In the 18th century, philosopher Jeremy Bentham envisioned the panopticon: a circular prison where guards could observe all inmates without the prisoners knowing whether they were being watched or not. The mere possibility of surveillance, Bentham theorized, would be enough to ensure compliance. The introduction of e-invoicing to monitor Value Added Tax (VAT) is the 21st-century tax echo of this concept. The results in Uganda reveal the promise of technological enforcement but also its limits without deeper institutional reform.
When the taxman can see everything
In January 2021, Uganda introduced the Electronic Fiscal Receipting and Invoicing System (EFRIS) to improve VAT compliance. EFRIS required businesses to transmit sales information to the Uganda Revenue Authority (URA) in real time. Like Bentham’s panopticon, EFRIS doesn’t just collect information, it creates the perception of constant observation. Every transaction becomes visible, every invoice verifiable, every discrepancy potentially detectable.
New research from UNU-WIDER, in collaboration with the ICTD and URA, examines what happened when the URA tightened the screws even further. In April 2022, approximately 5,000 larger firms were mandated to integrate their accounting systems directly with the government’s monitoring infrastructure. No more manual data entry. No more opportunities to massage the numbers before submission. The panopticon’s gaze appeared inescapable.
The results were dramatic. Firms subjected to this stricter enforcement reduced their reported purchases by 43% and increased their reported VAT liabilities—the total value of collected VAT a business needs to return to tax authorities—by nearly 150%. These aren’t small adjustments; they suggest that systematic misreporting had been widespread and that the digital panopticon successfully curbed it.
Ghost firms and invoice mills
To understand why digital surveillance matters, we need to understand how VAT fraud works in practice. Uganda’s VAT system, like those worldwide, operates on a simple principle: businesses charge VAT on sales, deduct VAT already paid on purchases, and pay the difference to the government. This creates a paper trail where buyers want receipts to claim deductions, theoretically keeping sellers honest.
But in contexts with weak enforcement capacity, this self-policing mechanism rarely works perfectly. Before EFRIS, researchers found that Ugandan buyers and sellers reported different amounts 79% of the time, with most discrepancies conveniently reducing tax owed. Some businesses ran sophisticated schemes involving ‘ghost firms’ or ‘invoice mills’; companies that generate and sell invoices for fictitious purchases, allowing others to inflate their costs and shrink their tax bills.
The traditional paper trail created incentives for honesty only if taxpayers believed the government would actually cross-check the receipts. The URA collected transaction-level data but historically only analyzed it during audits. The threat of detection remained largely theoretical.
The panopticon’s blind spot
Here’s where the story gets complicated. When enforcement tightened on input costs, the VAT firms paid for goods, services, or raw materials for their operations, they didn’t simply accept higher tax bills. They adapted, and not always in expected ways. The research shows that, on average, firms also reduced their reported sales by 39%. This was counter-intuitive, and the opposite of what e-invoicing is supposed to achieve.
Importantly, this sales suppression appears to be strategic rather than real. The researchers found no corresponding decline in reported revenues on corporate income tax returns or reductions in employment. Firms weren’t actually shrinking; they were shifting their evasion to less verifiable margins.
This finding reveals a potential limitation of surveillance-based enforcement. When firms could no longer exaggerate their input costs, they apparently resorted to under-reporting sales. The panopticon’s gaze, it turns out, has blind spots.
Spillover effects
After EFRIS was launched, the number of businesses filing VAT declarations surged. Some 12,000 firms that had never filed VAT returns began doing so, many of whom had previously filed corporate income tax returns showing sales well above the VAT registration threshold.
This finding suggests that digital surveillance changed perceptions of enforcement probability. Even firms operating outside of the VAT system began to believe that evasion would be detected. The panopticon’s power, Bentham understood, lies not in actual observation but in the uncertainty of being observed.
Yet this growth in VAT filers hasn’t translated to proportional revenue increases. Many new filers submit nil returns or declare minimal activity. And many existing firms carry enormous negative balances: the government technically owes them money from previous periods when their input VAT exceeded their output VAT. These balances, sometimes accumulated through years of dubious claims, now constrain how much revenue the system can generate.
The limits of seeing
Uganda’s experience with e-invoicing demonstrates that technology alone cannot solve governance challenges. EFRIS generated unprecedented data on business transactions, but firms found new margins for evasion. The system increased compliance with input claims while simultaneously being vulnerable to sales suppression. It brought thousands of new filers into the tax net while leaving existing structural problems, like massive refund backlogs, unresolved.
The lesson here isn’t that digital surveillance doesn’t work. The 150% increase in VAT liabilities among closely monitored firms proves it can be powerful. Rather, the lesson is that surveillance is insufficient without addressing the deeper architecture of tax systems, including taxpayer sensitization and integrating the new data into the URA’s enforcement processes.
Bentham’s panopticon was never built. Perhaps there’s wisdom in that. The fantasy of perfect surveillance, whether architectural or digital, founders on the same rocks: human adaptability, system complexity, and the stubborn reality that seeing everything is not the same as solving everything.
This article was first published by the WIDERAngle, UNU-WIDER. The original is available here.
