Psychological well-being during job search: Evidence from youth seeking employment in Ethiopia

Searching for a job is a dynamic and self-regulatory process in which effort and motivation are key, but it can be psychologically taxing. Unemployed individuals are more than twice as likely to experience symptoms of psychological distress as their employed peers, and the act of searching itself is associated with negative affect. Beyond the immediate welfare losses, psychological distress could directly undermine the job search process. An extensive literature in clinical psychology has linked distress with pertinent cognitive mechanisms -- including reduced motivation, overly pessimistic (self-)assessments, a tendency to overweight negative signals, and reduced cognitive capacity and endurance -- which may contribute to generating self-reinforcing cycles of unemployment and worsened well-being. This issue is particularly pronounced in low- and middle-income countries, where many young people experience extended periods of unemployment while searching for stable jobs with prospects for career growth. Youth and recent graduates are especially vulnerable; as they are typically overoptimistic about their labour market prospects but are then quickly faced with repeated rejection, which can give way to frustration and discouragement and ultimately depress labour market outcomes in the long-run.

In this project, we experimentally evaluate the impact of a psychological support intervention for young jobseekers in Addis Ababa, Ethiopia -- a context characterised by high youth unemployment, limited social safety nets, and high rates of psychological distress. Our intervention, delivered over five weekly in-person sessions, is based on principles of cognitive behavioural therapy and aims to improve young unemployed jobseekers’ psychological well-being and resilience during the job search process. The curriculum includes techniques to manage stress and anxiety, encourage reflection about job search strategies, develop strategies to break down and address practical problems, motivate engagement in meaningful activities and promote initiative-taking, and strengthen coping strategies for dealing with setbacks. We randomly assign jobseekers to either a group or an individual version of the program, both of which covers the exact same curriculum but where group discussions are replaced with individual reflection. This design allows us to estimate the overall effect of providing psychological support during job search, as well as any additional marginal impacts that may arise by leveraging group dynamics – including through exchanging information about the labour market, sharing personal experiences, and receiving encouragement from peers.

2025-12-08

2026-02-28

We define three families of primary outcomes: psychological well-being, job search behaviours, and labour market outcomes. All outcomes are measured in two follow-up surveys conducted shortly after the end of the intervention, and results are pooled across survey rounds.

1. Psychological well-being: Primary outcomes are: (i) a binary indicator for psychological distress (GHQ-12 ≥ 4); and (ii) a binary indicator for at least moderate depression or anxiety (PHQ-ADS ≥ 20). Secondary outcomes include standardized GHQ-12 and PHQ-ADS scores and their distributions.

2. Job search behaviours: We construct two inverse covariance weighted (ICW) indices. The first captures job search intensity using: time spent searching in the past seven days, job search expenditures in the past seven days, number of applications submitted in the past four weeks, and breadth of search channels used. The second captures the quality of jobs applied to using: reservation wage, average wage of the last ten jobs applied to, share of applications outside the preferred field or occupation, and an index of job quality characteristics (including field match, non-wage benefits, and opportunities for skill development or career growth). Impacts on index components are reported as secondary outcomes.

3. Labour market outcomes: We construct an ICW index of labour market participation using: economic attachment, formal full-time employment, hours worked in the past seven days, and total earnings. We also construct an ICW index of employment quality based on hourly wage, job formality, full-time status, job–field match, learning opportunities, and job satisfaction, measured with respect to the respondent’s main activity. Impacts on index components are reported as secondary outcomes.

Secondary outcomes capture potential mechanisms through which the intervention affects psychological well-being, job search behavior, and labor market outcomes. All secondary outcomes are collected in the same follow-up surveys as the primary outcomes but will be split across one of the two surveys (i.e. they are only measured once). The analysis of these outcomes is descriptive and exploratory, aimed at informing interpretation of the primary treatment effects.

1. Beliefs about the labor market and job search, including: (i) expectations about peers’ employment rates and earnings; (ii) perceived likelihood of finding any job and of finding a job in a preferred field or occupation over different time horizons; and (iii) expected wages in the next job and over the medium term. We also measure perceived returns to job search effort by eliciting subjective probabilities of finding a job at different levels of search intensity.

2. Preferences over job characteristics: Where participants are asked to rank job attributes, including earnings, job security, skill development, working conditions, work–life balance, and social impact.

3. Job search networks and information sharing: including the size and composition of personal and professional networks used for job search, the receipt of job referrals, and the types of information exchanged (e.g., job vacancies, labour market information, application or interview assistance, and emotional support).

4. Psychological costs of job search: Including a hypothetical willingness to pay to outsource a standardized application task, perceived avoidance or procrastination of job-search activities, available cognitive bandwidth for job search, and persistence following rejection or lack of response.

The study is an individually randomized controlled trial conducted among unemployed recent graduates actively searching for work in Addis Ababa, Ethiopia. Eligible participants are individuals who completed tertiary education within the last four years, are not currently employed full-time, and report having engaged in job search activities in the recent past. Participants are recruited through online and physical job boards and social media channels commonly used by recent graduates. Eligible participants are invited to complete an in-person baseline survey and those who complete the survey are recruited into the study.

Following the baseline survey, participants are randomly assigned at the individual level to one of three arms: (i) a group-based psychological support intervention, (ii) an individual (one-on-one) version of the same intervention, or (iii) a control group that does not receive any intervention. Due to implementation capacity constraints, the intervention is rolled out in two phases, with randomization conducted separately within each phase.

The primary estimand is the intent-to-treat effect of being offered psychological support during job search, estimated by comparing treated participants to the control group (i.e., pooling both intervention groups), with additional analyses comparing group-based and individual delivery formats.

Not available

Randomization done in office by a computer.

Randomization will be done at the individual level.

Assignment will be stratified by the following baseline characteristics: (i) gender (male or female); (ii) psychological distress, measured using the GHQ-12 score (range 0–12), divided into terciles within the male subsample and into two groups (above or below the sample median) within the female subsample due to the smaller female sample size; (iii) cognitive ability, measured using Raven’s Progressive Matrices, again divided into terciles for males and into two groups for females; and (iv) graduation year, defined as a binary indicator for having graduated in the class of 2025 (2017 in the Ethiopian calendar), which serves as a proxy for job search duration.

No

N/A

We aim to recruit a sample of approximately 1000 jobseekers. The final sample size will be determined by how many jobseekers complete the baseline survey between October 30, 2025 and January 26, 2026.

Approximately 1,000 jobseekers will be randomized into one of three treatment arms: (i) Group PM+, comprising approximately 340 individuals (34 percent); (ii) Individual PM+, comprising approximately 160 individuals (16 percent); and (iii) a control group, comprising approximately 500 individuals (50 percent). Randomization will take place within each phase using the same assignment probabilities.