What have the world’s central banks been up to? The latest Treasury survey provides some clues

I suspect I am one of a very small number of people who eagerly awaits the publication of the Treasury’s annual survey for foreign portfolio investment.     Wall Street doesn’t find it particularly useful: it looks backwards, and come out with a long lag.   The most recent survey tells us what happened between July 2005 and June 2006 – ancient history.     Academic economists don’t particularly like it either.    The survey data doesn’t form a nice time series that can be easily downloaded and analyzed.   It wasn’t even an annual publication until 2002.    

To the uninitiated, the survey’s results look like one page of numbers after another.  But those numbers tell stories.  If a country’s reported portfolio has lots of Treasuries and Agencies and no corporate bonds or stocks, it screams “central bank.”      Little quirks jump out.   Russia and India both hold a lot of short-term securities.    But not quite the same short-term securities.  The Bank of Russia holds Agencies and won’t touch Treasuries or short-term corporate paper, the Reserve Bank of India holds Treasuries and corporate paper, but won’t touch Agencies.   China’s state administration of foreign exchange has been a big buyer of “Agency ABS” – basically, mortgage backed securities with an agency guarantee.   “Official” buyers bought $54.5b of “Agency ABS’ between the end of June 2005 and the end of June 2006.   China bought $51.5b.   The Bank of Russia, by contrast, won’t touch Agency ABS, but it snaps up an awful lot of the debt the Agencies issue directly.  It bought about $30b of short and long-term Agency paper between the end-June 2005 and end-June 2006.   

And so on.   If you spend enough time with the data, you can learn a lot about “he who cannot be named.” 

Here are the five big stories that emerge – in my view – from the data.  

1.   China has not diversified its reserves.   Let me shout that: “CHINA HAS NOT DIVERSIFIED ITS RESERVES.”      According to the survey, China bought $193b of US long-term debt between the end of June 2005 and the end of June 2006.   To state the obvious, $200b (rounding up) is a lot.   Someone should write a story about how China influences the US economy that emphasizes the export of debt – not the export of goods.  

The US “exported” $87b of long-term Treasuries, $83b of long-term Agencies, $23 of long-term corporate debt to China between the end of June 2005 and the end of June 2006.   During that period, the US exported $48b of goods to China.    No wonder many on Wall Street like an unbalanced world … 

China financed its long-term debt purchases in part by reducing its short-term debt holdings (they fell by $22.5b) and its US bank deposits (down $1.7b).    But its total US holdings still rose by $170b.   

At the end of June 2006, China held $717b in US assets (by my count, which adds bank deposits to the data in the securities).      That works out to 76% of China’s end-June reserves ($941b), and to 69% of my estimate of Chinese government’s foreign assets ($1035b in foreign assets).  My broader measure counts the reserves shifted to central Huijin as part of the bank recapitalization and include an estimate of the central banks’ swaps with the domestic banks.    

Note that this estimate just covers the China’s holdings of US securities and its US deposits – it doesn’t count China’s holdings of dollar debt issued by other emerging economies or its dollar deposits in the international banking system.  I’ll have more on this, but the basic data remains consistent with a 70% or even a 75% dollar share in China’s portfolio.

2. Russia, by contrast, did diversify.   That isn't a surprise.  Russia said as much last June.  

Russia's total holdings of US debt rose by $35b between the end of June 2005 and the end of June 2006.   Its bank deposits – best I can tell – fell by about $15b over the same period, for a net inflow of $19b.  

During that period, the bank of Russia reports that it bought $91.4b of foreign exchange.    If the Bank of Russia had wanted to hold the dollar share of its reserves constant at 70%, given fluctuations in the euro/ dollar and pound/ dollar, it would have needed to have bought about $70b of dollars.   Cutting its dollar share from 70% to 60% would have implied about $45b of purchases.   Bringing its dollar share down from 70% to 50% would have implied $20b or so of purchases.

Draw your own conclusions.

Actually, the US data isn’t definitive.     Many countries hold a large share of their dollar reserves in the international banking system (India is one example, but Libya and Nigeria are others), so changes in US holdings are not a perfect proxy for changes in dollar holdings.  Indeed, the fall in Russia’s “onshore” dollar deposits was likely offset by a rise in Russia’s “offshore” dollar deposits. 

However, Russia also reports the valuation gains (and losses) on its fx portfolio.  Judging from the size of its reported valuation gains, Russia hadn’t diversified in advance of the big dollar move in q2 but it had diversified in advance of the big dollar move in q4.      

My guess is that Russia was actively diversifying during the second quarter, and quite possibly contributed to the dollar’s fall then.   It then drew down its dollar balances to cover the big payment to the Paris Club in q3, bringing the dollar’s share of its total reserves down to around 50% by the end of q3.

3.  Japan hasn’t been a big buyer of US securities.    Japanese holdings of US debt actually fell slightly (and the BoJ/ MoF looks to have swapped $15b of Treasuries for $15b of Agencies).    

Does this mean that the carry trade has been overstated?   I suspect not.   

Rather it suggests that “real” Japanese money wasn’t a big buyer of US debt between end-June 2005 and end-June 2006.  
 

That may be because Japanese investors preferred to buy kiwi or Aussie dollar debt.     

But there are other ways of doing the yen/dollar carry trade – ways that wouldn’t register in the US survey of portfolio holdings.  The forward market, for one.   And if a US or European bank borrows yen in Tokyo, lends those yen to its London office and the London office lends the yen to a London based hedge fund that buys US corporate debt that would register as an outflow in ‘other investment’ in the Japanese data, and a purchase from the UK (or Ireland) in the US data.  

Indeed, the absence of stronger signs of real money outflows from Japan actually suggests (I think) that true “leveraged carry trade” type flows have played a bigger, not a smaller, role in bringing Japan’s savings surplus to global markets.   It is also possible though that some Japanese real money flows may also have been structured in ways that make use of tax havens in the Caribbean.   (See this chart) 

4.   Flows from the Middle East picked up.   From June 2005 to June 2006, the US holdings of Asian oil exporters (basically the Gulf) rose by $82b.     Their short-term and long-term Treasury holdings jumped by $40b, and their equity holdings jumped by about $30b – but some of that rise reflects the rise in US equity markets.   Say $20b came from new inflows.  Throw in a $7b rise in “onshore dollar deposits”  and on a flow basis, inflows were probably closer to $77b.     That is still up from the $35-40b total inflow from June 03 to June 04 and from June 04 to June 05.

The composition of the Gulf's inflows (and total holdings) is interesting – basically, it is weighted heavily toward the safest of assets (Treasuries) and equities, while holdings of Agencies and corporate debt are low.   

The most obvious explanation?     The GCC’s assets are either in conservative central banks (SAMA) that buy Treasuries, or aggressive oil investment funds that buy equities. 

$70-80b is a reasonable total for the Gulf.  I don’t have a June-05 to June 06 estimate of the Gulf’s current account surplus, but it shouldn’t be all that different from the 06 surplus, which probably will come in a bit under $200b (I expect the GCC countries’ surplus to be in the 180b range, and the US data theoretically includes Iran as well).    Still a 50% (roughly) dollar share for Gulf flows seems seem a bit on the low side.   It is, after all, a region that pegs to the dollar.   

The total holdings of the GCC countries – roughly $250b in the survey – are also on the low side.   Total Gulf foreign assets should be in the $1 trillion range.   The Saudis had $186b in their central bank at the end of June 2006.  ADIA reportedly holds $600b (if not more)  in foreign assets. KIA (Kuwait), QIA (Qatar) and Dubai International Capital all have substantial holdings – and there are some rather large “private” accounts in Saudi Arabia as well.

How to explain the gap?   One possibility is that the dollar allocation of the Gulf’s various official investment funds is quite low.    George Magnus of UBS, for example, thinks oil funds have been big buyers of Asian equities.  

The other possibility is that the survey data doesn't capture all the Gulf's dollar holdings.   We know it doesn’t pick up offshore dollar deposits – but that is likely a bigger factor for North Africa than the Gulf.   The Gulf hasn’t been a big source of net deposits in the international financial system.   And I am not sure how assets that have been turned over to private managers are accounted for in the survey.  The Gulf states like private equity for example, and they also employ a range of private portfolio managers.    Some of those holdings may be reported by the private manager, not the country.  If so, they might show up in the data from Luxembourg, Ireland, the UK and Switzerland.   That though is just a guess.   I don’t know. 

5.  The US flow data has systematically understated official purchases, and specifically Chinese, Russian and oil purchases. 

Look at the revisions in the data series on foreign holdings of Treasuries.   The holdings of the “UK” dropped, and the holdings of China and the oil exporters rose. 

If the US published a similar table for Agency holdings, I suspect it would show something similar, but with a big rise in Russian holdings.   Russia seems to use a US custodian for its holdings of very short-term agencies, but a London custodian for its holdings of longer-term agencies.   

The gap between the “survey” data and the “flow” data is most obvious for China.     The TIC showed $43b of Chinese purchases of long-term Treasuries.   The Survey $87b.    The TIC showed $35b of Chinese purchases of Agencies.   The Survey $83b.   The TIC data also showed $31b purchases of corporate debt while the survey only showed a $23b in increase.    In total, the TIC data showed $109b of Chinese purchases of long-term debt, while the survey showed a $193b increase in Chinese holdings.    

The global data tells a similar story.    The TIC data (used to calculate flows in the US BoP) showed $96b of official purchases of long-term Treasuries.   The survey shows a $159b rise in stocks.    The TIC showed $96b in long-term Agency purchases, v $149b in the survey.     For corporate debt the difference is less dramatic -- $25b in the TIC data, and $36b in the survey. 

Sum it up though, and the survey shows a $344b increase in official holdings of long-term debt – v $217b in the TIC data/ the US BoP data.       China accounts for well over ½ of official purchases.

$344b is a lot closer to my expectations than the $217b. 

There are a couple of other things to note.   Official holdings of short-term debt fell by $17b.   Official holdings of foreign equities rose by $37b, but some of that came from the rise in the value of existing holdings, not new inflows.  Call it a $17b equity inflow and a $17b short-term outflow, or a wash.

However, not all dollar holdings show up in the US data.  If you add the $105b net increase in central bank dollar deposits reported by the BIS over this period, total dollar reserve growth was at least $450b.    Christian Menegatti and I estimated – based on the COFER data – total dollar reserve growth of $475b for the same period.    Our estimates of the global share of dollar reserves aren’t too far off.  

There are a couple of complications – the US data includes oil fund purchases, while the COFER data doesn’t.   But we have about $35b in swaps with the Chinese banks in our global reserves total for that period and the dollar deposits/ securities purchased by the Chinese state banks wouldn’t register as official purchases in the US or the BIS data (even though the Chinese banks are playing with money borrowed from China's central bank).    Getting everything to match up would take a bit more time …  

Jay Bryson of Wachovia argues that this data shows how little the US depends on official inflows -- $350b isn't that large relative to the $850b to the US current account deficit over that period. 

I would though note that that official inflows account for over ½ of total foreign purchases of US long-term debt ($350b of the $600b increase in the survey).  Moreover, other purchases of US assets were financed by the $100b or so in dollar deposits in the international banking system.   The IMF considers offshore dollar deposits a close proxy for onshore dollar deposits.  

All in all, the official sector probably financed about ½ the deficit between the end of June 2005 and the end of June 200.  That was time when US interest rates were far higher than European or Japanese rates.    It was also a period when the US economy looked good, and Europe – after the referendum – looked a bit shaky.    

The available data suggests that private flows – in good times – are enough to finance a $400 or even $500b US deficit, not a $850b deficit.    And in not-so-good times, well, the official sector has to provide more, not less, financing to keep the current system going.  

In q3 06, q4 06 and q1 07, the data that Christian Menegatti and I follow suggests that US dependence on official inflows rose (our latest analysis is here – RGE subscription required).

There is another story here too.  As the US fiscal deficit has shrunk, the world’s central banks shifted from financing the US Treasury to financing both the US Treasury and US households.   From mid-2004 on, there has been a very obvious surge in official demand for Agency bonds.   

That helps explain why the (significant) improvement in the US fiscal balance over the past few years hasn’t led to a fall in the US current account balance.   The US national accounts data shows that a fall in household savings (along with – until the last two quarters of 2006, a rise in residential investment) offset the fall in the fiscal deficit.    And the portfolio survey shows that the official sector shifted some of its financing from the US government to US households.

Official accounts now hold about 70% of long-term Treasuries held abroad – and about 50% of all Agency bonds held abroad.    If you look just at the debt the Agencies issue themselves, not the debt they guarantee, the share of official holdings is even higher, close to 60%.    And that is at the longer end of the curve.   The official sector accounts for 80% of all foreign holdings of short-term agencies.    

Fanny and Freddy increasingly intermediate between central bank demand for safe dollar assets that pay a little more than Treasuries and the US housing market.