News and Insights
Understanding poverty in rural Ethiopia: new evidence on programmes, measurement, and women’s time
Ethiopia has experienced sustained periods of macroeconomic growth over the past two decades. Yet for the more than 80% of its population that depends on smallholder agriculture, that growth has often felt distant. Recurrent droughts reduce harvests. Land degradation limits what the soil can give. Extension services do not reach everyone who needs them. And when shocks arrive, as they reliably do, households have little to fall back on. Chronic food insecurity is not an exception in rural Ethiopia; it is a reality for many people.
Understanding why poverty persists in this context and what interventions genuinely help requires evidence that is rigorous, longitudinal, and grounded in the realities of specific places and people. Three new working papers by Fisker, Mekasha, and Tarp, drawn from the same panel of 1,787 rural Ethiopian households tracked between 2021 and 2025, begin to provide that evidence. Each paper asks a different question and shed light on the scale of investment going into rural development in Ethiopia, the tools available to monitor its effects, and the constraints that continue to prevent households, particularly women, from building lasting resilience.
Do the flagship programmes work?
Ethiopia’s three major rural development programmes — the Agricultural Growth Programme (AGP), the Sustainable Land Management Programme (SLMP), and the Productive Safety Net Programme (PSNP) — represent enormous public and donor investment. They have reached millions of households, reshaped landscapes, redistributed food, and extended credit and technical advice to farmers who previously had none. But what has been harder to establish, until now, is what they have actually delivered, and for whom.
The first paper tries to answer this question. The findings point to some clear conclusions, but also important cautions. SLMP, which focuses on land restoration and ecological recovery, generated measurable resilience gains at relatively low marginal cost, making it well suited to climate adaptation objectives. AGP, which promotes agricultural technology adoption and productivity, appears associated with yield improvements though isolating its specific contribution is genuinely difficult where multiple programmes operate simultaneously in the same areas and communities. PSNP has been effective at smoothing consumption during lean seasons, but its pathways to structural transformation remain limited without stronger integration with nutrition, livelihoods support, or asset-building components.
Perhaps the most important finding is about coordination. These programmes often run concurrently, in overlapping geographies, with different implementing agencies and timelines. When they are not sequenced deliberately, for instance, pairing land restoration with subsequent extension support — it becomes hard to know which investment is driving change, or whether the potential complementarities between programmes are being realised. This leads to missed opportunities for design and evaluation in Ethiopia.
Can satellites tell us whether things are getting better?
A growing body of research has demonstrated that satellite imagery combined with machine learning can predict poverty with considerable accuracy. Freely available geospatial data such as vegetation indices, night time lights, building footprints, distance to roads and cities can identify which areas are poor well enough to support geographic targeting decisions. This has generated real enthusiasm, and genuine practical application, in contexts where household survey data is expensive to collect or takes years to produce.
The second paper asks an important question at this point: whether satellites can track whether poverty is changing. Using GPS coordinates for each household across three survey rounds, the researchers extracted time-varying spatial features and tested them against actual consumption data. The results are instructive. Accuracy for a single year is moderate and reasonably stable, the models can identify which communities are poorer, but when the task shifts to predicting whether households are better or worse off from one round to the next, performance drops substantially. Within villages, the satellite-based models provide essentially no ability to distinguish the households experiencing most poverty from its slightly better-off neighbour.
This does not make satellite-based poverty mapping less useful but they should not be expected to replace household-level data for identifying which individuals within a community need support, or for monitoring welfare change at the scale that social protection programmes require.
What the clock takes away
Climate adaptation has become one of the dominant frameworks in rural development programming. Governments, donors, and NGOs invest heavily in extension services, improved seed varieties, irrigation training, and climate information systems. The assumption embedded in much of this investment is that households, given the knowledge, the technology, and the opportunity, will adapt. That assumption may not be accurate.
The third paper looks at what happens when women face constraints on their time that are severe enough to prevent engagement with adaptation programmes, not from lack of interest, but from lack of hours. In rural Ethiopia, as across much of Africa, domestic responsibilities fall disproportionately on women. Cooking, water collection, childcare, and firewood gathering are not tasks that can simply be deferred. They happen daily, and they take time that is not recoverable.
The research finds that higher domestic time burdens significantly reduce women’s likelihood of contacting extension services, attending training sessions, or accessing climate information. The mechanism is not complicated: these activities require physical presence at a scheduled place and time. Women who are already working beyond a socially determined threshold of total hours cannot add more, however well-designed the programme on offer.
The study also draws an important distinction that is often missed in policy discussions. The constraint operates through both the total level of domestic work a woman carries, and through the distribution of work within the household. Reducing one woman’s burden in isolation does not address the underlying inequality that redistributes domestic tasks unequally in the first place. Technology-focused and institution-focused interventions which describe the majority of climate adaptation programming will remain only partially effective while this goes unaddressed.
What the three papers, read together, suggest
Ethiopia has invested heavily in rural development. The evidence is beginning to arrive on what that investment has delivered and where the gaps remain. Programmes that restore land and build ecological assets appear to generate resilience gains that are measurable and cost-effective. Large productivity programmes work but disentangling their effects from the broader environment in which they operate is difficult, and the evaluations need to be designed with that difficulty in mind. Satellite tools are genuinely valuable for geographic targeting and can maintain that value across time but they have clear limits, and those limits matter for how monitoring systems are designed and what claims are made about them. And the structural constraints on women’s time overlooked for too long in discussions about why adaptation programmes underperform are not peripheral to the question of whether rural poverty declines. They are central to it.
