The iPhone, Chinese Customs Data, and China’s Balance of Payments Puzzle

Almost all iPhones are still assembled in China.

The chips are generally imported, but most of the other components are Chinese. See Yuqing Xing’s update of the classic knock down study.

And it is well known that Apple itself doesn’t run the factories that produce its electronics; it relies on a set of contract manufactures, most famously Foxconn and Hon Hai.

The basics of Apple’s tax structure are also now pretty well known (see the Levin committee, and subsequent work by the New York Times and Seamus Coffey).

Apple Ireland (if the phone is intended for sale outside of the Americas) or Apple U.S. (if the phone is intended are for sale in the U.S.) will notionally buy the phone from the contract manufacturer. The cost of production (or the of goods sold) for a high-end smart phone is now generally estimated to be in the $300-$400 range (Xing put the bill of materials at around $400 a few years ago).

That is, in fact, the price that Apple (U.S. pays) when it imports a phone into the U.S., according to the U.S. International Trade Commission. Apple Ireland presumably pays the same price (it is the same phone!)

Apple (Ireland or U.S.) then adds a big markup and hands the phones over to its retail operations, who manage Apple Stores and the like. The net result is that the bulk of the profit on an iPhone is attributable to Apple’s United States headquarters, or to Apple (Ireland)—which “buys” the rights to Apple’s design and intellectual property outside of the Americas by paying something like 60 percent of Apple’s global research and development budget (this is a classic “cost-share” arrangement).

We know that Apple makes a large profit in Ireland for two reasons. One, of course, is that Apple reports a big international profit (and its tax disclosure highlights Ireland in particular); the other is that there is a big gap between Ireland’s reported customs goods exports and Ireland’s reported balance of payments goods exports.

Customs exports only include goods that actually cross a physical border; they correspond with the standard conception of exports. Balance of payments exports, by contrast, include the global operations of resident Irish companies (Apple’s Irish subsidiary became an Irish resident in 2015, creating “leprechaun economics”).

Apple Ireland’s markup effectively counts as an Irish export in the balance of payments data—it is recorded as a gain from merchanting (buying and selling goods produced around the world at different prices; it basically doubles Ireland's reported goods exports).

So far, so good—this is all now (fairly) well understood, and at this stage old news.

But there is now a twist—a big one: China.

China’s “balance of payments” goods surplus is now much smaller than its customs goods surplus, something that I have highlighted previously. China has told the IMF (see appendix 7) that this is because of changes in how it accounts for “factory-less” manufacturing (aka firms like Apple) operating in bonded or free trade zones in China (like Zhengzhou, China’s “Apple City” where Hon Hai makes many of Apple’s phones; the 2016 New York Times article on Zhangzhou's benefits for Foxconn/ Hon Hai real like evidence for a trade case, particularly this: “Foxconn receives a bonus when it meets targets for exports”).

According to the IMF, China’s National Bureau of Statistics stopped using customs data to calculate the balance of payments data (opting instead to use internal payments data and a survey to in theory better account for “factoryless” manufacturing in 2022. Since this change (which was statistically faded into the trade data in 2021), an enormous gap has developed between the customs data and the balance of payments data. Consequently, there is a real discontinuity in China’s reported balance of payments. I calculate that China’s good surplus (in balance of payments terms) would be about $360 billion more than is now reported (just under 2 percent of China’s GDP) if the balance of payments data was calculated using the customs data with the adjustments for the cost of insurance and freight that China used before 2020 (See Section 3.1.2 of China's SDDS data disclosure, which has not been updated to reflect the new methodology/source data).

It is thus interesting to ask if the operations of a firm like Apple—by far the biggest potential source of a discrepancy—could account for this gap. 

There are actually two issues here.

One is how the Apple’s contract payments to Hon Hai, Wipro, or Foxconn for phones sold outside China enter into the customs on and balance of payments data. The other is how Apple’s payments to the same firms for phones sold inside China enter into Chinese customs data and balance of payments data.**

The Chinese authorities apparently have told the IMF that under the old methodology Apple’s payments to Hon Hai and other contract manufacturers were entering the customs data at an inflated wholesale price well above the actual cost of production.

However, that is implausible for phones produced for sale outside of China. Apple isn’t handing its profit over to Hon Hai by paying Hon Hai $800 a phone. No, it pays Hon Hai $300 or $400 (per the U.S. International Trade Commission “the average unit value of a U.S. cellular phone import rose from $257 in 2019 to $352 in 2022”). The markup for design and software is then either applied in the U.S. or in Ireland so that the profit on the phone is booked outside of China. After all, the point of Apple’s tax structure is to move the profit to Ireland not to keep it in China (and out of the U.S. for non-U.S. sales, in the first instance).

Unless China has actual data showing that Apple was paying a Chinese entity something like $800 a phone, there shouldn’t be any gap between the customs value of iPhones exported from China and the balance of payments value of those exports. The export price from China should be the same in both datasets.

The IMF should obviously confirm this. Smart phones are HS 851713 in the harmonized trade data, and China should be able to provide this data to the Fund (it can be mapped to the U.S. import data, and presumably also cross-checked against Ireland’s data, though in Ireland’s case there may be more data confidentiality issues).

The only real possible source of difference between the customs data and the balance of payments data is for Apple’s in-China sales (Apple reports $30 billion in operating profit in China on $72 billion in sales, which sets an upper limit on the size of the impact the “Apple” adjustment should have on the balance of payments data).

An iPhone assembled in China for sale in China obviously doesn’t actually ever physically leave China’s borders, so it probably doesn’t enter into the customs data.

And, at least conceptually, there are three ways in which Apple’s in-China profits (generated via the employment of Chinese contract manufactures for its Chinese sales channels) could enter into the balance of payments data.

Suppose Apple China buys the phone from Hon Hai at Hon Hai’s “export” or contract price, applies a markup, and then sells the phone to retailers in China at Apple’s list price—with the difference, of course, being Apple’s Chinese profit.

That profit would then enter the balance of payments as a debit in the income line (a foreign company is making profit in China) and it would reduce the Chinese current account surplus by a commensurate amount.

There is also a second possible structure: a foreign firm could sweep most of that profit out of China by charging their Chinese subsidiary a large royalty for the use of the firm’s global design and software. The resulting royalty payment would enter the balance of payments data as a services import (Apple China is paying to “import” Apple’s global IP), and reduce the current account surplus by a commensurate amount. Apple Ireland would (presumably) get the royalty payment, adding to its Irish profit.

The third possibility: since Hon Hai operates in a bonded zone (“The customs operation is also in a so-called bonded zone, an area that China essentially considers foreign soil, subject to different import and export rules. This setup allows Apple to sell iPhones more easily to Chinese consumers”). It is possible that Hon Hai’s production of phones for the Chinese market already enters the customs data as an export (think of Apple Ireland importing a phone from a bonded zone in China) and then re-enters the Chinese data as an import after Apple Ireland applies its markup (think of it as onshore China importing from offshore China plus offshore Ireland). This should be something that China can easily clarify, as imports have to pay VAT at the border (there wouldn’t be a tariff on phones) and thus the tax records for Apple’s China operations should indicate how the transactions linked to Apple's in China sales are reported. 

Presumably, this is how Apple now enters into the new balance of payments data, with the contract manufacturer exporting a phone to a non-resident company (Apple) and then Apple’s Chinese operations making a payment to Apple’s global operation when they reimport the phone (at a marked up price).***

In either case, Apple’s profit should have already been somewhere in China’s balance of payments data before the new methodology was put into place—be it as a profit for Apple China (the income balance), a royalty payment (in the services balance) or a goods import (at a higher price than the initial export) and thus a debit in the goods balance that offsets surpluses elsewhere.

So, if Apple’s onshore operations were accurately captured in the pre-2022 data, the change in methodology in 2022 should have had no impact on the overall current account—though it would potentially shift a part of the services or income deficit over to the goods side of the ledger.

Bottom line: given that we know Apple’s operating profit in China is $30 billion, we also know that this accounting cannot explain the bulk of the $360 billion or so gap between the reported goods surplus in China’s balance of payments using the pre-2022 calculation method and the reported goods surplus using the new calculation method. The explanation that China and the IMF have put forward for this change just doesn’t quite yet add up. There is no good reason to think that improved accounting for the operations of Apple (the largest user of a “factoryless” manufacturing model) comes close to explaining the current puzzle in China’s balance of payments reporting.

As importantly, there are straightforward ways in which the IMF could check and see how China is accounting for Apple in its customs and balance of payments data. For example, does the export price for U.S.-bound, Chinese-made phones map to the U.S. import price reported by the ITC? What’s the export price for other phones destined for global markets, and how exactly was the VAT/VAT rebate handled on Apple’s in-China sales?

Chinese data isn’t sufficiently detailed for outsiders to be able to assess how China handled these issues before, but there is no good reason to think that a change in methodology (of a change in source data) should generate the kind of swing seen in China’s reported balance of payments over the last three years.

* Apple’s depreciation allowance for the purchase of Apple Jersey was fixed back in 2014, and it is now much smaller than Apple’s Irish profit. As a result, Apple has been paying the Irish tax rate (now 15 percent, previously 12.5 percent) on a much higher share of its Irish profit than in the past. There aren’t many secrets here, Apple is so big that it is hard to miss in both the balance of payments and the tax data.

** The number that is reported to the Chinese authorities when the phones leave China, which matters for things like the Value-Added Tax (VAT) rebate.

*** The balance of payments concept is a transfer of value to a non-resident, so the fact that Apple’s global operations are not Chinese residents is key. Apple’s retail operations in China would be structured so as to be Chinese tax residents, so the phone would be imported at the marked up price for sale in China.