Mirror Statistics

Reporting to international trade databases is uneven. Some countries report promptly and completely; others report with long lags, at coarse commodity detail, for some years and not others, or not at all. Some suppress individual cells to protect the confidentiality of a small number of traders, which removes exactly the flows a commodity analyst most wants to see.

Where the gap sits on one side only, the other side is still there. Mirroring exploits that redundancy: a flow has two potential witnesses, and one witness is better than none. The technique is standard practice in compiling international datasets, and the organisations that do it document where they have applied it.

• A non-reporting country’s trade can be reconstructed from what its partners report
• A reporter’s provisional or lagging figures can be anticipated from partner data available sooner
• A confidentiality-suppressed cell may be visible from the other side, where the same flow is not sensitive
• A reporter’s figures can be cross-checked against the aggregate of its partners as a coherence test

Bilateral asymmetry — the gap between what A says it exported to B and what B says it imported from A — is not an anomaly to be explained away. It is the expected consequence of several conventions operating at once, most of which are documented and none of which imply error.

Valuation basis

Exports are conventionally recorded at a value stopping at the exporter’s frontier; imports at a value including carriage and insurance to the importer’s frontier. An import figure is therefore expected to exceed the matching export figure, by the freight and insurance in between.

Timing

A cargo despatched near a period boundary is exported in one period and imported in the next. For long voyages the two series run out of phase.

Partner attribution

Importers commonly attribute the flow to the country of origin; exporters to the country of last known destination. Where goods move through a hub, or are re-sold in transit, the two name different partners for the same cargo.

Classification

The two administrations classify the same good independently, and may not reach the same code — especially near a processing threshold or after an edition change.

Coverage and thresholds

Trade system, statistical territory, low-value thresholds, and confidentiality suppression differ between the two reporters, so they are not counting the same universe.

Quantity units

Where the two sides record different supplementary units, or one records net and the other gross, quantity comparisons carry their own gap independent of value.

Bilateral gaps are sometimes read as evidence of under-invoicing, smuggling, or fraud. The inference is not available from the gap alone. Every mechanism in the previous section produces asymmetry without anyone having done anything wrong, and they operate simultaneously and in different directions.

Serious work on trade misreporting exists, and it proceeds by controlling for the known conventions first and examining what remains, with knowledge of the specific countries, commodities, and periods involved. A raw difference between two published numbers is the starting point of that analysis, not its conclusion. AgricultureID describes the mechanism and does not characterise any country’s reporting.

A mirror estimate is fit for some purposes and not others. It can establish that a trade exists, indicate its rough scale, show a direction of change, and identify partners. It is weaker for anything that depends on the precise level — a market-share calculation, a balance-sheet line, or a comparison against a directly reported figure on a different basis.

The presentational discipline follows from that. A mirrored figure is labelled as mirrored, its source side is stated, its valuation basis is stated, and it is not silently mixed into a series of directly reported values. Where an international dataset has mirrored a cell, that fact is in its documentation, and a user who does not read the documentation will not be able to tell from the number.

For agricultural commodities the technique carries an extra caution. Mirroring tells you what was declared as crossing a border; it does not tell you what was grown, consumed, or held. A country whose imports are visible only through partner records may also have production and stock figures of uneven quality, and combining an estimated trade figure with an uncertain production figure compounds the uncertainty rather than resolving it.