Quantitative Analysis · Trade Finance · Mirror Trade Discrepancy

Trade-Based Money Laundering

TBML hides inside the normal machinery of international commerce — invoice manipulation, phantom shipments, and multiple invoicing exploit the volume and opacity of trade documentation at scale. This project applies mirror trade discrepancy analysis to UN Comtrade bilateral data across SEA and South Asian corridors, mapping persistent anomalies against FATF typology indicators and MAS Notice 626.

Python · pandas UN Comtrade API seaborn · plotly FATF TBML MAS Notice 626 In Progress ◎

⬡ Python Notebook Full pipeline — API pull, discrepancy calculation, persistence flagging, corridor ranking

◈ Discrepancy Heatmap Anomaly magnitude by corridor & commodity category across the analysis period

◇ Time-Series Charts Discrepancy ratio over 5 years for the top flagged corridors — the persistence argument

Project Progress

Data Assembly

Analysis

Published

Why TBML, and why mirror trade

Trade-based money laundering is consistently identified by FATF as one of the three primary methods by which criminal organisations move and clean value globally — alongside bulk cash smuggling and the use of the formal financial system. Unlike the other two, TBML leaves its traces not in financial records but in trade documentation: shipping invoices, letters of credit, and customs declarations that move through a system handling hundreds of millions of transactions annually.

The core manipulation methods — over- and under-invoicing of goods and services, multiple invoicing for the same shipment, falsely described goods, and phantom shipments — all produce a predictable statistical signature when examined at the bilateral level. When Country A reports exporting a given value of electronics to Country B, but Country B reports importing a materially different value of electronics from Country A in the same period, that discrepancy is either a reporting artefact or a signal. At scale, across corridors and over time, the persistent cases become analytically distinguishable from the noise.

Singapore's position as a major regional trade hub makes this directly relevant. MAS Notice 626 identifies trade finance as a high-risk channel and specifies red flag indicators for TBML that mirror the analytical logic of this project. The ability to surface and interpret these discrepancy patterns is a concrete, transferable skill in compliance and transaction risk roles across the region.

Analytical question

Across a defined set of SEA and South Asian bilateral trade corridors and TBML-associated commodity categories, which country-pairs show persistent mirror trade discrepancies above a defined materiality threshold — and do those discrepancy patterns align with FATF-documented TBML risk corridors and MAS Notice 626 red flag indicators?

Persistence is the key analytical criterion. A single-year discrepancy may reflect reporting lag, classification differences, or transit trade accounting. A discrepancy that exceeds a defined threshold across three or more consecutive years — in the same corridor, in the same commodity category — is analytically significant and warrants annotation against the typology framework.

III

Scope & method

The analysis covers 8–10 bilateral trade corridors across Southeast and South Asia — geographies with documented TBML exposure and sufficient reporting volume in UN Comtrade to support reliable bilateral comparison. Commodity categories are filtered to four HS code clusters historically associated with TBML risk: electronics and electrical components, textiles and garments, precious metals and stones, and high-variability agri-commodities.

The analytical pipeline runs in four steps. Data is pulled from the UN Comtrade API using the Python wrapper, covering 2018–2023 to capture five years of bilateral trade flows. For each corridor-commodity pair, a discrepancy ratio is calculated as reported exports divided by reported imports, normalised for known structural factors (re-export volumes, CIF/FOB valuation differences). Pairs where the ratio deviates significantly from 1.0 are flagged; those exceeding the threshold in three or more consecutive years are classified as persistent anomalies. Finally, flagged pairs are annotated against FATF TBML typology indicators and MAS Notice 626 red flag language.

◈

Discrepancy Heatmap

Anomaly magnitude by corridor × commodity — 8–10 SEA/South Asian corridors, 4 HS code clusters, 2018–2023. In progress.

Data Sources

UN Comtrade API

Bilateral trade statistics reported independently by exporting and importing countries — the foundational dataset for mirror trade analysis. Accessed via the comtradeapicall Python wrapper. Free API access with registration.

FATF Trade-Based Money Laundering Guidance

Published typology indicators and risk corridor documentation — used to define which discrepancy patterns are analytically meaningful and which commodity-corridor combinations carry elevated TBML risk.

MAS Notice 626

Singapore's AML/CFT notice for banks and financial institutions, with specific provisions on trade finance risk assessment and red flag indicators for TBML — the local regulatory grounding for analytical interpretation.

Output & analytical deliverables

The project delivers three linked outputs. A Python notebook documenting the full analytical pipeline — API data pull, cleaning, discrepancy calculation, persistence flagging, and corridor ranking — with inline commentary explaining each methodological decision. A discrepancy heatmap showing anomaly magnitude by corridor and commodity category across the analysis period, making the highest-risk patterns immediately legible. And a persistent-anomaly time-series chart for the top five flagged corridors, showing discrepancy ratio over five years — the visual argument for why persistence is the analytically meaningful signal.

Each flagged corridor is annotated in a structured table against specific FATF typology indicators and MAS Notice 626 red flag provisions — so the output reads as both analytically rigorous and compliance-literate. A dedicated limitations section acknowledges structural data constraints: CIF/FOB valuation differences, re-export accounting, and reporting lag are all documented artefacts that affect the discrepancy ratio and must be controlled for before drawing analytical conclusions.

◇

Corridor Time-Series

Discrepancy ratio 2018–2023 for top 5 flagged corridors — the persistence signal visualised. In progress.

✦

Analysis in progress. Python notebook, discrepancy heatmap, and corridor time-series will be published here as outputs are completed.