Executive takeaway
A training place, a completed course and a sustained increase in earnings are different achievements. Each can matter, but public commissioners need to know which result a programme is intended to deliver and how it will be assessed. IDER’s editorial recommendation is to agree that result before procurement or expansion, then fund the information needed to judge it. Ghana’s latest verified labour observations in this article end in Q3 2025. An older Ghanaian experiment provides a useful design lesson, with clear limits on applying its findings to present programmes.
What the evidence shows
GSS’s current public panel reports national unemployment of 13.0% for Q3 2025. Its catalogue for the 2025 Q1–Q3 Labour Force Survey, updated on 18 August 2026, describes employment differences across sex, geography and economic activity. The metadata update is not a new observation period, and a higher employment count for one sex does not establish equality in rates, earnings or job quality. [S1] [S2]
A May 2019 World Bank working paper evaluated an earlier Ghana apprenticeship programme using random assignment. Placement occurred in October 2013–January 2014, with follow-up in August 2017–May 2018. It found limited evidence of improved average labour-market outcomes at that horizon, with differences across trades and some participants still training. This concerns a historical programme cohort, not the effectiveness of any 2026 initiative. [S3]
The World Bank’s impact-evaluation guidance centres causal analysis on a credible counterfactual: an estimate of what participants’ outcomes would have been without the programme. It emphasises choosing an evaluation design compatible with programme rules and planning it before implementation where possible. Simple before-and-after comparisons cannot by themselves establish an attributable effect. [S4]
Why this matters
Begin the commissioning brief with the decision that evidence must support. Is the question whether to expand an apprenticeship model, adjust placement support or change the mix of services offered? Specify the eligible group, expected pathway to work and realistic follow-up horizon. An evaluation commissioned without a decision in view can produce a technically competent report that arrives too late or answers the wrong question.
Define a small set of outcomes in operational terms. Sustained work could require observation at more than one follow-up point. Earnings should cover relevant wage and self-employment income, with a consistent period and treatment of business expenses. Collect working hours alongside income so a change can be interpreted. Include programme and participant costs when assessing whether an approach is affordable and worthwhile.
Protect the credibility of selection and reporting. Publish eligibility criteria and document changes to them. Keep a traceable record of offers, take-up, completion and follow-up without making identifiable participant data public. If a provider is rewarded for employment outcomes, examine whether its incentives could discourage recruitment of people facing greater barriers. A contract should make inclusion and evidence quality visible alongside performance.
Who is affected differently
Gender and youth inclusion require questions about the route through a programme, not just the composition of its intake. Ask whether schedules, travel, caring responsibilities, workplace conditions or entry requirements affect participation and completion. Investigate those barriers with participants and non-participants, then test practical adaptations rather than assuming one standard package serves everyone.
Age bands, trades, prior education and local labour demand should be specified before subgroup results are interpreted. A programme aimed at school leavers answers a different question from one supporting experienced workers. Small subgroup samples can produce unstable estimates, so explain uncertainty and resist selecting only favourable results. Differences between trades may also reflect the time required to move from training into paid work.
Recommended actions
What to watch next
Sources, limitations and next step
Evidence boundaries
This newsletter does not estimate the effect or cost-effectiveness of a current employment programme. The 2019 working paper provides historical experimental evidence, while GSS supplies national descriptive context. Neither can establish the performance of a newly commissioned intervention. The former’s follow-up timing and trade composition limit generalisation; the latter cannot identify a programme’s causal contribution.
An evaluation can also miss broader effects on other workers or firms if it observes only participants. Commissioners should discuss that risk when defining the study population and expansion decision. Where a credible impact design is infeasible, a well-documented implementation and outcome review can still be useful, provided its findings are not labelled causal effects.