Why have two official estimates of China’s trade surplus diverged by $270bn over the past year? It’s not data manipulation. Turns out, it’s another case of multinationals exploiting preferential tax rules and messing up the global balance of payments data. Let’s investigate!
Here’s the basic issue. According to China’s customs agency, its trade surplus was $970bn in the year through March, while the State Administration of Foreign Exchange data pegs the goods surplus at only $670bn in the balance of payments.
That divergence reduced China’s current account surplus by about 1.5% of GDP last year. That’s the difference between the IMF labeling China’s external balance as being in line with fundamentals or stronger than the level implied by fundamentals.
So SAFE would seem to have an incentive to manipulate the data to keep China off the IMF’s naughty list. And I came close to accusing them of this a few months back. Sorry!
The difference largely comes down the fact that the balance of payments measures changes in ownership between domestic and foreign entities, whereas merchandise trade tracks shipments across international borders.
SAFE explained in September last year that the gap comes from how it treats foreign-owned factories in China’s bonded zones. SAFE effectively treats those factories as foreign territory, while Customs counts them as being part of China.
My previous scepticism was largely due to a translation error. There are 2 bonded zones in the Customs data. One is just “bonded zones” the other is “integrated bonded zones,” usually translated as “comprehensive free-trade zones.”
I thought SAFE was referring to the former, which has declined as % of exports. But the latter has risen sharply as a share of total exports. And that rise coincided with the widening data discrepancy. 8/
So what’s going on in these FTZs? Here’s a recreation of SAFE’s key diagram. Assume the multinational firm owns the processing firm and warehouse in the bonded zone, probably through a holding company in Hong Kong. 9/
SAFE records the sale of inputs for production in the bonded zone (transaction 1) as an export, whereas Customs doesn’t. Neither captures the transfer of the finished goods from the processing firm to the warehouse (transaction 2). 10/
But Customs will record foreign sales from the warehouse as an export, whereas SAFE will not (transaction 3). And SAFE will record any sales from that warehouse into China’s domestic market as an import, whereas Customs will not (also transaction 3).
In net, SAFE will record more imports and fewer exports than Customs if these multinational firms are export oriented.
We can approximate the impact these transactions might have on SAFE’s estimate by netting out 50% of the exports that Customs recorded from the FTZs, to roughly account for the inputs from domestic suppliers that SAFE counts as exports.
A viola! That pretty much erases the Customs/SAFE discrepancy in the export level. This isn’t conclusive proof. Since Customs doesn’t record domestic sales from the FTZs we can’t verify SAFE’s higher import figures.
And SAFE seems to be missing some export earnings from “merchanting,” or purchasing and reselling goods abroad. China resold LNG it had purchased under long-term contracts to Europe at spot prices last year, but there’s no sign of those deals in the current account.
But I’d guess the profits accrued to shell companies registered in foreign tax havens, which often go missing from other countries’ balance of payments, too. So I’m now convinced that SAFE has not deliberately misstated its trade data to lower the current account surplus.
Still, from the rest of the world’s perspective, it doesn’t matter. Purchases of the goods produced in China’s FTZs by multinationals are likely to be treated as imports in the merchandise trade data and the balance of payments of other countries.
That’s especially true if the factories are registered to a shell company in HK or a tax haven, as they all likely are. Shell companies and tax havens are a major problem for economists tracking global capital flows