Trade Data: Use with Care
Not only do world exports and imports not balance, but large asymmetries are found in the balance of trade statistics between countries and regions. These errors do not cancel out on aggregation across countries.
Asymmetries in trade data occur when the declaration of the importer in country A is not consistent with the declaration of the exporter in country B. These asymmetries are derived either from accidental or deliberate errors in reporting trade as a result of incentives for misreporting such as tax and tariff minimisation, the circumvention of quotas or embargos and the evasion of capital controls. They are found across the world, but one notable regular series of discrepancies occurs between China and Hong Kong (SR). For example, in December 2016, customs data showed that China’s imports from Hong Kong rose by 64 percent year-year, while Hong Kong’s official data recorded an increase of only 0.9 per cent. In the same month China recorded a rise in exports to the territory of 11 percent year on year, while Hong Kong registered imports from China having fallen by 1 percent. There are also more serious conceptual problems relating to the realities of contemporary international trade patterns. Many goods classified as imports into are re-exported with no data linking source to final consumption. The ‘Rotterdam effect’ is so-named because Dutch trade flows are over-estimated because goods bound for other EU countries arrive in Dutch ports and, according to EU rules, are recorded as extra-EU imports by the Netherlands (the country where goods are released for free circulation). This in turn increases the intra-EU flows from the Netherlands to those Member States to which the goods are re-exported. To a lesser extent, Belgian and Republic of Ireland exports are similarly overestimated.
But much of international trade is not simply re-exported, but many products are intermediate inputs supplied into a value chain. Trade in intermediate goods accounts for around half of non-oil global merchandise trade and each nation’s contribution to value-added at each stage of production must be estimated to avoid double-counting. Goods in these Global Value Chains (GVCs) can cross borders many times. Conventional trade statistics assigning imports and exports on a dual basis do not reflect this complex reality ignoring the contribution of producers in many countries to value added. This problem has been popularised as the Apple ‘Made in China’ question. Conventional trade statistics consider the iPhone a Chinese export to the US, but the product is entirely designed and owned by a US company, and is made largely of parts produced in several Asian and European countries. China’s contribution is the last step – assembling and shipping the phones.
Conventional trade statistics provide a distorted picture of trade imbalances between countries since what counts is not the imbalances as measured by gross values of exports and imports, but how much valued-added is embedded in these flows. It has been estimated that conventional trade statistics overestimate the US bilateral deficit vis-à-vis China by around 30% as compared to measuring in value-added content based on input–out matrices. Official figures for the bilateral deficit would be cut by about 50% when the activity of export processing zones in China, and Hong Kong, China, re-exports are fully taken into account, but the bilateral deficit of the US with Korea or Japan, the main providers of electronic parts to Chinese assembly plants, would increase in proportion to the reduction of the US–China deficit.
Inaccurate trade statistics matter because they are employed to justify policies to correct supposed imbalances such as tariffs, quotas or currency revaluations. But poor data should not inflame poor reasoning. There are genuine fears that the world may be entering a period when protectionism is gaining political traction using rhetoric based on flawed conventional trade statistics.
Fighting the Gender Wage Gap: The Bahrain Experience
The Flaws in Population Data