Trekking the Gold Trail: Misinvoicing in Primary Commodity Exports

Léonce Ndikumana


Exports of primary commodities are an important driver of growth in many developing countries. However, high resource endowment exposes these countries to the vagaries of large swings in commodity prices. And if “natural capital accounting” is applied so as to count the full cost of non-renewable resource depletion, the World Bank finds that 88 percent of African countries are net losers: the incoming profits and investment are less than the outgoing value of the minerals (World Bank, 2014). Moreover, resource-rich countries suffer from losses in foreign exchange and tax revenues resulting from under-reporting of export proceeds. Indeed, trade misinvoicing is a major concern in a global system characterized by lack of transparency and skewed distribution of gains from trade. Incentives for trade misinvoicing arise from the existence of opportunities for profit maximization and access to foreign exchange out of the control of the regulating authority. These opportunities are made possible by regulation of tariffs, customs procedures, export subsidies, and exchange controls, among others.

The issue of trade misinvoicing has been a long-standing concern in the economics profession since the seminal work by Jagdish Bhagwati in the 1960s.[1] The work was inspired by an even older strand of literature concerned with the consistency of partner data on international trade back in the 19th century.[2] Interest in the problem of trade misinvoicing gained momentum in the 1980s in the context of the debt crisis;[3] since then it has taken prominence in academia and in the policy arena.[4]

Trade misinvoicing is defined as either perverse discrepancies or excessive normal discrepancies in partner trade statistics estimated through the comparison of exports as reported by the exporter and imports as reported by the importer. There are perverse discrepancies when recorded imports are significantly lower than recorded exports plus the cost of transport, insurance, and duties. This could reflect either export overinvoicing or import underinvoicing.  Trade misinvoicing may be driven by factors associated with the country of origin or the destination of trade flows or both. Estimates of trade misinvoicing do not therefore permit to assign a priori the respective share of responsibility on the basis of the estimates of trade misinvoicing alone.

A recent report published by the United Nations Conference on Trade and Development (UNCTAD) (prepared by the author of this note) examines the extent of trade misinvoicing of primary commodity exports from five resource-rich developing countries—Chile, Côte d’Ivoire, Nigeria, South Africa, and Zambia (UNCTAD, 2016). The publication of the first version of the report in July 2016 generated interesting debates, partly motivated by the sheer values of export misinvoicing but also by what appears to be misconceptions regarding the nature of the data, the methodology and concepts used in the analysis. Some reactions to the report warrant a clearer exposition of the concepts and methodology of estimation of trade misinvoicing—which the new version of the report sought to clarify. Most fundamentally, there is a need for further improvements in the transparency of the reporting of trade statistics. The UNCTAD report and the debates that it has generated can be considered as a boost to further investment in knowledge generation in this area. This note presents some of the updated results using the case of gold exports from South Africa as an illustration, while responding to the main criticisms to the initial report. More detail is provided in the revised version of the report just published by UNCTAD (December 2016).

Highlights from recent evidence

Large discrepancies in partner trade data

Key results emerge from the analysis of primary commodity exports from the five developing countries covered by the UNCTAD report. First, the data show large unexplained discrepancies between the values of exports as declared by the exporting countries and the values of imports as declared by the trading partners. These differences vastly exceed reasonable costs of freight and insurance, providing prima facie indication of trade misinvoicing. Table 1 summarizes the results. However, these differences may also be due to other problems associated with the reporting of trade data, notably inconsistency in the classification of products between trading partners, inaccurate reporting of origin and destination of products, and lags in recording of imports. To the extent that the discrepancies are systematic, this casts doubts on the justification based on any of these possible data problems, pointing to evidence of trade misinvoicing. Only detailed analysis by country, product, transaction, and partner pairs can shed light on the sources of the discrepancies and the share that is attributable to trade misinvoicing as opposed to other factors.


Blind spots in the trade chain

Second, the analysis of the data reveals ‘gray holes’ or blind spots in the trading chain whereby exports recorded at the origin cannot be traced at the declared destination on the exporter’s records. This phenomenon seems to be especially associated with what is referred to as commodity ‘trading hubs’, notably The Netherlands and Switzerland. For example, while Zambia’s data shows that Switzerland is the top buyer of its copper (51 percent), no copper imports from Zambia appear in Switzerland’s trade data. Similarly, a significant amount of Nigerian oil registered as exported to the Netherlands cannot be traced in the Netherlands’ bilateral trade data. Some highlights are provided in Table 2.


Transit trade has been suggested as a possible explanation for the large discrepancies between partner trade data.  The question here is why exports are recorded as destined to a country when they are not shipped to that country. If a commodity is just “transiting” in a country, it should not be recorded as an export to this country. When commodity exports cannot be tracked from the origin to their ultimate destination, when intermediaries fail to report the values of the transactions it becomes impossible to assess whether the gains from commodity trade are distributed fairly, specifically whether the producer is earning a fair share of the market value of the exported commodities.

One criticism leveled against the methodology used to estimate trade misinvoicing is that the proxy for the cost of freight and insurance used in comparing partner trade data leads to inflated estimates of trade misinvoicing. Because information on the cost of transport, insurance, and duties is not systematically reported, the practice is to use 10 percent of exports as a proxy for these costs. South Africa is one of the few African countries that publish imports in f.o.b. and c.i.f. values. Taking the ratio of these two series yields an average ratio of 11 per cent over the 1980-2014 period. The 10 per cent used as a proxy is therefore quite reasonable. Obviously, the cost of freight and insurance varies by industry and country position relative to the markets. In the case of gold, for example, the cost of transport is expected to be relatively low, so the 10% proxy could actually be too high.  Nonetheless, it is not possible to attribute the estimated trade misinvoicing to an inaccurate proxy for the cost of freight and insurance.

The case of gold exports from South Africa

Among the results from the UNCTAD report, those on gold exports from South Africa have ignited the most spirited debate, including questions about the source of the data used, challenges to the methodology used in estimating trade misinvoicing, and questions on the interpretation of the results. This note seeks to shed light on this debate and offer new insights into the analysis of South African gold trade data while raising issues that deserve attention from policymakers.

The analysis

First of all, it is important to establish clearly what is being measured and analyzed. Gold exports are reported under two categories: monetary gold and non-monetary gold. The UNCTAD report focused on non-monetary gold [SITC code 971] and compared the data reported by South Africa to the values reported by its trading partners. The analysis in the initial report published in July was based on the data reported in Comtrade, a database managed by the United Nations.

The results in the July version of the UNCTAD report showed that while partner data indicated a cumulative amount of $116 billion in non-monetary gold exports from 2000 to 2014, South African data in Comtrade showed only $34.5 billion. The South African Revenue Services contended that the true value was $54.5 billion but it did not provide details on how this amount was calculated. Investigation of published government data sources shows little difference between the values reported in Comtrade and those in government sources. This was confirmed using data from the Department of Trade and Industry (DTI). This similarity is expected given that Comtrade only records the data as supplied by government sources.

In September, the DTI data showed substantial upward revisions of the values of gold exports, raising non-monetary gold exports to $62 billion. But this still left a gap of $54 billion compared to the value reported by trading partners. The data are summarized in Table 3.


Apples + oranges = gold?

A report by consultancy firm Eunomix commissioned by the South African Chamber of Mines claims to have reconciled the data reported by South Africa and its trading partners. According to the report, gold exports over the 15 year period from 2000 to 2014 amounted to $87.9 billion.

There are two issues with the results in this consultancy report. The main issue is that it does not distinguish between monetary and non-monetary gold exports. This means that the values in the Eunomix report cannot be compared to those reported by South Africa’s trading partners. The second is that even if one accepted that apples could be compared to oranges, the comparison still leaves a substantial amount of gold unaccounted for, suggesting export misinvoicing of $19 billion.

Three key results

From the analysis of gold exports data in both the July edition of the UNCTAD report that relied solely on Comtrade and results in the revised version of the report using alternative government sources, the main conclusions remain unchanged. First, the data on non-monetary gold exports reported in Comtrade is similar to the data reported in government statistics, which is to be expected. Therefore, analysis of trade misinvoicing yields similar results regardless of which source is used.

Second, if we follow the classification of gold between monetary gold and non-monetary gold—as everyone should—then the analysis of export data shows substantial discrepancies between South Africa’s exports and the values of its trading partners’ imports of non-monetary gold.

Third, curiously, new government data series for monetary gold and non-monetary gold are merged starting from 2011. This seems to be a step back with regard to consistency and transparency in trade statistics. Such a move makes the analysis of trade misinvoicing impossible as it only considers non-monetary gold which is reported by trading partners.

Conflations and misinterpretations

The ongoing debate on trade misinvoicing in general and, in particular, some of the reactions to the UNCTAD report have revealed substantial confusion in the interpretation of the results and key concepts used in the analysis. First, partly because of inadequate explanation in the original UNCTAD report, the concept of trade misinvoicing clearly remains elusive for some uninformed readers and it gets conflated with other related but different concepts.

The concept of trade misinvoicing tends to be conflated with that of transfer pricing. Yet, the two phenomena are different. Transfer pricing generates no discrepancies between recorded exports and recorded imports for the simple reason that the same price is used and reported on both sides of the transaction. Firms resort to abusive transfer pricing by inflating prices so as to shift profits across territories and take advantage of differences in taxation regimes. Transfer pricing may indeed constitute an illicit financial flow; but it is not a mechanism for capital flight given that the outflow is recorded. Moreover, while transfer pricing results in tax revenue losses (e.g., as in the case of recorded profit repatriation), the associated transfer of profit itself does not constitute capital flight. It is difficult to ascertain the legality of transfer pricing because of the lack of consistent benchmarks for market prices especially for trade in services. The literature on capital flight has primarily been concerned with the fact that the flows are unrecorded, not with the legality of these flows. In that sense, the work on illicit financial flows is a welcome expansion of the analysis of capital flows and a major contribution to the policy debate.

There is also frequent confusion with regard to the implications of trade misinvoicing for capital flight. The literature defines capital flight as unrecorded outflows of capital from a country to the rest of the world through various mechanisms. Trade misinvoicing constitutes one such mechanism. Trade misinvoicing generates capital flight when exports are underinvoiced and when imports are overinvoiced. In contrast, underinvoicing of imports (under-reported or smuggled) does not result in capital flight. So is the more peculiar case of export overinvoicing. Note, however, that unrecorded imports constitute a cause of concern as they imply a loss of government revenue due to unpaid customs duties. Moreover, these imports must be paid for. Therefore, imports underinvoicing may be a symptom of a breakdown in regulation.

The existing evidence clearly demonstrates that while trade misinvoicing cannot be measured precisely, the sheer magnitude of the estimates suggests that the problem is real and it must be tackled with all the attention it deserves. In the meantime, a healthy, dispassionate debate on the data, methodology and other aspects of the research on trade misinvoicing constitute an integral part of the learning process towards generating policy relevant results. In this context, it is clear that the frontier of research in this area lies at the disaggregated level. Only micro-level and institutional analysis can generate the much needed insights on the mechanisms of trade misinvoicing and the associated political economy implications.


[1] Bhagwati (1964, 1967).

[2] See Ferraris (1885) as cited in Morgenstern (1963, Chap. IX).

[3]  Lessard and Williamson (1987).

[4] Ndikumana et al. (2015), Patnaik and Vasudevan (2000); Beja (2006)


Beja, E. L. (2006). Was capital fleeing Southeast Asia? Estimates from Indonesia, Malaysia, the Philippines, and Thailand. Asia Pacific Business Review, 12 (3), 261-283.

Bhagwati, J. (1964). On the underinvoicing of imports [with application to recent Turkish experience]. Bulletin of the Institute of Economics and Statistics (Oxford University), 26, 389-397.

Bhagwati, J. (1967). Fiscal policies, the faking of foreign trade declarations, and the balance of payments. Bulletin of the Institute of Economics and Statistics (Oxford University), 29, 61-77.

Ferraris, C. F. (1885). La Statistica del Movimento dei Metalli Preziosi fra l’Italia e l’Estero (Rome).

Lessard, D. R. and Williamson, J. (Eds.). (1987). Capital Flight and Third World Debt. Washington, DC: Institute for International Economics.

Morgenstern, O. (1963). On the accuracy of economic observations. Princeton, N.J.: Princeton University Press.

Ndikumana, L., Boyce, J. K. and Ndiaye, A. S. (2015). Capital flight from Africa: Measurement and drivers. In S. I. Ajayi and L. Ndikumana (Eds.), Capital Flight from Africa: Causes, Effects and Policy Issues (pp. 15-54). Oxford: Oxford University Press.

Patnaik, I. and Vasudevan, D. (2000). Trade misinvoicing and capital flight from India. Journal of International Economic Studies, 14, 99-108.

UNCTAD. (2016). Trade Misinvoicing in Primary Commodities in Developing Countries: The cases of Chile, Côte d’Ivoire, Nigeria, South Africa and Zambia (Vol. December). Geneva: UNCTAD.

World Bank. (2014). 2014 Little Green Data Book. Washington, DC: World Bank.

Triple Crisis welcomes your comments. Please share your thoughts below.

Triple Crisis is published by

2 Responses to “Trekking the Gold Trail: Misinvoicing in Primary Commodity Exports”

  1. […] 26 – Leonce Ndikumana article on the controversy ‘Treking the Gold Trail’. He says that the main conclusions of the study remain unchanged and argues “The existing […]

  2. Léonce,

    I agree that a healthy, dispassionate debate on the data, methodology and other aspects of the research on trade misinvoicing constitute an integral part of the learning process towards generating policy relevant results.

    It is a shame, therefore that the revised report and its accompanying note do not specifically acknowledge, reference or address the criticisms and comments received including from the South African Revenue Service, Chamber of Mines and South African Statistics agency as well as myself and others – which are neither confused nor uninformed. (For those who want to follow the debate the links are here

    The fundamental problem with the methodology, as the Eunomix report notes is that “there is no consensus, whether theoretical or empirical, that trade data discrepancies correlate with trade misinvoicing, much less that trade misinvoicing would be the primary cause of such discrepancies”. Yet the revised report simply defines trade misinvoicing as “discrepancies in partner trade statistics”.

    The note accompanying the report recognises that it “may be argued that the observed perverse or excessive normal discrepancies could be due to discrepancies between the data in official national statistics and the
    data in the databases compiled by international institutions such as COMTRADE”, but dismisses this by saying “such discrepancies are likely to be minimal given that these international institutions receive the data
    from national sources.”

    The South Africa gold and platinum cases highlight that such readily explainable statistical discrepancies are demonstrably not minimal.

    The idea that monetary and non-monetary gold are like “apples and oranges” is misleading. An apple cannot be reclassified as an orange. But a bar of gold bullion can be classified as monetary to to non-monetary. The UK for example in 2012 reclassified gold held as a store of wealth from ‘monetary’ to ‘non-monetary’ categories to bring it into line with international trade reporting standards.