Empirical studies on international trade extensively rely on the use of mirror trade statistics, i.e data reported by trading partners. However, while extensive reviews have been done on how to use mirror data to compensate poor quality data or to proxy transportation costs, very few has been done to see if and how the gap between the declared and mirrored disaggregated bilateral data could be used to capture informal cross border trade. Indeed, beyond the valid logistic reasons to explain why reported bilateral export flows from one country do not match the respective reported imports of its partner country, deliberate misreporting could significantly contribute to explain those discrepancies, either through misevaluation or misclassification of the imported goods, notably to evade tariffs and taxes. This paper proposes a review of the reasons for the gap between matched partner data, before investigating stylized facts from UN-COMTRADE data. Empirical analysis relying on econometrical panel data over a worldwide set of data at the 6 digits level evidences that discrepancies from the mirror data are not erratically driven. A statistically significant relationship between the gap and macroeconomic variables such bilateral distance, gdp per capita, average tariffs, foreign direct investments (FDI), implementation of regional trade agreements (RTA) have been evidenced. Based on these preliminary correlations, a probit has been run on orphan imports (imports reported by importing country without equivalent by exporting country) and predicts accurately up to 68% of these misclassification cases. Thus, part of the gap can be predicted by macroeconomic variables, some of them suggesting a relationship between cross-border trade flows misreporting and fraud opportunities to evade tariffs and taxes.
