In assembling the various components of our national accounts, the national statisticians often start with data from administrative sources and then modify them to fit the precise concepts necessary for the national accounts.
The Balance of Payments (BoP) provides a good example. In many countries, data on exports and imports of goods are collected by customs officials --- the administrators who physically control the border. These data must then be modified to capture goods trade that does not physically clear customs, or that does not otherwise accord with the concepts in the BoP. We'll focus on goods trade below, noting that, obviously, almost all services trade does not clear customs.
In the USA, the Census Bureau receives data on goods trade from Customs and Border Patrol and produces monthly estimates of goods trade "on a Census basis". These data are then modified in a number of ways to construct goods trade data on a "Balance of Payments basis" which is released quarterly by the Bureau of Economic Analysis (BEA). The adjustments can be positive or negative.
Anything that is bought from or sold to a foreign entity, but that remains in the US and so does not clear customs, requires an adjustment. This includes trade in non-monetary gold, which remains with a US custodian such as the New York Fed, as well as trade in artwork and other precious commodities that stay with a US custodian. It also includes purchases of fuel by foreign carriers at a port (customs only processes trade that leaves the port), as well as for "re-exports" (goods that are imported and then re-exported) without passing through customs ("net exports under merchanting"), and a host of smaller items. The BEAs Table 2.4. U.S. International Trade in Goods, Balance of Payments Adjustments reveals that these adjustments are on the order of one per-cent of goods exports and imports.
For other countries, the adjustments can be much larger. Today on Twitter there was a very informative exchange on China's trade balance. First, Adam K Wolfe of Absolute Strategy Research (ASR) showed that Chinese customs data is now recording an additional $270b trade surplus above that in the BoP data (put together by the State Administrator of Foreign Exchange, or SAFE):
Adam points out that this difference is enough to alter whether the IMF considers China's trade balance to be "in line with fundamentals" or not (in which case they woould be regarded as a suspected exchange rate manipulator).
There was a Twitter debate as to whether something nefarious is going on. I tend to agree with Adam that this is a not-unreasonable adjustment for trade occurring in China's various customs-free zones. Multinational firms set up factories in these zones which the BoP data treats as foreign territory (hence exports from these zones to the rest of the world are not included as China's exports) but the customs data treats as Chinese trerritory (and hence includes).
Adam helpfully provides a translation of a diagram that appears in a SAFE document (original can be found on page 11 of this document):
In a nutshell, BoP recognizes the "bonded zone border" as the border between domestic and foreign transactions, while the customs data recognizes the "international border". When the processing firm (owned by the multinational) buys inputs from a domestic Chinese firm, this counts as a BoP goods export but not a customs export. When the processing firm sells goods overseas, this counts as a customs export and not a BoP export. When the processing firm sells to the Chinese market, this is a BoP import and not a customs import. If the typical processing firm in a bonded zone buys Chinese inputs and exports both internally and back to China, BoP exports will be smaller and imports larger than in the customs data. Adam presents some back-of-the-envelope calculations to show that the orders of magnitude are about right to explain the difference. (Adam also argues that the BoP data are missing some re-exporting/merchanting).
If the rest of the world measures imports from China using the international border (for both Census and BoP trade measures), then the rest of the world's trade deficit with China should line up with China's customs trade surplus. Alex Etra of ExAnte Data presented this graph to show they do line up (at least, I think that is what it shows ... with the processing balance adjustment).
Brad Setser, who knows as much or more about this topic as anyone, argues that these measurements are not benign in that they have only begun to show up recently.
He also points to the probable existence of "stateless exports" ... exports by the processing firm do not appear in China's BoP exports and probably do not appear in any other country's BoP exports.
Brad has followed up on yesterday's post with a new short thread. In it, he points out that BoP exports and imports are both smaller than customs exports and imports.
The fact that BoP imports are smaller runs counter to the story I gave above, and so the data still remains a bit mysterious.
One possible explanation is that the processing firm is using inputs produced abroad. These would show up as customs imports but not as BoP imports.
