Does Tax Technology Pay for Itself?

W/P A

The first promises ran far ahead of delivery, then the official count stops

URA's own table for FY2019/20 sets what each measure was expected to raise against what it records. All administrative measures reached 8.39% of target.

This tests the forecast, not the systems. Stamps began only in November 2019, EFRIS was piloted in May 2020, and COVID-19 cut the year short. We do not know how URA measured the gain. And the forecast itself was not settled: URA's table says UGX 150 bn for stamps, while the Commissioner General reportedly told Parliament to expect UGX 47.7 bn instead.

After that, the count stops. In the reports we read for FY2020/21 to FY2022/23, tables give administrative gains only in total. For FY2022/23, URA credits arrears management with UGX 1,033 bn of UGX 1,474 bn, and calls EFRIS and stamps “major sources of information”, with no figure. We could not reach URA's reports for FY2023/24 and FY2024/25, so a figure may exist there.

Back to line 5, line 6 of the lead schedule

W/P B

Ten cost lines, and most of the slots are empty

A return starts with a cost. We listed every cost we could name, for government and for taxpayers. Empty does not mean zero, and it does not mean anything was hidden. It means we did not find it in public documents.

URA's budget does not help, as it is not split by system. In FY2024/25, it spent UGX 47.4 bn of a development budget of UGX 57.4 bn, with nothing to say how much went on either one. Our request to URA for these lines is drafted.

Back to line 1, line 2, line 3, line 8 of the lead schedule

W/P C

Collections rose about as fast as the economy

The rise in collections is often the first evidence offered for digital reform. Try it three ways, and from three starting years.

Whichever start you pick, the headline figure cannot be evidence for either system by itself. This neither proves nor rules out a contribution. It only says that collections rising is not a return. W/P H lists what else has to be ruled out.

Back to line 8 of the lead schedule

W/P D

EFRIS changed what its most tightly bound firms report

Lees and colleagues compare 3,511 firms mandated onto system-to-system integration from April 2022 with 5,931 others, using synthetic difference-in-differences, to June 2023. Tap a line to read it.

Fake or inflated input claims became harder, which is a real result. But it is not the same as more revenue: reported sales fell too, and the research brief speaks of “limited short-term revenue effects”. It also does not measure EFRIS as a whole, as all VAT firms were mandated at once, so there is no comparison group for that. About 12,000 first-time VAT filers coincided with EFRIS and enforcement, and the paper does not claim to have caused them. These authors include URA staff. Their own advice is not to mandate this for all firms, as smaller firms may not manage the cost.

Back to line 7 of the lead schedule

W/P E

Stamps: down in the first year, perhaps up in the second

Namunane compares treated and other excisable firms over 517 firms and stops at FY2020/21. In the first year government paid for the stamps, and in the second firms did.

The author warns of low power and possible pre-trends, and we repeat the warning. Since then, firms on stamps have grown from 143 to 925, and URA raised UGX 24.25 bn of assessments from stamp data in FY2022/23. Those show use, not revenue caused. No later study covers the wider products.

Back to line 7 of the lead schedule

W/P F

The stamp bill firms carry, as far as a model can say

Manufacturers and importers buy the stamps. Nobody publishes the bill, so we modelled one for FY2024/25. Move a pack size and see how far it travels.

Stamps have become cheaper

The direction is clear, even if the size of the bill is not. Below, prices from FY2019/20 to FY2024/25 on a log scale.

A 2026 press report gives lower prices still, and it conflicts with URA's guide. We have not used it. Comparisons with other countries' stamp prices are in dispute, so we do not publish one.

Back to line 3 of the lead schedule

W/P G

EFRIS is really two systems

The evidence and the burden sit with different firms. In June 2024, 13% of users issued 94% of invoices, and these are the firms W/P D studies. The web portal has 79% of users, and 5% of invoices. In a survey fielded during the 2024 traders' strike, 51% of firms reported harm from EFRIS enforcement.

The fiscal device barely happened: 13% of the 20,000 planned had been bought by June 2023. EFRIS adoption holds the detail.

Back to line 4 of the lead schedule

W/P H

What else has to be ruled out

Before any rise is credited to a system. None of these is separated from EFRIS or stamps in official reporting.

A larger economy

Nominal GDP rose 40% from FY2021/22 to FY2024/25. See W/P C.

Inflation

Part of that nominal rise is prices, which raises collections with no change in compliance.

New tax measures

Changes in the tax amendment Acts over the same years raise revenue on their own.

Arrears work

URA's own main attribution for FY2022/23. See W/P A.

Audits and enforcement

These came with both systems, and cannot be separated from the technology itself.

W/P J

Simpler alternatives, as questions

A fair appraisal asks what else could have done the job. These are not recommendations, and none has a costed comparison.

Risks on record: dependence on one supplier, single-sourced with terms not public, and a five-year term from October 2018 whose renewal we could not confirm. Exclusion through digital skills, connectivity and a UGX 1.1m device. Privacy and security were not assessed here.