This is a comment to the CMI Insight number 5, published November 2016. The Insight can be found here. 

Comment by
Matthew Salomon, 12 October 2016
Senior Economist, Global Financial Integrity

Broad response

Noting the increased prominence of IFFs on the development agenda in recent years and the role of trade misinvoicing in advancing general understanding of IFFs, the author (Maya Forstater) asks whether a focus on misinvoicing is not too narrow a frame for moving ahead on the IFFs agenda. She concludes that “it is not clear that the influential and widely quoted estimates of trade misinvoicing help us to understand the broader reality of illicit economies and networks in practice.” Further, the author suggests that focusing on misinvoicing is indeed looking under the wrong lamppost in the sense that it is an obstacle to progress: “continuing to allow [misinvoicing estimates] to shape understanding could impede rather than support progress in combatting corruption, organized crime, illegal exploitation of natural resources and tax evasion.”

Without question, focusing solely on misinvoicing as representative of the IFF issue is too narrow a frame—I absolutely agree with the author on this. I’d go further than the author has in this line of argument: the international community should cast as wide a net as possible in approaching a quantitative assessment of the IFFs problem. In fact, there appears to be growing international support to do this.  Among a recent gathering of IFF experts at the UN there was “strong support … to keep efforts (to estimate various sources of IFFs) disaggregated and to work on improving measurement for the separate components …” {1}

IFFs are unobservable and will remain so even as the international community makes progress in its attempts to monitor those flows. Basic principles of statistical science argue for broadening the scope of inquiry to include more indicators of illicit activity, even as indicators are uncertain and differ from each other in quality.

Translating the statistical principles into the terms of the exemplum of the man searching for his lost keys: we should be looking under all lampposts. In this context, there are no “wrong” lampposts.

However, while we welcome and support the author’s desire to broaden the scope of inquiry into IFFs, it is not at all clear why pursuing that objective requires her to discredit misinvoicing estimates. The author’s assertion that misinvoicing estimates are impeding progress is not supported in the note and remains, to us, mystifying.

Concerning methodology

While she acknowledges the existence of misinvoicing as a problem, the author also attempts to discredit the methodology used by Global Financial Integrity (GFI) and others (including IMF researchers) in estimating misinvoicing. In this context, she notes: “not all trade misinvoicing shows up as mismatches in the trade data, and conversely, not all mismatches in the trade data are evidence of misinvoicing.”  Here the author conflates issues of technique and data limitations in a way that obscures matters.

The basic truth of IFFs in general, and misinvoicing in particular, is this: nothing illicit just “shows up.” IFFs are unobservable and can only be gauged, however crudely, through reasonable inference applied to available data, warts and all. In an ideal situation, we might imagine an available administrative database covering all transactions with sufficient detail to identify the problematic passage of goods through intermediate trade hubs noted by the author as well as a host of other factors that earlier critiques have noted (e.g., temporal, geographic, and commodity aggregation effects, valuation effects). The point is this: even with such an ideal database, misinvoicing would still be unobservable. Even in that case, it would still be necessary to draw reasonable inferences about misinvoicing.   

The data available to researchers are far from ideal. Understanding the limitations of the data used is critical, and recognition of those limitations should pervade both the specific implementation of the partner-trade analysis and the interpretation of the results obtained. For practitioners, there is always room for improvement and necessary analytical adjustments are improvements. Such adjustments might be necessitated, for example, by the discovery of subtle deficiencies in the data for particular countries and commodities. Equally important, reporting practices must be fully transparent and must not misrepresent the findings.

GFI uses a variety of available databases and a variety of methods to draw inferences about misinvoicing. The IMF’s Direction of Trade Statistics (DOTS) are the basis of GFI’s annual estimates of trade misinvoicing for all emerging market and developing countries. The DOTS data are the most comprehensive available in terms of coverage, a fact that makes DOTS attractive for use in GFI’s global estimates. The DOTS data, however, do not provide the commodity-level detail that another, less comprehensive but more detailed, database do provide (e.g., UN COMTRADE). Furthermore, the DOTS data are adjusted when other data are available to increase the accuracy of the estimate to some extent; an example is the adjustment for re-exporting Chinese trade through Hong Kong, an adjustment that’s possible only because Hong Kong has made such data available.

In terms of its annual global estimates, GFI is attempting to provide a rough indication of the overall magnitude of misinvoicing, a measure that, by design, tends to underestimate the overall magnitude.{2} The result that a significant share of total developing country trade is potentially misinvoiced is alarming. Moreover, there are concrete (and relatively inexpensive) steps that developing countries can take to reduce such misinvoicing. No, that will not eliminate all illicit financial flows—but curtailing misinvoicing to some degree is surely a step in the right direction.


{1} “Summary of the IATF expert group meeting on Illicit Financial Flows – mapping out a way forward on tax avoidance and evasion,” 22 September 2016, Department of Economic and Social Affairs; Financing for Development Office, p.1.

{2} It is worth noting that the use of more detailed commodity-level partner-trade data, as some critics have suggested, would tend to increase the estimated magnitude of invoicing, not decrease it