Work on South African fiscal data began in 2014; the National Treasury Secure Data Facility was later opened as a secure installation where accredited researchers access data without removing raw data. This model has produced nearly one hundred published studies and is beginning to spread to Uganda and Zambia. It demonstrates that rigorous economic knowledge can be produced without entrusting data to outside actors, relying on infrastructure that is less spectacular than a supercomputer, but more strategic.
The essentials
- The South African Treasury’s secure lab has enabled nearly 100 economic studies over ten years, using 29 sets of administrative tax data, without raw data ever leaving the administration.
- Uganda opened its own lab with more than 40 researcher users; Zambia launched its own in August 2025, according to UNU-WIDER.
- The central mechanism: researchers work in a computing environment controlled by the state, and export only validated results, never individual data.
- The real tension is organizational, not technical: getting tax administrations, universities, and ministries to collaborate around data that each fiercely protects.
- Between 2030 and 2040, this model could determine whether Africa produces its own economic science or remains dependent on analytical frameworks designed elsewhere.
Sealed fiscal data, circulating studies
The South African National Treasury manages tax registers, customs data, and corporate files. These datasets rank among the continent’s richest for understanding who earns what, who pays what, and how public policies transmit through the real economy. They are also, by nature, confidential.
The National Treasury Secure Data Facility, established by the National Treasury, the South African Revenue Service, and UNU-WIDER, resolved this tension through a simple architecture: accredited researchers work on computers disconnected from the external network. They analyze the data, build their statistical models, produce their tables. Results to be exported are verified by the NT-SDF’s Data Team according to statistical disclosure control rules to prevent identifying a taxpayer or company. Raw data remains in the administration. Only analytical knowledge leaves.
Since 2014, this lab has grown from 5 to 29 available datasets. Nearly one hundred studies have been published, covering income inequality, effects of tax reforms, and the structure of formal and informal labor markets. These publications inform public policy debates in South Africa, and they are produced by South African teams, using South African data, under South African control.
The problem this model actually solves
To understand why this system matters, one must grasp what it responds to.
Low- and middle-income countries face a classic impasse. Their administrations accumulate data—fiscal, social, health—but this data is either buried in inaccessible systems or extracted by foreign researchers in projects financed by international donors, with analyses produced outside the country and results that don’t always make it back to local decision-makers. Two symmetrically problematic extremes: complete administrative secrecy, or extraction without return.
The secure lab proposes a third path. Data remains sovereign. Researchers access it under conditions. Results enrich national public policy. And the system can progressively integrate international researchers under the same rules, without requiring data to be transferred abroad.
It is the capacity to produce reliable internal diagnosis that this model protects. A government that cannot measure its own income inequality without asking an outside organization to do it for them remains dependent on that organization’s analytical framing—its assumptions, priorities, blind spots. This dependence is less visible than financial debt, but it weighs on the quality of public policy in the same way.
Adapting the model in Uganda and Zambia
Uganda opened a similar lab, backed by its national tax authority. Over forty researchers are accredited to it to date, according to UNU-WIDER. Early studies focused on tax compliance among small businesses and the redistributive effects of social transfers—two areas where administrative data produces results that surveys cannot deliver at reasonable cost.
Zambia launched its own lab in August 2025, as the country navigates fiscal reform and debt renegotiation. Internal analytical capacity allows it to negotiate with creditors or development partners by bringing its own numbers to the table.
These two cases also illustrate conditions for success. The model requires a fiscal administration sufficiently structured to manage quality data, a university or research network capable of training users, and political will to share sensitive data across ministries—which is often the hardest barrier to overcome. The technical obstacle is low. The organizational obstacle is real.
What administrative data offers compared to surveys
Studies produced in these labs are not interchangeable with what standard household surveys produce—the type of data national statistics institutes collect by sending surveyors into the field.
A household survey captures a snapshot: at a given moment, how many people report a certain income, certain spending. Fiscal administrative data is longitudinal: it tracks the same companies and taxpayers over several years. It allows measurement of trajectories—a business growing after a reform, an employee moving into a higher tax bracket, a sector responding to a trade policy shock.
This type of analysis is precisely what governments need to evaluate their own policies. Knowing that a tax reform increased revenues on average doesn’t say who paid more, who reduced activity, and whether effects were distributed evenly. Administrative data allows these questions. Household surveys alone cannot answer them.
This is why OECD countries have invested since the 1990s in similar infrastructure; Scandinavian tax registers became global references for inequality research, notably work that documented the concentration of wealth over the long term. South Africa is today creating comparable infrastructure, with ten to twenty years of lag behind Nordic countries, but with a clarity about data sovereignty that these countries did not need to cultivate in the same way.
Data governance and funding: the two fragile points
The model is conceptually robust. It remains fragile on two concrete points.
The first is inter-institutional governance. An administrative data lab requires that the tax authority, finance ministry, university partner, and researcher users operate under common rules. In South Africa, this agreement took time to stabilize. The National Revenue Service (SARS) long protected its data in silos—a comprehensible posture for an institution whose mandate is to collect taxes, not facilitate research. Progressive integration of its datasets into the lab required specific procedural guarantees and trust built over time.
The second fragile point is funding. These labs are not expensive: secure computing infrastructure and analyst salaries represent modest budgets compared to other government expenditures. Yet they often depend, in startup phases, on external financing: foundations, multilateral donors, international research programs like UNU-WIDER. This dependence creates long-term vulnerability, precisely in the area where the model seeks to build autonomy. Finland or the Netherlands finance their administrative data infrastructure from the national budget.
For the African model to hold, the same trajectory is necessary.
2035 horizon: producing economic knowledge about Africa
The issue is epistemological: in ten to fifteen years, the economic analyses guiding African public policies will be produced either by local institutions with their own infrastructure or by outside actors.
Two trajectories are plausible. In the first, secure data labs spread beyond South Africa, Uganda, and Zambia. Tax administrations in countries like Ghana, Kenya, Rwanda, or Côte d’Ivoire, which have invested in digital collection systems, progressively open their data to national research under similar conditions. A critical mass of researchers trained in this method forms, producing publications in English, French, and Swahili that feed regional debates. Governments arrive at international negotiations—with the IMF, private creditors, trading partners—with their own diagnoses.
In the second trajectory, organizational constraints prevail. Administrations remain compartmentalized. External financing dries up after pilot phases. Trained local researchers leave for international institutions offering better conditions. Data continues to be extracted case by case in projects led by institutions based in Washington, Helsinki, or Paris, whose research questions reflect their own analytical priorities.
This trajectory is not catastrophic—many useful analyses will continue to be produced—but it perpetuates an asymmetry.
What will likely separate these trajectories is less technology than institutional politics. Signals to watch for: the shift to national budget funding of existing labs, the emergence of regional networks sharing protocols (similar to what Europe built with Eurostat for harmonized data), and the capacity of African universities to train enough economists skilled in administrative data analysis to not depend on a handful of experts.
There is an instructive precedent in an adjacent field. Mobile payment systems allowed countries like Kenya or Tanzania to develop autonomous financial infrastructure, partly because local regulators retained control of the rules of the game—a dynamic documented in analysis of the effects of payments interoperability. Research data sovereignty follows analogous logic: what matters is who sets access rules, not who owns the servers.
UNU-WIDER, which supports several of these labs on the continent, framed the question in direct terms in its 2026 publication: Africa has the data. The question is whether it will build the institutions to analyze it itself. The South African lab has offered for ten years a partial and verifiable answer to this question. The challenge for the next decade is whether this answer will remain an exception or become the norm.
Sources
- UNU-WIDER, Secure research data labs: Unlocking evidence-based policy in Africa through data (2026)
- UNU-WIDER, Africa’s Next Move and Why Data Sovereignty Is the Missing Key (2026) [UNU-WIDER, no guaranteed URL]
- South African Revenue Service (SARS), annual reports on tax data management
- National Treasury of South Africa, documentation of the Secure Micro Data Laboratory, University of Cape Town
