How
An online field experiment in real comment sections: we post one visible, gender-equal comment on randomly chosen videos and measure whether others then speak up. (Detail next.)
Attitudes on gender are shifting faster in private than in public. When support stays silent, outdated norms survive even after most people have abandoned them — slowing progress on women's work.
An online field experiment in real comment sections: we post one visible, gender-equal comment on randomly chosen videos and measure whether others then speak up. (Detail next.)
35 Spanish-language channels across 8 countries — Colombia, Mexico, Argentina, Chile, Peru, Spain and more.
Díaz-Escobar & Boltz. Data pipeline already live; one research assistant runs the experiment.
Within two hours of a new video, we post one norm-challenging comment, assigned at random. The four treatments cross how they challenge the norm with which norm they target.
"Women working outside the home benefits the whole family."
"Men who share housework and childcare are better fathers and partners."
"In Colombian surveys, the majority — over 59% — support women working outside the home."
"Surveys show more men than it seems want to share housework and childcare."
Tiered deployment: the expression arms (T1/T3) are withheld from male-centered channels to avoid backlash contaminating the outcome; social-proof arms and polls run everywhere.
Flexible work is reshaping careers, but women take it and pay in wages and promotion. If the driver is a norm of maternal availability rather than preference, flexibility policy needs a different fix — and that distinction is what we identify.
Couples face one scarce flexibility package they assign to either partner. We randomize the framing and the scarcity — first with hypothetical offers, then with real packages. (Two interventions next.)
Couples across Colombia, recruited and run online — beyond Bogotá, at scale. Optional firm arm later.
Díaz-Escobar & Boltz, with IPA on implementation and firms for the workplace arm.
"Either of you can use it."
"…so a parent can be there for the children."
"…so the recipient does better at work."
"…so the household protects its earnings."
Each partner nominates privately, then the couple decides together. Does the frame push the package toward the wife? Cross-cut by scarcity (one package vs. one each); a subset wins its choice for real.
The benefit stays with the employee.
The worker can pass it to the spouse.
Pre-assigned to one partner unless changed.
Worker keeps it, after a joint info session.
Take-up, who actually uses it, job retention, hours, earnings, childcare time and spousal bargaining — does making flexibility transferable change who bears the cost?
On freelance platforms, women earn less, rank ~12 places lower, and hold far fewer skill certifications than comparable men (0.4% vs 5.2%) — despite slightly higher ratings. Controlling for writing style does not absorb the gap, so it isn't just women presenting themselves modestly.
A three-arm RCT giving freelancers access to a generative-AI writing assistant. If AI closes the gap, the constraint is supply-side self-presentation; if it persists, it is demand-side discrimination on signals AI can't change — photo, name, country. (Detail next.)
~1,200 active Workana freelancers in Argentina, Brazil, Colombia, Mexico — on a live panel of 7,712 profiles across 18 countries.
Díaz-Escobar (PI) & Boltz, in partnership with OpenAI (model access + privacy-safe usage data). Panel already built.
Freelancers are randomly assigned (stratified by gender × country) to one of three arms. The design separates access to AI from structured guidance on how to use it.
Pre-loaded prompts for rewriting bios, drafting proposals and portfolios, and translation, with a short onboarding.
Access without onboarding or curated prompts — isolates the effect of access from structured help.
The baseline of what freelancers achieve on their own; receives access after the study.
Fund one RA for ~6 months to run the seeded-comment experiment on data already built.
Fund the online couples pilot (conjoint + real-package lottery), then scale to real packages nationwide.
AI access in-kind from OpenAI; IPA could support Latin American recruitment (~US$10K ads).