Afrimintel · Countries · Screening · Methodology · Audit Log
Mirror-trade discrepancies on a jurisdiction's mineral exports (UN Comtrade) read as a fiscal-governance risk indicator — a prompt for enhanced scrutiny, not a measured revenue loss and not proven illicit flows. This surface feeds the country risk layer as separate context; it is not a component of the published country composite. Status: method + worked jurisdictions Zambia (copper) and South Africa (gold / iron ore) on historical UN Comtrade (2000–2018), with continent coverage mapped; live, current-year monitoring is a pending ingestion step.
Read this first. A mirror-trade gap is an indicator, not a measurement. Large gaps are routinely produced by ordinary, legal trade — freight/insurance (CIF vs FOB), merchanting and transit through hubs (the "Switzerland effect" / "Rotterdam effect"), bonded warehouses, and commodity classification. Afrimintel surfaces the gap and the legitimate explanations, and asserts no headline "loss" figure. Treating a gap as proven misinvoicing is the error this surface is built to avoid.
Definition. For a jurisdiction's primary mineral exports (selected HS codes — e.g. refined copper HS 7403), the mirror gap compares what the reporter country records as exports against what its partner countries record as imports from it, with a standard ~10% CIF/FOB allowance applied to the import side. A persistent asymmetry is the indicator.
What it can flag. Direction and persistence of a gap can prompt questions on fiscal-governance, transit-trade transparency, and export-valuation oversight — inputs a DFI or fiscal-risk analyst would want surfaced before underwriting.
What it is NOT. It is not a customs-fraud finding, not a quantified loss, and not asset-level by default. Asset/operator attribution is attempted only where defensibly traceable and is explicitly flagged when it is.
| Cause | Effect on the gap |
|---|---|
| CIF/FOB asymmetry | Partner imports include freight + insurance the exporter's FOB value excludes; ~10% allowance is an assumption, and real freight varies by route and commodity. |
| Merchanting & transit ("Switzerland effect" / "Rotterdam effect") | A trader's hub is recorded as destination though goods never physically arrive there; creates a negative gap to the hub and an offsetting positive gap to the final buyer — the pair nets out. |
| Bonded warehouses / exchange networks (e.g. LME) | Storage and re-sale in transit break the origin→destination chain customs statistics assume. |
| Special vs general trade systems | SEZs, bonded zones, free zones record some flows differently between partners, producing gaps unrelated to misinvoicing. |
| Classification & price volatility | The same good coded differently on each side; volatile commodity prices and in-transit pricing inflate apparent gaps. |
| Low-transparency trading hubs (e.g. UAE) | Flows routed through low/nil-tax SEZ hubs are hard to attribute; large gaps to such hubs warrant extra caution and are not, by themselves, evidence of an origin-country flow. |
| Non-producing re-export hubs | Hubs with little or no domestic production (e.g. some West African transit points for gold) can show very large export gaps that are re-export artifacts, not producer underinvoicing. |
Indicator: Derived Underlying figures: Sourced Live current-year snapshot: Absent
The gap. Zambia's data has recorded Switzerland as the leading destination for its copper (~51% in the UNCTAD analysis; ~40–50% in every year 2006–2017 in the academic record), while Switzerland records almost no corresponding copper imports from Zambia. One peer-reviewed audit put recorded Swiss copper imports near CHF 4.2m/yr against ~CHF 2.3bn/yr of refined-copper exports booked to Switzerland over 2011–17 — a roughly 500-fold asymmetry (this is the aggregate "Switzerland effect" across all origins, evidencing the mechanism behind Zambia's gap, not a Zambia-only figure).
The contested headline. A naïve reading of the mirror gap yields a large "underinvoicing" total — Afrimintel surfaces it only to contest it. UNCTAD (2016) computed a cumulative Zambian copper export-side mirror gap of roughly US$12bn against major partners (about US$14.5bn against the rest of the world) over 1995–2014, excluding Switzerland. Afrimintel does not assert this as a loss. Subsequent peer-reviewed analysis attributes a large share to Swiss merchanting and CIF/FOB rather than illicit flows: Swiss-based trading firms buy Zambian copper and on-sell it in transit to China, Italy, Korea and elsewhere, so the offsetting hub/final-buyer gaps cancel. The cited studies identify Swiss-based trading firms as the dominant intermediaries over their 1995–2017 window; the indicator is jurisdiction-level and does not turn on any single named firm. Period-bounded public-record context, not a current-operator or illicit-flow claim. The gap also moves inversely with the copper price, consistent with valuation-timing rather than fraud.
Indicator read: elevated, persistent mirror gap — predominantly merchanting/transit-explained. Surface it as a prompt for fiscal-governance and export-valuation transparency, not as a loss estimate. Vintage: cited gap analyses run on 1995–2017 UN Comtrade; a current-year snapshot is Absent pending a dated Comtrade pull.
Indicator: Derived Underlying figures: Sourced Live current-year snapshot: Absent
The contested headline. A naïve reading produced one of the most-cited African IFF figures — Afrimintel surfaces it only to contest it. UNCTAD (2016) put South African gold export underinvoicing at US$78.2bn, ~67% of total gold exports, over 2000–2014 (top partner gaps: India ~$40bn, Germany ~$18.4bn, Italy ~$15.5bn, UK ~$13.7bn), plus ~US$3bn on iron ore to China. Afrimintel does not assert these as losses.
Why the gap is largely an artifact. South Africa does not report gold export destinations (a hangover from gold's former monetary classification), and reclassified gold from monetary to non-monetary in 2011 — so partners record South-Africa-origin gold that South Africa's own series does not match by destination. A Eunomix review commissioned by the Chamber of Mines (an interested party) found that mining companies and agencies do report gold exports, just not in COMTRADE's format, with roughly three-quarters of the discrepancy explained from official statistics; the platinum estimate rests largely on two years (2000, 2002) missing from COMTRADE entirely. UNCTAD itself later stated that, after South Africa changed its export statistics and methods, the gold discrepancy can no longer be calculated with certainty. The CIF/FOB proxy is 10%; South Africa's own published ratio averaged ~11% over 1980–2014.
A separate, differently-evidenced concern — do not conflate. Distinct from the COMTRADE mirror gap, a separately documented artisanal/illegal-gold concern is evidenced through other channels (SARS enforcement cases, the national ASM policy paper's ~R70bn/yr estimate, ~70–80 t/yr moving outside official Rand Refinery / Reserve Bank channels; UN Comtrade shows the UAE receiving ~41 t from South Africa in 2021 though no large-scale miner exports there). A 2025 audit also notes refined-and-re-exported gold (Rand Refinery's total exports exceeded SARS's reported figures by >2× in some years), which complicates partner mirrors further. The mirror gap does not measure this flow — real illicit flows require enforcement-grade evidence, not a Comtrade asymmetry.
Indicator read: the headline mirror gap is predominantly a gold-reporting and refining/re-export artifact (UNCTAD walked back its own number); a separate artisanal-gold concern is real but differently sourced. Surface both, conflate neither, assert no loss. Vintage: 2000–2018 (UNCTAD/EDAR); re-export nuance to 2025.
Every jurisdiction Afrimintel covers, each carrying its three-state trade-gap status. Magnitudes are third-party estimates (UNCTAD 2016/2020; SWISSAID 2024/2025), attributed — not Afrimintel loss claims. SWISSAID frames undeclared/smuggled gold as revenue lost; Afrimintel surfaces the gap as a risk indicator and asserts no loss figure. Most jurisdictions are honestly Absent: no published mirror-gap study exists, and none is invented to fill a row. Gap-type is flagged so a producer concern is never confused with a merchanting, reporting or re-export artifact.
| Jurisdiction | Commodity | Gap type (per cited literature) | Estimate (attributed) | Status |
|---|---|---|---|---|
| Worked cases — Sourced magnitude + read | ||||
| Zambia | Copper | Merchanting artifact (Switzerland) | Mirror trade gap (Comtrade) — UNCTAD 2016: ~$12–14.5bn 1995–2014, contested | Sourced |
| South Africa | Gold / iron ore / platinum | Reporting artifact + refining / laundering hub (+ separate artisanal concern) | Mirror trade gap (Comtrade) — UNCTAD: $78.2bn gold 2000–14, walked back; SWISSAID: 139–143 t imported in 2022/23 (hub) | Sourced (contested) |
| Ghana | Gold | ASM-smuggling (per SWISSAID / OECD) | Mirror trade gap (Comtrade) — SWISSAID: ~229 t / ~$11.4bn 2019–23 | Sourced |
| Sourced ASM-smuggling estimate (SWISSAID) — attributed, not a loss claim | ||||
| Mali | Gold | ASM-smuggling concern (also a rerouting destination) | Undeclared ASM-production estimate (modeled) — SWISSAID: 30–57 t/yr (range) | Sourced (range) |
| Burkina Faso | Gold | ASM-smuggling concern | Undeclared ASM-production estimate (modeled) — SWISSAID: ~57.7 t | Sourced |
| Zimbabwe | Gold (+ PGM / diamond) | ASM-smuggling concern (largely via the SA hub) | Smuggling estimate (modeled) — SWISSAID: ~60 t 2022 | Sourced |
| Guinea | Gold (+ bauxite) | ASM-smuggling concern | Undeclared ASM-production estimate (modeled) — SWISSAID: 16–32 t/yr (range) | Sourced (range) |
| Sudan | Gold | Active-conflict caution — data shown, no Afrimintel interpretation | Qualitative — SWISSAID / media: UAE imports from Sudan ~+70% in 2024 | Sourced (qualitative) |
| SWISSAID African Gold Report coverage / ranking-named — per-country magnitude pending | ||||
| Tanzania | Gold | Producer; also feeds the SA refining hub | Named in SWISSAID continental data; no extractable per-country magnitude yet | Absent (named) |
| Niger | Gold | Producer / transit | In UNCTAD 2020 gold-gap ranking; magnitude pending | Absent (ranking-named) |
| Mauritania | Gold | Producer | SWISSAID country analysis + in UNCTAD ranking; magnitude pending | Absent (analysis exists) |
| Senegal · Liberia · Sierra Leone · Côte d’Ivoire · Kenya · Chad | Gold | Producer / transit (mixed) | SWISSAID African Gold Report analysis available; magnitude pending | Absent (analysis exists) |
| Nigeria | Gold | ASM concern (gold) | SWISSAID gold coverage; mining-gold magnitude pending | Absent (analysis exists) |
| Botswana | Diamonds (minor gold) | Minor / transit — likely artifact | SWISSAID coverage; little ASM gold | Absent (analysis exists) |
| Namibia | Gold / diamonds / uranium | Refining-feed artifact (SACU inward processing to SA — legitimate) | SWISSAID coverage; flagged as feed, not leakage | Absent (analysis exists) |
| Independence firewall — data-layer only, no editorial interpretation (see note below) | ||||
| DRC · Rwanda · Burundi | — | Under independence firewall | — | Carried data-layer only; no editorial interpretation |
| Uganda | Gold (re-export) | Re-export conduit for the firewalled complex | — | Any read would be a read on firewalled flows; data-layer only |
| Absent — no published mineral mirror-gap study; method ready to apply | ||||
| Angola · Gabon | Diamonds / gold | — | SWISSAID sought disaggregated data; none published | Absent |
| Egypt · Eritrea · Ethiopia | Gold (largely LSM) | — | No mirror-gap study in hand | Absent |
| Morocco · Algeria · Libya | Phosphate / hydrocarbons | — | Not a gold-smuggling subject; no study | Absent |
| Mozambique · Madagascar · Malawi · Cameroon | Graphite / gems / mixed | — | No mirror-gap study in hand | Absent |
| Republic of Congo · CAR · Djibouti · Somalia | Mixed / minimal | — | No mirror-gap study in hand | Absent |
Reading the matrix. Sourced means a published study reports a magnitude — it does not mean Afrimintel asserts a loss. Gap-type is what the cited literature attributes the gap to, not an Afrimintel verdict. The magnitudes are also different constructs and are not directly comparable: a mirror trade gap (Comtrade exports vs partner imports, e.g. Ghana / Zambia / South Africa) measures a reporting discrepancy, whereas an undeclared ASM-production estimate (modeled, e.g. Mali / Burkina Faso / Guinea) is the source’s estimate of unrecorded output — a softer, model-based figure. Ranges (Mali, Guinea) reflect the source’s own uncertainty. Absent “(analysis exists)” means SWISSAID has a country page but no single magnitude is safely extractable yet; bare Absent means no study is in hand. ASM-smuggling estimates double-count where gold is rerouted across borders (e.g. Ghana→Togo/Burkina/Mali), so continental and per-country figures cannot simply be summed.
On the independence firewall. Where Afrimintel’s editorial leadership has any affiliation or appearance of interest touching a jurisdiction, that jurisdiction is carried data-layer only, with no editorial interpretation — independence is protected by declining to opine, not by opining carefully. The Great Lakes gold complex (DRC, Rwanda, Burundi, and Uganda as a regional re-export conduit for that complex) is held under this firewall. This is a disclosed editorial limit, not an absence of data, and it is deliberately distinct from the active-conflict caution applied to Sudan, where there is no such affiliation but the wartime context means data is shown without Afrimintel interpretation. The firewall removes the highest-magnitude cases from editorial reach by design; it is not a selection of which gaps to report. This firewall is the governing editorial disclosure documented in the Independence Policy.
Continental context (SWISSAID 2024, attributed). SWISSAID estimates 321–474 t/yr of artisanal gold produced in Africa without being declared (~$24–35bn), with ~435 t smuggled out in 2022 (~$30.7bn) and ~2,596 t entering the UAE from Africa over 2012–2022 without matching export declarations (~$115bn). Top destinations: UAE, Switzerland, India. These are SWISSAID’s own continental estimate — not a sum of the matrix rows above — surfaced as indicator context — Afrimintel does not adopt them as measured losses. The earlier UNCTAD EDAR 2020 picture (rest-of-world gold imports ~2× the continent’s reported exports 2011–2018; South Africa ~67% of the absolute gold gap; gap tracks the gold price) remains the structural backdrop and is heavily shaped by destination-reporting, rules-of-origin and refining, not a single “loss.”
UNCTAD, Trade Misinvoicing in Primary Commodities in Developing Countries (2016). UNCTAD EDAR 2020 background paper on mirror-trade analysis. Mineral Economics / Springer, "Misinvoicing in mineral trade: what do we really know?" (2018). Center for Global Development, "Gaps in Trade Data ≠ Criminal Money Laundering" (2017). G. Dobler & R. Kesselring, "Swiss extractivism: Switzerland's role in Zambia's copper sector," Journal of Modern African Studies (2019). Curbing-IFFs / R4D, "Abnormal Pricing in International Commodity Trade" (2020). On South Africa: UNCTAD (2016) and its revised statement (2016); Eunomix review for the Chamber of Mines (via CGD, 2017); D. van Rensburg / City Press via tralac (2016); tralac, "South Africa's Intractable Illegal Gold Mining Problem" (2023); SWISSAID, African Gold Report — South Africa (2025); UNCTAD figures as reported by Mining.com (2016). On the continental / gold picture: SWISSAID, On the Trail of African Gold (M. Ummel & Y. Schulz, 2024) and the country analyses at africangoldreport.org (2024–2025); A. Ndoricimpa, “The ugly side of the Africa–UAE gold trade,” Resources Policy / ScienceDirect (2024); OECD / M. Hunter, GI-TOC (2019–2020); and contemporaneous reporting (Reuters, Ecofin Agency, TRT Afrika, swissinfo.ch, 2024–2025). SWISSAID frames undeclared / smuggled gold as revenue loss; Afrimintel attributes those figures and surfaces the gap as an indicator without adopting a loss claim. All cited per their published terms; no commercial relationship with any named body is implied.
Subordination & scope. This indicator feeds the country risk layer as qualitative fiscal-governance context. It does not alter the published country composite (Fraser / TI CPI / NRGI RGI / EITI) and applies no numeric adjustment to any score. Headline aggregate "loss" numbers are never asserted. Independence firewall: the Great Lakes complex (DRC, Rwanda, Burundi, Uganda) is carried data-layer only, no editorial interpretation (see matrix note). Worked jurisdictions: Zambia (copper), South Africa (gold / iron ore / platinum), Ghana (gold); full continent coverage mapped above. Other jurisdictions Absent pending dated Comtrade snapshots.