Data shared through digital public infrastructure (DPI) can make tax administration more effective and compliance easier. By connecting tax data with information held by other agencies or the private sector – like payrolls, bank records, or property registers – tax authorities can reduce paperwork and fraud.
But these benefits aren’t automatic. As institutions become better at connecting information, citizens need clearer ways to understand and challenge how their data is used. That means setting clear rules governing data exchange, how resulting decisions are made, and how taxpayers can challenge errors.
Our recent policy brief on governing the exchange of tax data, part of ICTD’s research on DPI and Tax, examines this challenge. We argue that data sharing must be governed as an exercise of public authority. Its legitimacy depends on the purposes it serves, the limits placed on it, and the protections available to those affected.
More connected data means more public power
Data sharing can bring substantial benefits. It allows authorities to cross-check declarations, detect inconsistencies and simplify compliance, while also enabling taxpayers to spend less time supplying information already held by the state.
But these connections can also determine which taxpayers are flagged for scrutiny or asked to explain a discrepancy. Once information feeds into an assessment, penalty, or enforcement action, the consequences can be significant.
The legal challenge is to make concrete rules governing those connections. Broad powers to obtain information should be subject to clear conditions for each exchange: which data is necessary, who may access it, how it may be reused, and who is responsible for its accuracy. Tax confidentiality, data protection, and administrative justice must work together across the entire data lifecycle, from collection and exchange to decision-making, correction, and deletion.
Who carries the cost of a mistake
An incorrect employer report or a mistaken identity match can affect a tax decision. In a connected system, an error in one institution may also travel to other institutions and influence further decisions.
While the administration gains a wider view of the taxpayer, the taxpayer may see only the resulting demand. Resolving that error can require discovering which organisation supplied the information, obtaining the relevant records, and persuading several bodies to correct them – a major undertaking.
Even accurate data can be misleading without context. A payment record shows money received but may say little about whether it represents business revenue, a loan, or a household transfer. Irregular earnings can also be misrepresented by assumptions designed around stable employment. Importantly, the burden of correcting errors falls unevenly. People with incomplete documentation, limited digital access, or fewer resources for professional advice may find it particularly difficult to contest a decision. Fair administration requires accessible explanations and workable procedures for review and correction.
A warning from Robodebt
Australia’s Robodebt scheme illustrates harm that can result when data collected for one purpose is reused without adequate legal and procedural safeguards. Between 2016 and 2019, the government used an automated system to compare the income welfare recipients reported each fortnight with their annual income from tax records, averaged evenly across the year. For many people whose earnings fluctuated, this produced debts they did not owe. The government later accepted that around 470,000 debts had been wrongly raised, and a Royal Commission on the Robodebt Scheme was launched, which found it was “neither fair nor legal”.
The lesson applies directly to tax administration – information that is suitable for one administrative purpose may require different checks before supporting another. Institutions need to establish that the data, its context, and the assumptions applied to it can justify the decision being made. Data availability should not be treated as proof that is use is lawful, appropriate, or sufficiently reliable.
What we are finding in Kenya and South Africa
Our comparative research on Kenya and South Africa points to where the gaps lie. Both countries give their revenue authorities road powers to obtain data, and both have modern data protection laws. Tax law in both is detailed about what the revenue authority may disclose and to whom – but it says much less about the data the authority receives and how it may use that data afterwards.
Those questions are left to general data protection law, which allows any processing that another law requires. This points back to the tax statutes. As a result, neither country has published tax-specific rules for individual exchanges. These would cover who asked for the data and on what basis, what record is kept, what else the data may be used for, and how errors are corrected.
The legislative tools to fill this gap already exist. South Africa has approved data protection codes of conduct for banking and credit bureaus. Kenya’s regulator has issued a general data sharing code for government and guidance for sectors such as health and education. Neither country has a code or guidance for tax. Data sharing agreements and impact assessments, if they exist for tax data, are not published, and taxpayers, courts, and oversight bodies cannot check rules or assessments that are not public. Most of what is needed could be done under existing law.
What should taxpayers expect?
Our brief sets out six dimensions of data governance, from legal clarity and oversight to recourse and inclusiveness. These principles have practical implications. A taxpayer should receive an intelligible explanation of a decision and be able to access the information needed to challenge it. A correction process should make it clear which institution must act and ensure that corrections reach every institution that has received or relied on the faulty record.
Data-sharing arrangements should specify permitted purposes, access, retention, and onward disclosure. Independent oversight bodies need the powers, expertise, and resources to examine whether those limits are respected. Where automated tools inform consequential decisions, meaningful human review requires staff who can question the output and reconsider the case.
In taxation, these safeguards must be adapted to the state’s extensive authority to compel information and enforce payment.
Build accountability into implementation
Governments can begin mapping data flows and identifying responsibility at each stage. UNDP’s Governance Assessment Framework for Data Exchange Systems can help with this.
Implementation also requires listening to those who use and are affected by these systems. Small businesses, informal workers, and taxpayer associations can identify confusing notices, difficulties with identity verification and correction procedures that are hard to navigate. Their involvement should continue through design, testing, and review.
Success should be assessed through both administrative results and taxpayers’ experience, including whether compliance becomes easier, decisions are better explained, and mistakes are resolved promptly. Those are tangible reasons for people to trust a more connected tax administration.
Two forthcoming ICTD papers will take this work further: the first develops a normative framework for governing tax data exchange; and the second sets out our findings on Kenya and South Africa in full.
