Table of Contents
Governments increasingly recognize that administrative data are not just a by-product of public services. Used responsibly and effectively, they can improve service delivery, play a vital role in the national digital agenda, and inform policy against a global backdrop of declining response rates to statistical surveys.
But the path from “data exists” to “data improves decisions” is rarely straightforward and often requires data sharing across institutions. Indeed, evidence shows that data sharing initiatives can generate US$5 in social returns for every dollar invested. Data sharing by mobile operators supported Nepal’s 2015 earthquake relief; while in Brazil, the government has linked health and social data coming from different public agencies and given access to accredited researchers in order to study the impact of health and social policy on over 100 million citizens.
The harder truth is that most governments do not fail at data sharing because they lack databases. They fail because the institutional conditions for safe, sustained, and useful sharing are weak. Public agencies often assume that another department has a reservoir of clean, well-documented, timely data, and all it takes is for the gatekeepers to switch the taps on. But in practice, administrative systems have usually evolved over decades and come with no instruction manuals. With each fiscal policy change, new tables appear in a legacy database. Decades later, understanding this data requires tacit knowledge spread across multiple individuals and teams. The data may be valuable, but it is rarely ready.
Recent World Bank Group data governance assessments in the Indian state of Kerala, as well as in Mozambique, point to the same lesson from different contexts: data governance is “soft infrastructure” as much as technology. The state of Kerala had already developed sophisticated resilience-related platforms and applications, but the diagnostic found that many initiatives were not underpinned by a common set of standards, methods, and policies. Mozambique’s diagnostic similarly emphasized that a robust data governance framework is required to support data collection, management, and sharing within the national digital ecosystem
That is why public sector data sharing needs to be designed as an operating model, not a one-off transaction. Elements of technical architecture such as APIs, Trusted Research Environments, and interoperable standards are important, but not sufficient by themselves. Governments also need to establish a vision for data sharing, assign leadership, fund stewardship, build trust, and create incentives for agencies to participate.
A pragmatic roadmap should begin with five priorities:
1. Start with a clear public purpose
For data sharing to be sustainable, governments must define the benefits they aim to generate and ask themselves: Which specific government priority or transformation program will it support? Alignment with a strategic program creates accountability and urgency, as well as buy-in from agencies and the public whose data is being shared. This matters because vague data-sharing mandates quickly lose momentum. A request to “share more data” often sounds more like a new administrative burden than a mission to improve public services and transparency.
Purpose also helps governments decide what not to do. Not every dataset should be linked. Not every data project deserves equal priority. A clear use case helps focus scarce capacity on the datasets, standards, safeguards, and partnerships that can generate the greatest public value.
2. Build governance around an anchor institution
Cross-government data sharing requires a sustainable data ecosystem and a credible institutional anchor. For statistics and policy research, national statistical offices are the obvious place to start, while central digital authorities (the government entities responsible for digital standards and shared platforms) and/or digital service delivery units are best for operational data sharing. Health, tax, education, social protection, and civil registration agencies often bear the cost and risk of sharing data while others reap the analytical benefits. Expecting busy operational teams to curate, document, anonymize, negotiate, and support data users on the side is unrealistic. Dedicated gateway teams, with data engineers, metadata specialists, legal and privacy expertise, and domain knowledge, can make sharing safer and more predictable.
Finally, governments should look beyond the public sector to the many successful examples of public-private partnerships on data sharing that can generate significant impact, for instance by integrating retailers’ price data in the production of inflation statistics. At a global level, the Development Data Partnership has shown how collaboration with the private sector can support projects to model the impact of transport policy choices on air quality in Southeast Asia, identify cost-effective solutions to closing internet connectivity gaps for schools in Jamaica, and monitor the cross-border mobility of AI talent at the global level.
3. Fix the incentive problem
Data sharing stalls when data-rich departments bear all the political risk of data being used to highlight shortcomings in their own policies, as well as the costs of curating and sharing their data.
The temptation, then, is to mandate data sharing. Unfortunately, this rarely works, as there are legitimate reasons (as well as convenient excuses) why a department might struggle to share data.
The first challenge is political. This is why governments must align data sharing with clear government priorities. For example, sharing children’s health and education data across agencies works better within a broader reform program than as a standalone effort.
The other challenge is linked to resources. A central agency is better positioned to coordinate a system of incentives than multiple departments acting separately. This could include grants to curate linked datasets that meet the needs of multiple departments, capacity building aimed at a cohort of smaller agencies, as well as core funding to ensure stability beyond time-bound programs. Incentives should reward reuse of data as well as sharing. A linked and anonymized dataset created for one analysis may be valuable for planning ministries, local governments, gender agencies, researchers, and service delivery teams. Treating reusable data assets as data infrastructure changes the investment logic. It makes the case for funding data stewardship before the urgent analytical request arrives.
