A Hundred Pay Packets

One

The claim, rebuilt

The World Bank's figure comes from UGAMOD, a tax model built on household survey data. An earlier run gave UGX 145.1 bn for the same schedule; its 2025 report gives UGX 148.7 bn. So we rebuilt it: first on a base resembling that model, then on what URA records.

Reproducible on a survey-based model; higher on URA's base. On a base resembling UGAMOD's (about UGX 2,666 bn of income tax across 1.71m payers), this model lands on UGX 140 to 162 bn. It also comes near Reform 2, though not Reform 1, which it overshoots. On URA's figures, the same schedule gives UGX 464 to 486 bn, before any response.

Why the difference? UGAMOD's own report says it simulated 52% of reported income tax for 2022, and reads that as high earners being under-represented in the survey. That is common for surveys anywhere. A 35% band falls chiefly on those earners.

Which base is right? One check was not used to build this model. The Budget Speech puts the cost of raising the threshold at UGX 96 bn. URA's base gives UGX 98 to 108 bn. The base like UGAMOD gives UGX 168 to 185 bn.

What we are not saying. That the World Bank made a mistake. Its report says the model is static, and a household model does things this one cannot, such as poverty and inequality results. “Mostly the base” is our inference: version 3.0 of UGAMOD's validation is not published, and the 2,666 bn is inferred from version 2.1's shares. The Budget's method is not published either.

Two

The half not adopted

The 2026 Act took the threshold and left out the 35%. It also added a 25% band, which lowers tax where the World Bank kept 30%. So what would a 35% band do now, on top of the law? There is no relief left to offset it.

That is 0.18% to 0.19% to 0.26% of FY2024/25 GDP. With top incomes doubled as a deliberate stress, and the high response, it is still UGX 324 bn. And 61 to 67 of every 100 monthly returns would pay more. This is arithmetic, not a recommendation.

Three

Where higher earners begin

The World Bank writes that only “the relatively higher income individuals” would pay more. Under Reform 3, tax starts rising at UGX 685,000 a month. A 35% band on the 2026 law would start at UGX 485,000. For comparison, the median in URA's FY2014/15 returns was UGX 440,000. Whether that is high is for the reader to judge.

• UGX 335,000Tax from July 2026 0Reform 3, against the old law -10,00035% band, against the law now no change
• UGX 485,000Tax from July 2026 33,750Reform 3, against the old law -10,00035% band, against the law now no change
• UGX 685,000Tax from July 2026 93,750Reform 3, against the old law no change35% band, against the law now +10,000
• UGX 1,000,000Tax from July 2026 188,250Reform 3, against the old law +15,75035% band, against the law now +25,750
• UGX 2,000,000Tax from July 2026 488,250Reform 3, against the old law +65,75035% band, against the law now +75,750
• UGX 10,000,000Tax from July 2026 2,888,250Reform 3, against the old law +465,75035% band, against the law now +475,750

UGX a month, on monthly chargeable income. These are exact, from the schedules. Type your own in the desk above.

Four

What the Budget's figure names

The Budget Speech gives UGX 96 bn for raising the threshold from UGX 235,000 to UGX 335,000 a month. The Act also added the 25% band, so the saving above UGX 5.82m a year is a flat UGX 165,000, not UGX 120,000. On URA's base, the threshold alone costs UGX 98 to 108 bn, and the whole schedule UGX 128 to 140 bn.

Model bars show the middle of each range. UGX bn a year.

The speech does not say whether its figure covers the band. So this is a question for the Ministry of Finance, not an error: does the UGX 96 bn include it? The relief also reaches every packet above the threshold. Someone on UGX 400,000 a month saves UGX 10,000 a month; someone on UGX 40m saves UGX 13,750.

Five

If people respond

Raise a rate and some reported pay may fall. Uganda has its own evidence, from URA's records, on the 2012 rise from 30% to 40% at the top. The same study has changed its answer between versions. In 2021 it reported large responses. By 2023, with a tighter comparison, its preferred estimate is small and not statistically significant, though top pay kept falling and some income may have moved into dividends.

None of these measures what Reform 3 would change. It leaves the 40% top rate alone and raises the rate from UGX 5.82m to 120m a year, where no Ugandan study has measured a rate rise. Carrying these figures down the schedule is our assumption. Here is what moves the answer, one thing at a time.

On URA's base the high case removes about a third of the gain and the sign holds. On a base like UGAMOD's, the same assumption takes it to between UGX -7 bn and UGX 37 bn. Leaving the formal payroll is not modelled at all.

Six

What this cannot tell you

The biggest limit: the shape of pay is over ten years old. URA's band table has been published only for these five years. We searched URA's data books, revenue reports, research bulletins and Research Lab pages, the UGAMOD reports, the national strategy documents, and the wider web, and found nothing later.

1. 2010/11
2. 2011/12
3. 2012/13
4. 2013/14
5. 2014/15
6. 2015/16
7. 2016/17
8. 2017/18
9. 2018/19
10. 2019/20
11. 2020/21
12. 2021/22
13. 2022/23
14. 2023/24
15. 2024/25

• Bands published
• Totals only, from returns
• Collections only

So the model keeps that FY2014/15 shape and scales pay until it raises each year's total. The latest public totals are FY2022/23: 11,402,638 returns from 37,120 employers. PAYE collected reached UGX 5,386 bn in FY2024/25. A current band table exists only inside URA, and it would replace this assumption.

• Grouped, not individual. One packet is 1% of monthly returns, not a person. Results hold at band level only.
• PAYE only. Individuals with business income face the same schedule. Their revenue is not modelled, which is not the same as zero. Flat-rate second jobs are inside the totals, which slightly overstates the base.
• Statutory, not final, burden. Who finally bears a tax may differ from who pays it.
• Not modelled: growth, jobs, moving off payroll, shifting into dividends, the cost of administration, and inflation wearing thresholds down again.
• No poverty or inequality results of our own. Those need household data. UGAMOD's earlier run put this schedule at 21.8% to 21.65% poverty and a Gini of 39.83 to 39.6.

Side by side

No best option

These sit in schedule order, not ranked. Revenue is one column. Each design also trades relief, simplicity and what people can bear.