Understanding Barriers to Youth Employment in Kenya

How to cite

Amit, Inbar; Barker, Nathan; Andrew, Alison; Orkin, Kate; Garlick, Robert; Nekesa, Carol. 2026. "Understanding Barriers to Youth Employment in Kenya." AEA RCT Registry. AEARCTR-0018220. https://doi.org/10.1257/rct.18220-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Inbar Amit* , Nathan Barker , Alison Andrew , Kate Orkin , Robert Garlick , Carol Nekesa

* Primary investigator

Registry ID
AEARCTR-0018220
DOI
https://doi.org/10.1257/rct.18220-1.0
First registered
Last updated
Status
On going
Trial dates
2026-02-16 – 2027-12-31
Countries
Kenya (Nairobi)
Registry record
http://www.socialscienceregistry.org/trials/18220

Abstract

This project will study high youth unemployment in large, growing cities in developing economies. We will study the constraints that prevent young adults in Nairobi’s informal settlements from accessing job opportunities in commercial areas, and evaluate whether an intervention that relaxes multiple constraints at once can change jobseekers’ job search behavior and improve their labor market outcomes.

Intervention

Our theory of change is that young jobseekers face multiple, mutually reinforcing constraints, each of which can independently prevent effective job search. In this setting, relaxing any single constraint in isolation is unlikely to substantially improve labor market outcomes. Instead, meaningful improvements require simultaneously addressing these constraints.

First, many jobseekers hold inaccurate or overly pessimistic beliefs about hiring practices. In particular, they often believe that personal connections or formal credentials are necessary to be considered for jobs, especially in commercial areas. As a result, many restrict their search to their home neighborhoods, sharply limiting their opportunity. Second, jobseekers face severe liquidity constraints. Living largely hand-to-mouth, they are unable to absorb even small short-term costs associated with search. The costs of transport to commercial areas, combined with the opportunity cost of forgoing casual work, limit both the intensity and duration of search (Franklin, 2018; Abebe et al., 2021; Banerjee and Sequeira, 2023). Third, many jobseekers lack the soft skills required for effective search. Low self-efficacy, limited resilience to rejection, and weak communication skills reduce search effort and persistence, consistent with evidence on the role of planning, counseling, and behavioral support in job search (Abel et al., 2019; Witte et al., 2025). Consistent with this, field observations indicate that many jobseekers struggle to approach employers confidently and often discontinue search after only a few rejections.

The intervention is designed to jointly relax these constraints through a multi-day “guided job search” program with three components. First, a half-day workshop provides accurate information about hiring practices in Nairobi’s commercial areas—emphasizing that firms do consider walk-in applicants without formal credentials, but that success typically requires mul- tiple applications—alongside training in resilience and professional communication, delivered through role-play. This information will be based on information collected during a survey of firms (Barker et al., 2024). Second, participants engage in 3 days of guided job search in commercial areas, working in groups of 3 accompanied by trained guides who encourage them to engage with potential employers. Third, the program provides transport subsidies and daily stipends (approximately $4) to relax liquidity constraints during the search period. The intervention is delivered by non-specialist guides with undergraduate degrees and two weeks of training, with an emphasis on cost-effectiveness and scalability.

Individuals will be randomized into receiving the treatment based on their plot of residence. Following the baseline survey, all individuals residing in the same plot will be jointly assigned to either treatment or control status. We estimate that we will have approximately 850 clusters. Individuals assigned to treatment will be scheduled to receive the intervention on a rolling basis, determined by the timing of their baseline survey. In a second stage, jobseekers will be randomly assigned to groups of three to take part in the guided job search activity together. Randomization will be done using a computer.

The commercial areas in which the guided job search activity will take place will be identified based on the firm survey conducted in Barker et al. (2024). These areas will be further partitioned into 12 zones with roughly equal numbers of large firms (defined as those with more than five employees). Each group of jobseekers will then be allocated to a zone within the commercial areas that has a relatively high concentration of businesses in the sectors they report interest in, based on the distribution of firms across zones.

Experimental design

We will estimate average treatment effects (ATEs) by comparing mean outcomes between treatment and control groups, including randomization stratum fixed effects, and when collected, the baseline value of the outcome in question. In cases where the baseline value is missing, we will code the baseline value as equal to 0, with an indicator variable for whether the value was missing.

We will estimate ATEs separately for the midline, high-frequency panel, and endline, to examine treatment effect dynamics over time. For outcomes that are measured with greater noise, we will also pool data across waves to improve statistical precision, following McKenzie (2012). When pooling across survey waves we will add time-period fixed effects. We will assess treatment effect heterogeneity by gender for all outcomes. Standard errors will be clustered at the plot level, which is the unit of randomization. In cases where we have multiple measures that all capture very similar concepts, we will adjust for multiple hypothesis testing over the measures using sharpened q-values that control the false discovery rate following Benjamini et al. (2006).

Design details

Randomization method
Randomization will take place remotely using a computer.
Randomization unit
We will randomize at the "plot" level. This corresponds to all households that live behind the same gate.
Clusters
We anticipate having about 450 plots.
Observations
We anticipate having 1,600 job seekers.
Arms
We anticipate allocating 640 job seekers to the treatment arm and 960 to the control arm.
Primary outcomes
Our main hypothesis is that our intervention will shift jobseekers’ beliefs about their job search, change where they search for work, increase job search effort, and increase their probability of being employed. To study this hypothesis, we will investigate impacts on the following primary outcomes: 1. Job search location (own neighborhood vs. commercial areas) 2. Job search effort and methods 3. Employment status 4. Quality of employment, based on factors such as job location, earnings, hours and formality
Secondary outcomes
In order to better understand the mechanisms through which the intervention operates, we will investigate its effects on the following secondary outcomes: 1. Beliefs about future job search activities and outcomes 2. Beliefs about the returns to search in different areas 3. Preferences over job characteristics 4. Soft skills (perseverance, assertive communication, self-efficacy, resilience, behavioral activation, initiative) Finally, we will also examine the impacts of the intervention on a number of downstream outcomes, including jobseekers’: 1. Food security 2. Consumption 3. Cash-on-hand 4. Mental health

Cited by

No citation data available for this registration yet.

Keywords