Job Seekers' Beliefs and Labor Market Demand

We provide beneficiaries of two labor market programs access to a labor market information platform that uses artificial intelligence to help users build a personalized skills profile. The platform then recommends training programs and careers that best match their abilities and are in demand. All treated user receive access to the CV feature of the app, which helps job seekers identify their skills by guiding them through a series of questions based on their past jobs, education, and other experiences. Once users have finished entering in their experiences, they have the option of generating a CV which displays standard education and work experience information, but also the identified skills associated with each experience. Two treatment arms will receive access to course and career recommendations which use a Large Language Model to suggest careers and government-provided training programs which match the users' skills. One last treatment arm will receive recommendations based both on their skills and based on careers highlighted as being in-demand based on real-time job-postings data.

2025-05-01

2025-07-31

The surveys will measure five key outcomes: 1) skills knowledge and career aspirations,
2) job-search behavior, 3) reservation wages; 4) employment outcomes; and 5) mental health.

See Pre-Analysis Plan

We conduct an individual-level randomized controlled trial to evaluate how access to information 1) on skills, 2) course and career matching, and 3) labor market demand influences job seekers' ability to find employment. Specifically, Secretariat of Employment and Vocational Training (SEFP) program beneficiaries are assigned to one of three versions of the AI platform and a comparison group. The first version (SkillLab CV) enables users to input their labor market experiences, identifies their skills, and generates a tailored CV. The second version (SkillLab “Recommended Careers/Courses”) includes an additional feature that recommends careers/courses tailored to the user's skill profile. The third version (SkillLab “In Demand: Careers/Courses") introduces a new feature that recommends careers and courses tailored to the user's skill profile and that align with current labor market demand. Finally, users in the comparison group will not receive access to SkillLab in 2025. We observe course enrollment and employment outcomes via administrative data, and conduct follow-up surveys via phone and online platforms to measure outcomes on beliefs, reservation wages, mental health, and other outcomes.

Not available

Randomization will be conducted in Stata.

The randomization will occur at the individual level, stratified by municipality, baseline program participation, and above/below median age.

No

N/A

We will send invitations to 500,000 beneficiaries. We expect a take-up rate of around 10%, resulting in 50,000 participants for which we observe

125,000 participants will be assigned to control, 125,000 will be assigned to T1 (CV treatment), 125,000 will be assigned to T2 (CV + skill recommendations), 125,000 will be assigned to T3 (CV + skill and demand recommendations).

In the binary outcome calculation, we are powered to 80% to detect effects ranging from 2 to 4.5 percentage points, depending on resulting take-up rates. In the continuous case, we are powered to 80% to detect effect sizes ranging from 0.075 to 0.225 SD.