Funding Shocks and ICT4D Evaluation: How Research Teams Are Measuring Impact With Less Money

The OECD’s preliminary figures for 2025 put official development assistance at $174.3 billion, a 23% decline from 2024 and the largest annual contraction on record. Spending on development programmes, projects and technical co-operation fell by 26.3%, the biggest drop in that component the OECD has recorded. Five donors, led by the United States, account for almost all of the fall.
For ICT4D researchers, the headline is less important than the plumbing. Impact evaluation of digital interventions depends on survey programmes, donor-funded evaluation budgets and partner organizations that were themselves financed by the programmes now being cut. The question for the field is what can still be measured, and how, when each of those inputs is smaller.
What was lost: the data infrastructure
The clearest documented example is the Demographic and Health Surveys (DHS) Program. It runs nationally representative household surveys in more than 90 low- and middle-income countries, with 452 surveys conducted since 1984 and more than 6,000 peer-reviewed papers drawing on its data, according to a review published in a peer-reviewed journal. When US funding was suspended in early 2025, staff could not approve new data users, and surveys in progress or awaiting implementation in 23 countries were disrupted. Data processing for completed surveys in Malawi, Zimbabwe, Zambia and Nigeria was among the casualties.
DHS is a health programme, not an ICT4D one, but its fate matters here. DHS also records some ICT indicators, such as mobile phone ownership and internet use, so researchers can use its microdata as a baseline or comparison source. When a shared dataset goes dark, every evaluation that depended on it loses a reference point.
A bridge arrived: ICF received a three-year emergency grant from the Gates Foundation in July 2025 to keep the DHS website running, maintain access to data and tools, and complete some surveys. The programme’s longer-term future remained unsettled. That pattern, where a philanthropic stopgap keeps core data alive without resolving who pays after it ends, is likely to recur.
What evaluation teams can still do
Evidence for how ICT4D teams specifically have adapted is thin; most of the published guidance is general to development evaluation. Several lower-cost approaches have a documented track record, with limits.
Use data someone else already collects
J-PAL’s Innovations in Data and Experiments for Action initiative was set up to expand the use of administrative data in randomized evaluations. Its premise is that experimental evaluations using administrative data can be cheaper and faster than those requiring new primary surveys. For digital programmes, relevant administrative records might include health facility reporting, education management data, or transaction logs held by a service provider.
The constraint is access and quality. Administrative data exist only where a system captures the outcome, and they reflect what the administrative system counts, not necessarily what the evaluator needs. In low-capacity settings, the records may be incomplete.
Shift from in-person to remote surveys
Phone-based methods, including computer-assisted telephone interviews, SMS, interactive voice response and online surveys, expanded rapidly during COVID-19 and are documented in resources such as Innovations for Poverty Action’s phone survey guidance. They cut travel and enumerator costs.
They also introduce selection problems that matter acutely in ICT4D. A phone survey reaches people who have phones and network coverage, which is exactly the population an inclusion study is trying to look beyond. Results for the least connected can be biased by construction.
Combine new data sources with representative surveys
A 2026 International Growth Centre policy brief on fragile and conflict-affected settings describes three practical innovations: mobile phone data, rapid-response phone surveys and satellite imagery such as nighttime lights. Its central caution is that these are complements, not substitutes, for official statistics, because they face coverage gaps, model drift, privacy risks and representativity constraints. Hybrid designs that integrate them with representative surveys and administrative data perform best.
Why digital programmes have an edge, and a catch
Digital interventions often generate their own records: platform logs, message delivery reports, transaction histories. That makes them better candidates than most sectors for evaluation built on existing data. The catch is that these records usually describe users of the service, not the people who never reached it, and they are held by providers who may have little incentive to share them. Data-sharing terms need to be negotiated before a programme starts, not after its funding changes.
Where the evidence is contested
Two debates sit underneath all of this.
First, whether cheaper methods can preserve causal credibility. Administrative-data trials and remote surveys can sustain rigorous designs, but some of the cost savings come from narrowing the outcomes measured, so the evidence base risks tilting toward whatever is easy to count.
Second, whether evaluation itself should be protected in a downturn. The argument for doing so is that scarce money should go to what works. J-PAL, in its own post-cuts communications, has pointed to more than 1,200 completed randomized evaluations among its network’s work as a basis for directing constrained funds.
A practical takeaway for research teams
For a team designing an ICT4D evaluation under a smaller budget, three decisions carry most of the weight. Decide which outcome genuinely needs primary data collection and which can come from an existing administrative or platform source. State, in the design document, who the chosen method will fail to reach, and report that limitation next to the findings rather than in an appendix. Finally, identify early which external datasets the study leans on and what happens to the design if one disappears, since the DHS episode showed that even long-standing public datasets can go offline. J-PAL’s IDEA initiative and the International Growth Centre’s innovative data solutions brief are useful starting points for the methods discussed above.