Counting Power: what censuses reveal about aid, governance and knowledge

It’s a subject that’s deeply political, but often treated as a technicality: censuses, surveys and production of development data. On 23 February, I had the privilege of delivering   a guest lecture on the subject at Sciences Po.

I called my lecture Counting Power: The Political Economy of Censuses, Surveys, and Foreign Aid. My aim was not to explain questionnaires or sampling frames, but to explore a more fundamental set of questions: who counts, what counts, and why that matters for power, governance, and aid.

What follows are some reflections from that session — and the beginnings of an argument I hope to develop into a longer essay.

The hidden infrastructure behind development decisions

When we talk about development, we often focus on the visible outcome of decisions: infrastructure projects, social programmes, economic reforms, diplomatic negotiations. Think of debates around large-scale investments such as the Ethiopian Renaissance Dam, or social protection programmes like cash transfers under the Urban Productive Safety Net Programme. These discussions tend to revolve around cost, impact, or geopolitics.

What we rarely discuss is the invisible layer that underpins those decisions: data.

Population figures shape how resources are distributed. Poverty indicators determine where programmes expand or contract. Education and employment statistics influence how countries are perceived internationally. Behind every headline number lies a complex process of negotiation — between governments, donors, statisticians, and communities.

Yet counting people, something that sounds neutral, is never purely technical. It is a deeply contested political activity.

Why counting can be politically risky

One of the key arguments I shared with students is that the decision to run a census is itself political. Governments weigh constitutional requirements, electoral timing, donor expectations, and social tensions before deciding whether — and how — to count.

During my time working on census processes in Ethiopia, debates around ethnicity illustrated this clearly. Ethnic categories were constitutionally tied to representation and resource allocation, meaning demographic numbers carried significant political weight. Fears that groups might inflate their population counts made enumeration itself politically sensitive. In that context, postponing the census was not simply a technical failure — it was a political decision.

A different dynamic unfolded in Sierra Leone, where concerns about legitimacy led the World Bank to withdraw funding close to an election period. The government proceeded, but the financing shift reshaped timelines, staffing, and ultimately public perceptions of credibility.

These examples highlight a broader lesson: data collection can destabilise political equilibria.

Surveys as negotiated instruments

Another theme that resonated strongly with students was the idea that survey questions are never neutral.

Deciding whether to include modules on disability, gender empowerment, or specific health issues often involves negotiation between donors pushing global agendas and ministries prioritising national plans. Sociologists of statistics describe this as the “politics of classification”: categories do not merely measure society — they help construct it.

In practice, entire topics may disappear from questionnaires because they are politically sensitive or lack institutional backing. What emerges as a final survey instrument is therefore a map of power relations.

Aid doesn’t just fund data — it shapes it

Development data is rarely neutral because the act of financing it shapes what becomes visible. Censuses and surveys are expensive, and in many countries depend heavily on foreign donors. When you fund measurement, you inevitably influence priorities — sometimes explicitly through conditions, but often more subtly through what gets supported, what gets delayed, and what never gets measured at all.

We are seeing this dynamic very clearly today with the uncertainty surrounding large survey programmes such as the Demographic and Health Surveys (DHS). When global funding priorities shift, entire data ecosystems become fragile. Surveys are postponed, modules disappear, and long-running time series risk breaking — not because countries no longer need the data, but because the financial architecture that supports them becomes unstable.

This volatility reveals something uncomfortable: development statistics are often built on funding structures that can change faster than the realities they aim to measure. When financing shifts, operational decisions follow — fewer enumerators, reduced supervision, narrower questionnaires. What later appears as a methodological problem is often the downstream effect of earlier political or financial negotiations.

At the same time, the question of ownership remains unresolved. Many argue that governments should finance their own statistical systems to strengthen sovereignty and priority-setting. Yet data collection is frequently treated as an invisible layer beneath development projects — essential, but rarely politically rewarding. Roads, dams, and social programmes are visible achievements; high-quality population data is not. As a result, statistical systems struggle to compete for limited domestic resources.

The paradox is that data sits beneath everything: infrastructure planning, social protection targeting, climate adaptation, and international financing decisions. And yet it is often funded last and questioned first.

This is why I increasingly think of development statistics as co-produced within aid architectures. Donors do not simply fund data; they influence how knowledge is produced.

Trust, labour, and the human side of data

Enumerators are not neutral data collectors; they are temporary street-level bureaucrats translating official rules into everyday interactions. Who gets hired, how they are paid, and how they are supervised all shape the quality of the data that emerges.

In many contexts, census jobs are among the most visible short-term employment opportunities, meaning recruitment reflects local expectations as much as formal criteria. Incentives matter deeply. Payment delays, weak supervision, or unclear contracts can erode morale and trust, influencing response rates and how strictly protocols are followed. What is often described as a “methodological problem” is frequently a labour and governance issue.

Trust sits at the centre of enumeration. Citizens decide whether to participate based on how they perceive the state and the individuals representing it, while enumerators rely on community relationships to gather accurate information. Training aims to standardise behaviour, but interpretation always exists — whether in how dates of birth are recalled or how time spent collecting water is measured. Small variations in judgement can translate into significant shifts in national indicators.

Seen this way, data quality is not just technical — it is social. Statistics emerge from incentives and trust, negotiated daily between enumerators and the communities they serve.

Global standards, local realities

Working across multiple countries has also highlighted the tension between global statistical standards and local realities. International frameworks aim to ensure comparability, but on the ground, conditions rarely align perfectly with methodological ideals.

Field teams adapt to seasonal migration, connectivity issues, or institutional constraints. The resulting datasets appear uniform at the global level, yet they often conceal layers of negotiation and improvisation. Global indicators can obscure the lived complexity behind the numbers.

Digitisation and the new politics of data

One of the most visible changes in recent census cycles has been the transition to digital enumeration. Across Kenya, Ghana, Mauritius, Sierra Leone, and Ethiopia, tablets replaced paper forms and real-time monitoring transformed supervision.

Digitisation has improved efficiency and traceability. But it has also redistributed power. Software embeds decisions about question logic and validation rules, reducing local discretion while expanding central oversight. At the same time, reliance on donor-funded platforms introduces new dependencies.

Technology does not remove politics — it reorganises it.

A future of shrinking aid and growing knowledge gaps?

The lecture concluded with a question about the future. As foreign aid becomes more contested globally, what happens to development data?

Censuses are often among the first activities to be postponed when funding declines. Surveys may be scaled back, and statistical offices face difficult choices about what to measure. The risk is not only fewer datasets but widening knowledge gaps between countries with strong domestic capacity and those reliant on external support.

In a world of shrinking aid, we may also see shrinking knowledge.

Flying blind – development without data

Delivering this lecture reminded me how central data is to debates about governance, legitimacy, and development. Statistics are not just technical outputs; they are negotiated realities shaped by power, funding, trust, and technology.

I hope to expand these reflections into a longer essay exploring the political economy of development data — and the implications of a world where counting itself is becoming more uncertain.

Because behind every indicator lies a simple but profound question:

Who counts — and who decides what counting means?