Apples and oranges: making sense of employment outcomes

Why access, job creation and job quality interventions shouldn’t be compared by headline numbers alone.

This first blog draws on findings from the MSD4E landscape assessment and explores why access, job creation and job quality interventions shouldn’t be compared by headline numbers alone. Recognising these differences helps us interpret employment outcomes more clearly and choose the mix of interventions that best fits the context. It will be followed by a second blog exploring incentives for more and better employment, and a third on why MSD4E needs more ex-post evaluations.

Bigger (numbers) aren’t always better

While sharing finding from the landscape assessment, one question kept coming up: “Which type of intervention is best?”

This is understandable but hard to answer because it’s incomplete. We looked at over 50 initiatives aiming to improve employment outcomes - more jobs, better jobs, or better access to jobs - and their results looked wildly different. One connected over a million people to work, while another created a few hundred jobs. Both were well-designed, yet their outcomes were on different scales. These numbers describe fundamentally different things, so to answer “which is better,” we also need to ask “better at what?”

Distinguishing between employment outcomes matters. Without that nuance, headline numbers dominate priorities and pull resources toward larger, more visible results, regardless of what the context needs most.

Understanding different employment outcomes

All MSD4E interventions aim to change the systems that shape employment outcomes, but not every intervention is changing the same thing or operating in the same part of the system.

Some interventions focus on access to work, usually through skills development or job matching that help jobseekers find existing unfilled vacancies. Taking job matching as an example, results can emerge quickly and at scale once platforms (and the systems that support them) are up and running, because each additional match costs little to replicate, with network effects drawing in more employers and jobseekers as the platform grows. EYE Kosovo, for example, spent over ten years building the foundations of a private job-matching market that now connects hundreds of thousands of young people to work.

So under the right conditions (i.e. existing unfilled vacancies and a workforce that's ready to fill them but unable to get matched), job matching can impact significant numbers of people. But this doesn't make sense in every context - so we can't just 'trust the numbers' if we're working in a context with low demand for labour and limited vacancies.

Other interventions focus on job creation, addressing the factors that shape demand for labour. Because this usually relies on firm growth - which is influenced by ‘upstream’ factors like access to finance, the business environment, and market linkages - results can take some time to appear. Ultimately, committing funding for a new position depends on a firm’s confidence in future demand for its products or services and its ability to grow profitably. The marginal cost of adding a worker is therefore much higher than in the job matching example, because firms must commit real resources such as wages, equipment and inputs to expand production.

To put this in context, MSD4E interventions targeting formal wage work typically created a few hundred to around a thousand jobs. This is objectively impressive, especially given that many were also building the foundations for future growth and job creation. However, it sits far below the scale achieved by job-matching or skills interventions.

Job quality interventions aim to improve how people experience work once they have it. These interventions change the rules, standards, and incentives that govern safety, stability, and fairness. Their results may not appear in 'number of jobs' figures but are valuable in the many contexts where most work is low-paid or insecure. Because they often involve institutional or behavioural change, effects can take time to embed but can influence whole sectors. For example, in Rwanda the ILO’s support to the construction sector led to safety standards being written into the national building code - a reform with the potential to protect thousands of workers each year.

None of these outcomes are inherently better than the others, but they differ in how they’re achieved, the impact they have in context, how quickly results emerge, and how visible those results are. Yet we often still treat them as if they were comparable, placing access, job creation and job quality outcomes side by side, or overlooking differences within the same category (for example by considering the creation of lower-value jobs as equivalent to higher-value ones).

What happens if we don’t compare like-for-like?

Funding and accountability pressures can pull attention towards what is easiest to see and count, side-lining slower, more complex changes. This is especially relevant in employment programming because of the significant differences between the three outcomes outlined above.

Once expectations for ‘scale’ are set according to the biggest results reported elsewhere, they tend to ripple through programme design and delivery, shaping what gets funded, how success is measured and which employment outcomes are valued most. This can push programme towards a ‘path of least resistance’ - focusing on activities that generate quicker, lower risk and more countable results, even when the real constraints lie in slower, harder, but more transformative areas.

What we need to be mindful of

As MSD4E continues to grow and evolve, there are practical steps that can help steer attention and resources towards the areas that matter most.

Firstly, we can start with diagnostics that identify the real binding constraints to more and better work. This means resisting the temptation to begin with a preferred solution or familiar model. If the core problem is a shortage of good-quality jobs, then focus should be on helping firms expand or improve the quality of existing work, not creating more low-quality jobs or matching people with low-value opportunities they might have accessed anyway.

We also need to manage expectations around the pace and scale of change. Meaningful shifts in employment systems can take a decade or more to materialise. Rigid short-term job targets can be counterproductive, especially in nascent markets. Patience, flexibility, co-design, and recognition of intermediate results, such as investment, innovation, and market access improvements, are important.

Comparisons between initiatives need the same nuance. A programme creating formal wage jobs in an emerging sector can’t be directly compared with an entrepreneurship programme or a job-matching platform. Each is addressing a different part of the system and will show impact on a different scale and timeline.

As we step back from individual interventions, another question emerges. Many employment programmes combine or sequence approaches across access, job creation, and job quality. Understanding how these elements interact may be just as important as distinguishing between them. A more complete picture will come from studying whole programmes, not just single interventions, to see how different levers reinforce or undermine one another over time.

Reconciling the logic of each outcome while examining how they work together remains one of the biggest opportunities for learning in this field.

See also our:

• second blog in this series: Cracking the incentives puzzle in MSD4E
• third blog in this series: Why MSD4E needs more ex-post evaluations