Information Frictions, Screening, and Matching in Apprenticeship Markets

800 young jobseekers residing in Uganda’s West Nile region submit Expressions of Interest (EOI, containing basic demographic information), to be selected for ACAV’s 10-month vocational training program. This program consists of formal training, and an apprenticeship. In the EOI, jobseekers indicate their preferred areas of employment (e.g., tailoring, mechanic).

Based on the demographic information collected in the EOIs, we will construct 4,000 randomized candidate CVs that reflect the population demographics of the 800 young jobseekers. The hiring managers of the 120 firms (with 216 vacancies) participating in this study will then evaluate a random subset of 32 of these hypothetical CVs, rank their interest in the profile on a scale of 1-10, and shortlist 8 CVs. As such, firms undergo an incentivized résumé-rating (IRR; Kessler et al., 2019). The firm-side IRR will provide insights about which traits firms value, and will be used to calculate a predictive hiring probability for each worker–firm pair.

Subsequently, the 800 young jobseekers will complete an IRR, by evaluating 18 hypothetical firm profiles across their preferred areas of work, drawn from firm profiles. The hypothetical firm profiles will be based on data of the 120 participating firms. For each profile that workers see, they must indicate their interest in working for this firm (with a binding cap that limits how many firms workers can signal their interest to).

Among the young jobseekers who complete the IRR, there are two treatment arms. In the Control arm, jobseekers merely see the hypothetical firm profiles. In the Treatment arm, the jobseeker sees the hypothetical firm profiles along with their personalized predictive hiring probability per firm profile, based on the calculations done following the firm-side IRR.

After the firm- and worker-side IRRs have been completed, the preferences of both firms and workers will be used to select the 216 jobseekers for the available vacancies.

Workers and firms will be monitored during the vocational training program, apprenticeship, and for 6 months afterwards.

2026-06-01

2027-04-30

Firm-side IRR:
For each hypothetical worker profile:
- Interest in hiring worker for apprenticeship program (intensive margin, 0-10)
- Interest in hiring worker for apprenticeship program (yes/no; extensive margin)
- Justification of the rating of each profiles

Worker-side IRR:
- Preferences of workers about workplace characteristics;
- Application/signaling choices across firm profiles;
- Profile rankings and stated workplace preferences;
- Subjective probability of being shortlisted by each firm profile.

The primary outcomes are intended to capture both efficiency and allocation: how the information treatment changes application targeting, realized matches, and the quality of those matches, while the firm-side IRR measures the screening object used to construct the signal shown to treated workers.

Firm-side IRR:
- Firm-level characteristics (number of employees, annual revenue, value of capital, working hours, workplace amenities)
- Firm-level hiring constraints (difficulty of hiring workers with desirable characteristics)
- Social desirability and experimenter demand effect related questions.

Worker-side IRR:
Perceived job value, belief accuracy (the gap between subjective and model-implied screening feasibility), perceived discrimination/safety, perceived job burden, and task difficulty / effort.
Interpretation checks on what workers believe the signal represents, plus worker-side experimenter-demand-effect questions.

During Apprenticeship Period:
Attendance, completion, separation, retention by the firm, earnings, and job satisfaction measured during the program and in the 6-month follow-up.

Secondary outcomes are mainly mechanisms and longer-run performance measures used to understand whether information changes beliefs, self-selection, and post-placement retention.

The study uses a two-sided IRR design. First, firms participate in a within-firm 2x2 randomized CV-rating exercise that varies refugee status and ACAV quality certification across hypothetical worker profiles. Second, workers are randomized to Control or Treatment in the worker-side IRR, where the treatment reveals a personalized screening-feasibility signal. The study then follows the resulting allocations into real vocational placements and post-program outcomes.

Not available

Randomization is done by computer. In the firm-side IRR, refugee status and ACAV quality certification are randomized within firms in a balanced 2x2 design across the 32 evaluated CVs. In the worker-side IRR, jobseekers are randomized to Control or Treatment, stratified by gender, refugee status, and preferred vocational sector/area group.

Profile-level within firms (firm-side IRR) and worker-level (worker-side IRR).

No

Firm-side IRR analyses cluster at the firm level and, in preferred specifications, also by base CV profile. Worker-side IRR analyses cluster at the worker/respondent level (Abadie et al., 2023).

120 firms, 800 workers.

400 workers in the Control group, 400 workers in the Treatment group.

The attached technical appendix provides power calculations for the firm-side 2x2 IRR with m = 32 CVs per firm. This is the experiment for which we have the least power. With 100 firms, the main-effect MDE is about 10.0 percentage points under ICC = 0.10 (interaction MDE approximately 20 p.p.); if all approximately 200 planned firms complete the IRR, the same benchmark falls to about 7.1 p.p. under the same assumption.