Active Labour Market Policies in Addis Ababa: 12-Year Follow-Up

The study follows up on two active labour market interventions implemented among young jobseekers in Addis Ababa in 2014-2015. The first reimbursed transport costs for job search. The second provided skills tests, certificates reporting test performance, and guidance on using the certificates in job applications.

2014-09-01

2015-02-28

Endline welfare
EW1. Current monthly labour earnings (one constructed outcome).
EW2. Current formal wage work (one binary outcome).
EW3. Current job quality (one index).

Productivity and human capital
PH1. Human-capital accumulation over the career (one index).
PH2. Match Quality (one derived outcome).
PH3. Skill use, advancement and responsibility (one index).

Cumulative career welfare
CW1. Cumulative labour earnings (one person-level aggregate).
CW2. Share of the observed career spent in work (one person-level proportion).
CW3. Career security and formality (one index).

EW1 is monthly wage earnings plus the monthly equivalent of bonuses, overtime pay and other employer payments received during the previous 12 months, plus monthly self-employment profits. Respondents without current work receive zero. EW2 equals one when the respondent currently holds a wage job with either a written fixed-duration agreement or a written agreement of undetermined duration, and zero otherwise.

EW3 is an Optimus index with 11 fixed components: (1) permanent wage work; (2) a written agreement of undetermined duration; (3) expected tenure of at least five years or until retirement; (4) employer pension contributions; (5) employer health insurance; (6) severance protection; (7) workers' compensation or injury insurance; (8) paid sick leave; (9) paid annual leave; (10) paid maternity or paternity leave; and (11) reliable pay. These components equal zero for respondents without current wage work.

PH1 is an Optimus index with 10 fixed components: (1) any formal training during the previous ten years; (2) number of training programmes; (3) total months in formal training; (4) any useful skill reported from formal training; (5) any useful skill reported from formal education; (6) any useful skill reported from work experience; and, for one past wage job selected at random by the survey software during the interview, (7) employer-provided training; (8) weeks of training; (9) a portable certificate; and (10) mentoring or informal training from senior colleagues.

PH2 is one measure of the congruence between workers' skills and the jobs or occupations they enter. The exact measure will be specified after treatment-blind assessment of which occupational categories and worker characteristics can be measured reliably in the data collected before registration.

PH3 is an Optimus index with 12 fixed components: (1) skill or training level required by the current wage job; (2) reading and writing at work; (3) arithmetic at work; (4) English use at work; (5) share of work time using a computer; (6) promotion by the current employer; (7) responsibility for at least one other worker; (8) share of current wage-job tasks spent on supervision or coordination; (9) mean within-spell change in real log wages across wage-job spells; and, for one past wage job selected at random by the survey software during the interview, (10) an automatic wage increase with experience; (11) a performance-related wage increase; and (12) applying for promotion.

CW1 is the sum of monthly wage earnings and self-employment profits from the first-endline interview through the 2026 interview. CW2 is the number of observed months in wage work, self-employment or farming divided by the number of observed months. CW3 is an Optimus index with three fixed components: (1) the negative of the share of observed months in involuntary non-work; (2) the negative of the number of adverse wage-job or business exits; and (3) the share of observed months spent in wage work with a written agreement.

All index components will be oriented so that higher values represent better outcomes. We will construct the Optimus indices following Anderson and Magruder (2022, Highly Powered Analysis Plans, NBER Working Paper 29843). Their procedure uses sample splitting to estimate weights that combine the fixed component outcomes and then applies those weights to held-out observations. Component definitions, sample means and estimated index weights will be documented. Conventional inverse-covariance and equal-weight indices will also be reported.

Event studies
Treatment-effect paths for work, labour earnings, formal wage work, open-ended wage work, involuntary non-work, training, promotion and responsibility.

Alternative definitions and decompositions
Wage-only current and cumulative earnings; current earnings excluding additional employer payments; self-employment earnings; hourly wages; and the individual components of the four Optimus indices.

Open-ended questions
Pre-specified questions about respondents' own decisions and external constraints, their experience of substantial periods without work, and whether their careers have moved them forward or held them back.

Event studies will use the finest credible interval supported by the recall data. We will estimate treatment-effect paths for: (1) any work, including wage work and self-employment, measured monthly where dates permit; (2) labour earnings, combining wage earnings and self-employment profits and normally measured annually; (3) wage work with a written agreement; (4) wage work with a written agreement of undetermined duration; (5) involuntary non-work, identified from non-work spells during which the respondent was actively searching; (6) dated education or training spells; and (7) promotion and responsibility where their timing can be measured. Current outcomes will use the 2026 follow-up. Cumulative outcomes will combine the life history with earlier survey waves. Where two sources cover the same period, the main measure will normally use the report collected closest to that period, while the overlapping report will be used to assess consistency.

Alternative earnings definitions and the individual component outcomes will show which measured outcomes account for any effect on an aggregate outcome. These do not constitute additional primary outcomes.

The open-ended analysis will use three pre-specified questions: whether respondents attribute where they are today mainly to their own decisions and effort, circumstances outside their control, or both; whether substantial periods without work felt like being stuck, waiting, or trying to move forward; and whether their work life so far has mostly moved them forward, held them back, or done both.

The original experiment used two levels of randomization. Geographic clusters were assigned to transport-subsidy, skills-screening, or control arms. Within treatment clusters, eligible jobseekers were randomly selected to receive the assigned intervention. The 2026 long-run follow-up covers the transport group, the screening group, the pure control group, and untreated jobseekers living in transport or screening clusters. The latter groups allow us to estimate spillover effects. The study builds on AEARCTR-0000911.

Not available

Randomization was conducted by computer in the research office. Cluster assignment and individual assignment within treatment clusters were randomized. Individual assignment was blocked by education and rerandomized to improve balance.

Geographic cluster and individual jobseeker. Clusters were assigned to the transport-subsidy, skills-screening, or control arms; individuals within treatment clusters were then randomly selected to receive the assigned intervention.

Yes

178 geographic clusters in the 2026 long-run follow-up

3,052 jobseekers in the intended 2026 long-run follow-up sample

Group 1, transport subsidy: 774. Group 2, skills screening and certification: 768. Group 5, pure control: 823. Group 6, untreated in transport clusters: 500. Group 7, untreated in screening clusters: 187.