The re-export puzzle: how the OECD addresses the largest source of distortion in merchandise trade statistics

International trade statistics offer a picture of trade flows between countries and play a critical role in shaping economic policy, guiding trade negotiations, and supporting evidence-based research and decision-making. These statistics are reported from two different perspectives: that of the exporter and that of the importer. Users expect the two numbers to match but this is not always the case. When the two numbers do not align, statisticians refer to the discrepancies as ‘trade asymmetries’.

Trade asymmetries, whether in goods or services, can obscure policy decisions, confuse businesses, and complicate research efforts aimed at measuring and understanding the rapidly evolving global economy. Persistent asymmetries may even undermine trust in official statistics. Addressing them is essential for achieving a coherent and accurate view of international trade.

In merchandise trade statistics – which record all goods crossing an international border – asymmetries between reported and “mirror” import flows can arise for several reasons. First, exports and imports are valued differently: exports are reported on a “free on board” (FOB) basis, whereas imports are recorded with additional costs on a “cost of insurance and freight” (CIF) basis, increasing their value. Other factors include differences in customs regimes, confidentiality policies, time of recording, and product code misclassifications.

One of the most important, and least understood, sources of asymmetry is the presence of re-exports in trade statistics. These are goods imported into a country and then exported again without any significant processing. Unlike transshipments, which either do not enter the customs territory or do so only temporarily and are therefore excluded from trade statistics, re-exports are formally recorded and included in total exports.

To illustrate the magnitude of imbalances caused by re-exports, consider semiconductor exports from the United States to Mexico. According to official national statistics, the gap between the value of semiconductors exported by the United States to Mexico and the value imported by Mexico from the United States amounted to approximately USD 10.2 billion in 2024 (Figure 1). Interpreting such a large discrepancy requires some additional context.

The International Merchandise Trade Statistics Manual (IMTS 2010) provides the basic framework and guidance for compiling merchandise trade statistics. Regarding partner attribution, the IMTS recommends recording the country of origin for imports and the country of last known destination for exports.

While making sense in theory, this approach creates asymmetries in practice. Let’s return to the example of semiconductors above. The semiconductors originally produced elsewhere in the world, are imported first by the United States, and are subsequently re-exported to Mexico (Figure 1). Ideally, the imports of semiconductors recorded by Mexico (from the USA) should match the exports recorded by the United States (from Mexico). However, following IMTS guidelines, Mexico records the imports from the country of origin, (i.e. not the US), whereas the US reports the exports to Mexico, the last known destination, creating a statistical inconsistency.

As global production chains grow more complex and as more goods cross borders multiple times before reaching the final consumer, the IMTS attribution guidance produces asymmetries between recorded exports and corresponding mirror imports. This is especially true for major trade hubs – such as the Netherlands, Belgium, and Germany in Europe; Hong Kong and Singapore in Asia; and the United States and Panama in the Americas – which re-export a larger proportion of their imports.

For example, re-exports account for roughly 90% of Hong Kong’s total exports. In 2024, most of these re-exports originated from Mainland China, Taiwan, Korea, and Japan, and comprised mainly of electrical machinery and appliances, telecommunications equipment, and automatic data processing machines. Their main destinations included Mainland China, Unites States, Viet Nam, and India (Figure 2). A major challenge with re-exports is that they are not consistently reported. According to the UN Comtrade+ database, for the latest available years, about only 30% of countries report the value and destination of their re-exports, and around 20% also provide information on their origin. In most cases, re-exports are not reported at all.

To address the persistent asymmetries observed in international merchandise trade statistics, the OECD has developed the Balanced International Merchandise Trade Statistics (BIMTS) dataset. BIMTS provides a harmonised and internally consistent view of global merchandise trade by systematically correcting the main sources of distortion found in reported and mirror flows. One of its most important innovations is the treatment of re-exports, which play a central role in shaping bilateral trade asymmetries.

BIMTS tackles re-export related asymmetries with an innovative methodology. First, re-export flows are identified or estimated; then, they are allocated to the producer countries (rather than trade hubs). The identification process combines three complementary methods, applied according to the type and quality of information available for each re-export flow. This approach leverages recently available data fields from Comtrade+, the United Nations’ comprehensive platform for international trade statistics, along with supplementary information provided by national statistical offices.

Accurately adjusting for re-exports requires detailed knowledge of the value, origin, and destination of each shipment. However, countries differ significantly in the granularity of their reporting. In some cases, all three elements – value, origin and destination – are available, allowing for precise corrections. In others, only the value and destination are reported, with the origin missing. Sometimes, no explicit re-export data exists at all, but patterns in trade statistics suggest triangular trade activity, where goods pass through an intermediary country before reaching their destination.

To address this variability, BIMTS uses a tiered correction approach that adjusts trade flows according to the completeness of the available information. When full information is available, re-export flows can be reassigned accurately to their country of origin. Where only partial data exists – typically the value and destination, but not the origin – the most likely source country is inferred from observed trade patterns. Similarly, when re-exports are not explicitly reported but ‘symmetric asymmetries’ suggest the presence of triangular trade – particularly in the case of major transit hubs such as Singapore or the Netherlands – BIMTS infers both the value and the country of origin of re-exports based on observed patterns in the data. In these situations, targeted adjustments are applied to ensure that the corrections remain both accurate and relevant. Through this process, re-export flows are effectively reallocated to the producing countries, offering a clearer picture of the true origin of value added (Figure 3).  As data availability improves, a growing share of re-export flows can be identified and corrected. Identified re-exports accounted for around 1.5% of total world merchandise exports in 1995, rising to 4.0% in 2024, equivalent to approximately USD 905 billion (Figure 4).

After all corrections have been implemented, BIMTS applies a final adjustment that balances the mirror flows into coherent figures.

BIMTS provides two balanced measures. The first is the total balanced value, which reconciles the exports and mirror imports and includes both domestic exports and re-exports. However, as discussed earlier, the presence of re-exports leads to double counting, since an export may be recorded once by the producer country and again by the re-export hub. To account for this, BIMTS also provides a balanced value adjusted for re-exports, which removes the re-export component from the flow between a producer and the trading hub. The balanced value adjust for re-exports offers a cleaner, origin-based measure that is particularly useful for avoiding double counting and for analysing global value chains.

Returning to the semiconductor example between the United States and Mexico, Figure 5 illustrates the impact of these adjustments. The dotted line shows the corrected exports, with re-exports reallocated to their most likely producer countries. The orange line depicts the balanced value adjusted for re-exports, which arguably offers a more accurate representation of the underlying economic reality than either country’s raw statistics.

The novel treatment of re-exports is one of BIMTS’s most significant advances, substantially reducing the observed asymmetries in merchandise trade statistics. It should be noted, however, that the ability to track and account for re-export transactions crucially depends on data availability. To continue improving trade statistics, national statistical offices are encouraged to report re-export flows in as much detail as possible, disaggregated by destination, by product, and ideally by origin.

The Balanced International Merchandise Trade Statistics (BIMTS) dataset – recently released with data covering up to 2024 – provides a complete, internally consistent, and globally balanced merchandise trade matrix. Covering roughly 200 reporting economies and their trading partners, it spans over 5,500 products and is published in both the Harmonised System (HS 2017) and the Statistical Classification of Products by Activity (CPA v2) formats.

Besides the re-exports adjustment described in this article, BIMTS addresses many more of the quality issues found in international trade statistics in a transparent, systematic manner. It adjusts import values from CIF to FOB to ensure comparability with export data and addresses product misclassifications. BIMTS also reallocates unspecified or misattributed trade flows, such as those labelled under “unspecified product” or “unspecified partner”, to their most likely product categories or partner. Finally, it offers users consistent time series for the period 1995-2024.

BIMTS is an analytical construct built on top of official national statistics: it aims to enhance, not replace, the figures reporter at national level. As such, its balanced values may diverge from nationally published estimates, but they offer researchers and policymakers a unified, comparable lens through which to study global trade dynamics and the structure of international value chains.