences in application rates and show up ratios to the examination sessions. We also identify differences in the characteristics of applicants (in terms of cognitive and non-cognitive abilities) as well as gender biases through various tests performed in the examination sessions.
Primary Outcomes (explanation)
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
We randomize the information provided in the NGO’s program application webpage which in both cases ends in a registration form. Traffic to the registration webpage comes from different sources. The main one is social media where the NGO advertises its program. In addition, the program also gets publicity in radio, TV, newspapers, and other local media. In all cases, interested applicants are directed to the application/registration webpage.
Experimental Design Details
Randomization Method
Randomization done with the VWO software.
Randomization Unit
Individual (unique clicks)
Was the treatment clustered?
No
Sample size: planned number of clusters
0
Sample size: planned number of observations
6000
Sample size (or number of clusters) by treatment arms
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Assuming a total traffic of N=6000 (unique) clicks in the recruitment webpage, and a baseline (control) application rate of 6%, we are able to detect a MDE of 0.08 standard deviation.
Assuming a baseline (control) show-up ratio to the examination sessions of 40% (of those registered), we are able to detect MDE of approximately 0.25 standard deviation.
For the various test scores, we are able to detect MDE of between 0.30 and 0.35 standard deviation.