Tax Lab · Illicit financial flows
Can the Stamp Cross the Border?
Excise smuggling, fiscal-marking data and Articles 3(2) and 11 of the draft UN Tax Convention: what the data show, what the treaty text should say, and how two revenue authorities could check each other's stamps without sharing a single taxpayer's records.
Companion to the ICTD working paper of the same name (workshop draft, November 2026). Every figure below is recomputed in this site from the paper's own inputs and checked against its Stata output. The treaty text is still under negotiation.
The question
Uganda's Digital Tax Stamp proves that excise was paid on a bottle, a pack or a bag inside Uganda. It proves nothing once that good crosses into Kenya, South Sudan or DR Congo, because no neighbour can routinely read Uganda's marks, nor Uganda theirs. The Co-Lead's zero draft of the UN Framework Convention on International Tax Cooperation (A/AC.298/CRP.32, 21 July 2026) would make this worse in two places. Article 3(2) excludes from tax-related illicit financial flows those that "consist primarily" of non-tax crimes, and gives smuggling as an example. Article 11(1)(b), still in brackets, would allow exchange of "transaction information" to match exports with imports, but says nothing of fiscal marks. The paper asks whether the convention can reach excise smuggling, and how stamp data could cross the border.
What the data show
- The stake is large. In FY2024/25, domestic excise on goods now under the stamp was UGX 1,246.8 billion, 51.3 per cent of all domestic excise, up from UGX 385.1 billion in FY2015/16. The three largest heads (beer; soft drinks; spirits and waragi) account for 41.4 per cent of all domestic excise on their own.
- Neighbours record less than half of what Uganda says it ships. From 2015 to 2025 Uganda reported US$2.77 billion of stamped-goods exports to its six neighbours; they recorded US$1.21 billion of imports from Uganda, or 44 per cent. Of 403 product-partner-years with exports, 134 have no import record at all on the other side.
- The stamp has not closed the gap. Where both sides report, the median discrepancy on stamped goods was -0.39 before the stamp and -0.43 after it. Aligning each product to its own phase gives -0.46. The share of export records with no import record moved from 32.6 per cent to 33.8 per cent. Rice and steel moved from -1.53 to -0.14, but on 26 observations in each period, too few and too noisy for a difference-in-differences claim. This is description, not an estimate of the stamp's effect.
- The stamp reached the tax return five years late. Stamps were activated from November 2019, but for five years the stamp ran alongside the excise return rather than inside it. In URA's return extract, none of the 123,735 lines dated 2008 to 2023 carries a stamp identifier; the first appears in November 2024, when a web return pre-filled from the stamp system replaced the Excel return, in all ten stamped categories at once. In 2025, 91.4 per cent of lines carried one. URA now holds the stamp identity, the quantity and the duty in one system.
Paper, Figure 3.1
Domestic excise on goods now under the Digital Tax Stamp
Show the numbers as a table
| Financial year | Stamped goods | All domestic excise | Stamped share |
|---|---|---|---|
| FY2015/16 | 385.11 | 671.14 | 57.4% |
| FY2016/17 | 495.54 | 819.78 | 60.4% |
| FY2017/18 | 628.49 | 953.91 | 65.9% |
| FY2018/19 | 692.32 | 1317.04 | 52.6% |
| FY2019/20 | 656.67 | 1266.08 | 51.9% |
| FY2020/21 | 760.98 | 1479.98 | 51.4% |
| FY2021/22 | 818.61 | 1646.70 | 49.7% |
| FY2022/23 | 978.08 | 1920.66 | 50.9% |
| FY2023/24 | 1107.91 | 2171.17 | 51.0% |
| FY2024/25 | 1246.76 | 2429.71 | 51.3% |
Paper, Figure 5.1
Stamped goods: what Uganda reported exporting, and what its neighbours recorded
Show the numbers as a table
| Year | Reported by Uganda | Recorded by partners | Recorded share |
|---|---|---|---|
| 2015 | 164.8 | 92.3 | 56.0% |
| 2016 | 160.1 | 95.9 | 59.9% |
| 2017 | 164.9 | 79.3 | 48.1% |
| 2018 | 206.3 | 105.9 | 51.3% |
| 2019 | 188.2 | 74.5 | 39.6% |
| 2020 | 205.6 | 58.3 | 28.4% |
| 2021 | 261.7 | 98.4 | 37.6% |
| 2022 | 333.5 | 103.7 | 31.1% |
| 2023 | 272.4 | 104.3 | 38.3% |
| 2024 | 353.4 | 124.6 | 35.3% |
| 2025 | 458.9 | 275.1 | 59.9% |
Paper, Table 5.1
Where both sides report: the median gap, 2015-2019 to 2020-2025
Show the numbers as a table
| Series | 2015-2019 | n | 2020-2025 | n |
|---|---|---|---|---|
| Stamped goods, median D (value) | -0.39 | 122 | -0.43 | 147 |
| Stamped goods, median D (net weight) | -0.62 | 111 | -0.35 | 142 |
| Stamped goods, Kenya corridor, median D (value) | -0.48 | 21 | -0.64 | 28 |
| Comparator goods (rice, steel bars and rods), median D (value) | -1.53 | 26 | -0.14 | 26 |
| Imports into Uganda of stamped goods, median D (value) | -0.17 | 66 | -0.10 | 90 |
Paper, Tables 5.2 and 5.3
From the stamp to the tax return: a five-year wait
Show the numbers as a table
| Return period | Lines | With SKU | Share |
|---|---|---|---|
| 2008-2023 | 123,735 | 0 | 0.0% |
| 2024 | 35,119 | 6,571 | 18.7% |
| 2025 | 47,937 | 43,806 | 91.4% |
| 2026 | 1 | 0 | 0.0% |
What the text should say
Four amendments, each drafted to survive the objections already on the record: Kenya's concern about proving intent, the South Centre's objection to smuggling as an example, and the fear of overlap with organised-crime instruments. Struck-through words leave the zero draft; underlined words are new.
Article 3(2), second sentence and a new third sentence
It does not include illicit financial flows that consist primarily of non-tax crimes (e.g., [deleted: trafficking or smuggling][inserted: trafficking in narcotic drugs, firearms or persons]) even if the crime would also give rise to a revenue claim. [inserted: Notwithstanding the preceding sentence, a flow arising from the movement, or declared movement, across borders of goods whose importation or exportation is not prohibited by the law of a State Party, in breach of the laws of that State Party relating to customs duties, excise duties or other taxes on goods, is a tax-related illicit financial flow, without regard to the purpose of that movement, where it gives rise to a revenue claim of that State Party.]
Keeps the exclusion for crimes where tax is incidental, takes smuggling out of the examples, and adds an objective test for taxes on goods: a revenue claim, not the smuggler's motive.
Article 11(1)(b)
(b) [deleted: transaction information to allow the matching of exports and imports of goods and services][inserted: information, including transaction, fiscal-marking and tracking information, that enables States Parties to verify cross-border movements, and declared exports and imports, of goods and services and to match exports with imports, for the detection and prevention of tax-related illicit financial flows, in accordance with modalities to be set out in a protocol];
Names fiscal-marking data, reaches declared as well as physical movements, and leaves the mechanics to a protocol, so the convention carries the legal basis and the protocol carries the detail.
Article 7, purpose clause
... to ensure the effective taxation of income and profits from[inserted: , and the effective collection of taxes on goods and services connected with,] tax-related illicit financial flows by, and in accordance with the national laws of, the jurisdiction whose tax base has been reduced as a result of the tax-related illicit financial flow.
Whether or not Article 11 survives, the purpose clause should cover the collection of taxes on goods, not only the taxation of income and profits.
Article 12(2)(c)
(c) Setting up data analytics, exchange of information [inserted: and interoperable verification frameworks];
Capacity support should build the verification systems that make exchange usable, not only the legal channel.
If Article 11 is deleted, the anchor can move to Article 7: after "developing effective tools for the detection and prevention of such tax-related illicit financial flows", insert "including arrangements for the cross-border verification of fiscal-marking, tracking and other transaction information on goods". The paper's Annex B gives the full set of fallback and conforming texts.
How verification would work
The paper proposes an excise-verification protocol in three tiers. Only the first needs anything new at framework level, and it discloses no taxpayer data. Parties could reserve on the second and third while accepting the first.
| Tier | What moves | Mechanism | Disclosure risk | Legal basis |
|---|---|---|---|---|
| 1. Authentication | Valid or not valid; product class; issuing state | Real-time lookup, or a privacy-enhanced query of a hashed register of stamp serials | Very low | Framework commitment and a protocol on standards |
| 2. Matching | Aggregate stamped volumes declared for export, by product, month and destination | Periodic automatic exchange of anonymised statistics, with Article 15(3)-style cell suppression | Low to moderate | Protocol (opt-in) |
| 3. Investigation | Unit and transaction records for named producers or consignments | Exchange on request; simultaneous or joint audit | Higher | Article 11 safeguards; the MAAC or CRP.34 where in force |
Tier 2 is buildable now from data both sides already hold. URA's excise returns carry the stamp identifier on 91.4 per cent of 2025 lines, and every neighbour in the paper except South Sudan already reports its imports from Uganda to UN Comtrade. Monthly stamped volumes declared for export, matched against the destination's recorded imports by product and month, and suppressed wherever fewer than ten firms sit behind a cell, would turn the annual ceiling above into a monthly risk signal without a single taxpayer record crossing the border.
Stamp to Arrival builds it: a published matching engine run on these data, with an audit ranking, the excise the unmatched exports carry by corridor and product, and a working demonstration of the Tier 1 stamp check that catches forged, moved and cloned stamps.
A second lens: the PCM+ model
Data pulled 14 October 2026. A wider, exploratory model built before the paper: seven neighbours, eleven goods, both trade directions, and UNCTAD's PCM+ adjustments.
The model measures the same borders differently. For each neighbour, good and year it sets Uganda's customs record against the partner's, converts imports from CIF to FOB, removes re-exports, and classifies what remains as inward or outward under the UNCTAD-UNODC convention. It uses soap and bakery products, never stamped, as controls, and for cigarettes adds a market identity from WHO prevalence, World Bank population and URA stamp deliveries. Its results agree with the paper's: no effect of the stamp can be claimed either way.
- 836 comparisons are possible across 6 of 7 neighbours. Somalia filed no comparable report in any year and South Sudan only for 2025. The paper works with six: Kenya, Tanzania, Rwanda, Burundi, South Sudan and DR Congo.
- Cigarettes from Kenya: in 2019 and 2020, either side of the first phase, Uganda recorded US$13.8 million and US$13.1 million of cigarette imports from Kenya while Kenya declared exports of US$7.0 million and US$6.1 million, a remaining asymmetry of US$5.6 million and US$5.8 million after adjustments. Which side is wrong cannot be told from these data alone. The gap closed in 2021.
- The phase 1 difference-in-differences estimate is +0.29, against 4 placebo estimates reaching 0.26 in size. The real estimate sits only just outside the placebo range, too thin a margin on so few placebos to call an effect. Read it, as the paper reads its own comparison, as description.
- Only 39% of the measured gap on stamped goods sits with partners whose cigarette marks WHO records as carrying track-and-trace data. For the rest, Article 11 would deliver a legal basis to exchange information that does not yet exist in a matchable form, which is why Tier 1 needs a standard as well as a treaty.
Model settings, applied to every chart on this page
Comparisons
660
partner x good x year x direction, both sides reported
Inward, all years
$641M
money into Uganda: export over-invoicing or import under-invoicing
Outward, all years
$74M
money out of Uganda: export under-invoicing or import over-invoicing
Verifiable today
39%
of the gap sits with partners whose marks carry track-and-trace data
Module 5: IFF country risk assessment
Where the mirror gaps concentrate
Each neighbour is scored on two things: how often its records and Uganda's disagree by more than 50% of the flow (likelihood), and how large the disagreement typically is in dollars (impact). Somalia filed no comparable report in any year and cannot be scored.
SDG 16.4.1, trade-misinvoicing component
Inward and outward flows, year by year
The remaining asymmetry after PCM+ adjustments, split by the direction the money moved. This is the quantity the UNCTAD-UNODC framework asks countries to report under SDG indicator 16.4.1, restricted here to the goods Uganda stamps and its regional neighbours.
Hover a year. The sign convention follows the UNCTAD-UNODC statistical framework: export under-invoicing and import over-invoicing are outward flows; export over-invoicing and import under-invoicing are inward. Totals depend on which partners reported that year, so a dip can mean a missing report, not less leakage.
Module 1: PCM+ step by step
How much of the gap is just valuation and re-exports
Each row sums the absolute gap across the goods and directions selected, for one partner and year. The distance between the dots, which can widen as well as narrow the gap, is what valuation (CIF versus FOB) and re-exports explain; what is left is the remaining asymmetry that UNCTAD's PCM+ method treats as a signal of misinvoicing or unrecorded trade.
Module 3: did stamping change anything?
Asymmetry before and after each DTS phase
If digital tax stamps made under-declaration harder, the mirror asymmetry for stamped goods should fall after their start date by more than it fell for goods that were never stamped. The difference-in-differences estimate below measures exactly that, and the placebo strip repeats it with fake start dates to show how large a change arises by chance.
Goods in this phase: Beer, wines, spirits, soft drinks, bottled water, cigarettes. Comtrade covers only those in the HS list above.
Stamped goods
0.47 to 0.83
before vs from 2020; 438 comparisons
Unstamped controls
0.27 to 0.34
176 comparisons
Difference in differences
+0.29
negative = asymmetry fell more for stamped goods
The real estimate is larger in size than every placebo. That is a necessary, not a sufficient, condition for reading it as an effect.
Claim 1: the incentive
Cigarette price gaps and cigarette mirror gaps
The paper argues that a price gap across a border is what pays for smuggling. If so, partner-years with a larger price gap should show larger cigarette asymmetries. The vertical axis is what the importing country recorded minus what the exporting country declared, summed over both directions, after PCM+ adjustments.
Hover a dot. Pearson r = 0.60 on n = 19 partner-years. Hollow dots are excluded from the fitted line and r. WHO publishes prices every second year, so odd years are linearly interpolated inside the published range and never extrapolated.
Module 4: the market identity
Tax-paid cigarettes against what Ugandan smokers consume
Consumption equals the WHO adult smoking prevalence times Uganda's adult population times sticks smoked per smoker per day. Whatever consumption the stamped supply does not cover is the illicit residual. Daily consumption per smoker is not published in a source this page could confirm, so you set it; the tooltip shows the break-even level at which stamps account for everything.
Article 3(2) of the draft UN Tax Convention
What gets counted under each reading
One calendar year, from the raw gaps both customs administrations reported to the amount that would be measured as tax-related illicit financial flows under the Article 3(2) reading chosen above. Switch the reading and the share judged primarily smuggling to see what the drafting choice is worth.
Misinvoicing (always counted)
$59M
both sides recorded the flow at different values
Smuggling components
$185k
one-sided flows plus the smuggled market
Excluded by this reading
$185k
0.3% of the measured total
Module 5: scoring choices
The risk matrix behind the map
Moving the cut-offs shows the trade-off the course calls inclusion versus exclusion error: lower cut-offs flag more partners for scrutiny, higher ones risk missing a real problem.
Triangulation with URA enforcement data
How much of the inward gap enforcement recovers
URA's outright smuggling recoveries (URA Data Book, Enforcement sheet) as a share of the measured inward gap for the selected goods, converted to shillings at the World Bank average exchange rate. Fiscal years are matched to the average of the two calendar years they span. Recoveries cover all goods, not only the ones selected, so this is an upper bound on the share for the selection.
Article 11(1)(b): exchange of information
How much of the gap could be checked against fiscal marks today
Article 11 would let administrations exchange information on goods crossing borders. That is only useful where both sides' marks carry data that can be matched. This splits the measured gap by whether WHO records the partner's cigarette marks as carrying track-and-trace information. Uganda's do.
- Partner marks carry track-and-trace data: $281M (39%)
- WHO has no answer for the partner: $422M (59%)
- Partner marks carry no track-and-trace data: $11M (2%)
Every comparison behind the charts
Pair-level PCM+ results
| Partner | Good | Year | Direction | Uganda (adj.) | Partner (adj.) | Remaining gap | Ratio | Channel |
|---|---|---|---|---|---|---|---|---|
| DR Congo | Cement | 2021 | Out of Uganda | $50M | $26M | -$23M | -0.61 | Both, different values |
| Kenya | Sugar | 2025 | Out of Uganda | $121M | $98M | -$23M | -0.21 | Both, different values |
| Kenya | Cement | 2015 | Into Uganda | $36M | $57M | $21M | 0.46 | Both, different values |
| DR Congo | Sugar | 2022 | Out of Uganda | $21M | $572k | -$20M | -1.89 | Both, different values |
| DR Congo | Beer | 2024 | Out of Uganda | $20M | $221k | -$19M | -1.96 | Both, different values |
| DR Congo | Cement | 2024 | Out of Uganda | $41M | $23M | -$18M | -0.57 | Both, different values |
| Kenya | Cement | 2025 | Into Uganda | $100M | $118M | $18M | 0.16 | Both, different values |
| DR Congo | Sugar | 2024 | Out of Uganda | $18M | $334k | -$18M | -1.93 | Both, different values |
| DR Congo | Beer | 2023 | Out of Uganda | $17M | $282k | -$17M | -1.93 | Both, different values |
| Kenya | Sugar | 2015 | Out of Uganda | $34k | $16M | $16M | 1.99 | Both, different values |
| DR Congo | Cement | 2022 | Out of Uganda | $41M | $26M | -$15M | -0.45 | Both, different values |
| DR Congo | Beer | 2021 | Out of Uganda | $16M | $477k | -$15M | -1.88 | Both, different values |
| DR Congo | Spirits | 2024 | Out of Uganda | $15M | $7.5k | -$15M | -2.00 | Both, different values |
| DR Congo | Sugar | 2018 | Out of Uganda | $15M | $335k | -$15M | -1.91 | Both, different values |
| Kenya | Sugar | 2021 | Out of Uganda | $29M | $43M | $14M | 0.40 | Both, different values |
| DR Congo | Beer | 2022 | Out of Uganda | $14M | $255k | -$14M | -1.93 | Both, different values |
| DR Congo | Cement | 2020 | Out of Uganda | $42M | $28M | -$14M | -0.39 | Both, different values |
| Kenya | Sugar | 2022 | Out of Uganda | $43M | $30M | -$13M | -0.35 | Both, different values |
| DR Congo | Beer | 2020 | Out of Uganda | $12M | $145k | -$12M | -1.95 | Both, different values |
| DR Congo | Sugar | 2021 | Out of Uganda | $12M | $377k | -$12M | -1.88 | Both, different values |
| DR Congo | Spirits | 2023 | Out of Uganda | $12M | $194k | -$12M | -1.94 | Both, different values |
| DR Congo | Soft drinks | 2024 | Out of Uganda | $12M | $245k | -$12M | -1.92 | Both, different values |
| DR Congo | Sugar | 2019 | Out of Uganda | $12M | $759k | -$11M | -1.77 | Both, different values |
| DR Congo | Sugar | 2020 | Out of Uganda | $12M | $1.0M | -$11M | -1.69 | Both, different values |
| DR Congo | Beer | 2019 | Out of Uganda | $11M | $12k | -$11M | -2.00 | Both, different values |
Showing the top 25 of 660. The download has every row with every intermediate step.
Rules, not judgement
Data-quality flags
Every series this page leans on was checked by fixed rules: a year-on-year move of ten times or more (a unit or coverage change), two URA sheets disagreeing by more than half on the same quantity, values reported before the system that should have produced them existed, and Comtrade years a partner did not report.
| Series | Rule | Severity | Detail |
|---|---|---|---|
| DTS cigarette delivery quantity | 10x year-on-year move | high | 2023/24 to 2024/25: x647.4 jump |
| DTS locally produced construction materials quantity | 10x year-on-year move | high | 2019/20 to 2020/21: x57.4 jump |
| DTS locally produced construction materials quantity | 10x year-on-year move | high | 2020/21 to 2021/22: x20.2 drop |
| Sugar production (tonnes), Excise sheet | 10x year-on-year move | high | 2015/16 to 2016/17: x494.2 jump |
| Cigarette excise: DTS sheet vs Excise sheet | Sheets disagree | medium | 2015/16: 44.6 vs 12.6 (72% apart) |
| Cigarette excise: DTS sheet vs Excise sheet | Sheets disagree | medium | 2016/17: 74.2 vs 12.1 (84% apart) |
| Cigarette excise: DTS sheet vs Excise sheet | Sheets disagree | medium | 2022/23: 8.4 vs 19.5 (57% apart) |
| DTS cigarette delivery quantity | Predates the system | medium | 4 fiscal years (2015/16 to 2018/19) are reported before DTS went live on 1 November 2019, so these values come from an earlier system and are not stamp counts. |
| UN Comtrade, Burundi as reporter | Missing reports | medium | No report for 2024, 2025; comparisons with Burundi exist only for 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023. |
| UN Comtrade, South Sudan as reporter | Missing reports | high | No report for 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024; comparisons with South Sudan exist only for 2025. |
| UN Comtrade, DR Congo as reporter | Missing reports | medium | No report for 2025; comparisons with DR Congo exist only for 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024. |
| UN Comtrade, Somalia as reporter | Missing reports | high | Somalia filed no report on these goods for any year 2015 to 2025, so no comparison is possible. |
Reading this carefully
A mirror asymmetry is a signal, not proof. Timing differences, misclassification between HS codes, goods in transit, re-exports, confidential trade and plain recording error all produce gaps without any illicit intent, which is why the UNCTAD-UNODC framework treats these estimates as indicators that point an audit at a flow, not as findings against anyone. The paper's aggregate gap is therefore a ceiling, and its medians are not adjusted for product mix, partner or the exact phase date.
The model's labels "inward" and "outward" describe the direction of money under the statistical convention, not where tax was lost. 63% of its inward total on stamped goods comes from goods Uganda records as exported to DR Congo that DR Congo records at a lower value or not at all. The plainer reading is that the goods entered DR Congo without being fully declared, so the revenue lost is DR Congo's. Uganda holds the stamp and export records that DR Congo would need: the case for Article 11 in its simplest form.
Decisions made, and why
- CIF/FOB margin of 10% in both the paper and the model, the IMF Direction of Trade Statistics convention, with 5% and 15% as bounds in the model. Short overland routes in the region plausibly sit nearer the lower bound.
- Before and after. The paper splits at 2020, the first calendar year under phase 1, and checks the split by aligning sugar and cement to 2021. Comtrade is annual, so no finer timing is possible from trade data.
- Medians, not means. A few hundred dollars recorded on one side can produce a log ratio that swamps a twenty million dollar flow; the median is robust to that. The model uses value weighting for the same reason, with equal weights as a check.
- Comparators. The paper uses rice and steel bars and rods, which are never stamped; the model uses soap and bakery products. Neither set is large enough to identify an effect, and neither page claims one.
- Only released aggregates. The URA Research Lab figures are category-month aggregates released after disclosure review; cells with fewer than ten firms are suppressed, which is why cigarettes and cement show no post-2024 share.
Limitations
- Mirror data see only goods that at least one customs administration recorded. Goods that crossed with no record on either side appear only through the model's market identity, and only for cigarettes.
- South Sudan and DR Congo report little to Comtrade, so part of the aggregate gap is missing reports, not missing goods.
- The return extract starts carrying stamp identifiers in November 2024, so it cannot evaluate the 2019-2022 phases; that needs the earlier stamp-activation data linked to returns.
- HS heading 2202 holds both phase 1 soft drinks and phase 3 non-alcoholic beverages, so phase 3 goods cannot be separated at four digits.
- The WHO fiscal-marking answers come from a single questionnaire round (2022) and describe cigarettes only.
- The treaty text is under negotiation and may change.
Sources
- Kaijamahe, W.H. (2026) Can the Stamp Cross the Border? Excise Smuggling, Fiscal-Marking Data and Articles 3(2) and 11 of the Draft UN Tax Convention, ICTD workshop draft, November. Not for citation or circulation without the author's permission.
- Domestic excise: URA Annual Data Book FY2024/25, Table 1.1.1, p. 14.
- UN Comtrade Database, HS 2402, 2203, 2208, 2204, 2201, 2202, 1701, 2523, 1006 and 7213, 2015-2025, downloaded 13 October 2026.
- URA excise return extract RESEARCHLAB_LED_RETURN, URA Research Lab, October 2026; aggregates released after disclosure review and published with URA's permission.
- United Nations (2026) Co-Lead's Zero Draft of the UN Framework Convention on International Tax Cooperation, A/AC.298/CRP.32, 21 July, and draft protocols CRP.33 and CRP.34, Intergovernmental Negotiating Committee on the United Nations Framework Convention on International Tax Cooperation.
- DTS phase dates: Uganda Gazette No. 53, 14 October 2019 (phase 1); allAfrica, 20 April 2021 (phase 2); URA, The Taxman, 2022 (phase 3). Web excise return: URA public notice, 2024.
- Model: UN Comtrade (comtradeapi.un.org, /data/v1/get/C/A/HS), annual, HS as reported, totals only (partner2Code=0, motCode=0, customsCode=C00).; WHO Global Health Observatory OData API (ghoapi.azureedge.net): R_total_tax_estimate, R_Price_mp_estimate_usd, R_admin_tax_stamps_cig, R_admin_unique_identifier, R_admin_identifier_track_and_trace, M_Est_cig_curr.; World Bank WDI API: SP.POP.1564.TO + SP.POP.65UP.TO (population aged 15+), PA.NUS.FCRF (official exchange rate, UGX per USD, period average).; URA Data Book: Digital Tax Stamps (DTS system export, extracted 27 July 2026).
- UNCTAD and UNODC (2020), Conceptual Framework for the Statistical Measurement of Illicit Financial Flows; UNCTAD, PCM+ methodology for trade misinvoicing.
- Engines and tests: lib/stampBorder.ts with lib/stampBorder.test.ts (the paper), lib/exciseIff.ts with lib/exciseIff.test.ts (the model).