from Follow the Money

So, just who is financing the US deficit? (My work on global reserve growth)

September 13, 2006

Blog Post

More on:

Monetary Policy

I spend a lot of time thinking and writing about world’s growing stock of reserves.  Some of that material appears on this blog, some lies hidden behind RGE’s firewall and some never quite sees the light of day.   

The basic story that emerges from the US data – and the global data published by the IMF – is that central bank reserve growth and central bank financing of the US soared from 2001 through 2004.    Central bank inflows and global reserve growth (measured in dollar terms) both peaked in 2004.  And from what we know – and everyone now recognizes that there is a lot that we don’t know because of holes in the IMF’s data – the growth in dollar reserves peaked in 2004 as well.    In 2005, recorded central bank inflows into the US, global reservesgrowth (in dollar terms) and, from what we know, dollar reserve growth all slowed a bit.

I have challenged some key aspects of that data.   The fall in official inflows to the US seemed too large to be fully believable, at least to me.    And the fall in global reserve growth seemed to be more a product of moves in the euro/ dollar – and Saudi Arabia’s narrow definition of its reserves – than of a real fall off in the pace of global reserve growth.   The only part of the story that does seem true is that central banks bought more euros and fewer dollars in 2005.

One graph, I think, is enough to make my point about the impact of valuation gains.    In dollar terms, global reserve growth slowed in 2005.   But not in euro terms.   The growth in the world’s reserves – measured in euros – set a record (the growth in euro-denominated reserves also set a record, but that is a slightly different story).


If adjustments are made for valuation, all of the Saudi Monetary Agency’s foreign assets are added in and the reserves transferred to a Chinese state bank are counted, global reserve growth in 2005 more or less matched 2004’s record levels.    

Total estimated reserve growth in 2004, after these adjustments: $679b

Total estimated reserve growth in 2005, after these adjustments: $658b 

Connoisseurs of my work on reserves will note that I have in the past also added Taiwan’s reserves to the global total reported in the IMF’s data.  I had assumed that the IMF did not include Taiwan’s reserves because Taiwan isn’t a member of the IMF.   But Ted Truman – who knows more about international financial data than anyone – prompted me to double check.   And it turns out that the IMF quietly works Taiwan’s reserves into the global total.    My bad.  

Incidentally, the BIS offers a great summary of the different data sources on reserves, and what is included and excluded from each source.

Fortunately, Taiwan had a bigger impact on the 2002-2004 data than on the 2005 data. 

Deriving an estimate for dollar reserve growth from the IMF’s data requires making a few assumptions about the currency composition of the reserves of those countries who do not report data on the currency composition of their reserves to the IMF.    Yeah, that is you, China.   The currency composition of the Saudi’s large non-reserves assets is also an unknown.  

I assumed that the emerging economies that don’t report actually hold a higher fraction of dollars than those that do.    China recently confirmed that it holds around 70% of its reserves in dollars (I would guess the real answer is a bit over 70% of its reserves) – a higher fraction that those countries that do report.   I have other reasons for this assumption, but well, I cannot give away the entire store. 

That assumption allowed me to estimate total dollar reserve growth and non-dollar reserve growth for 2005.    After a lot of digging (and a few big assumptions), I estimate central banks added $405b to their dollar portfolio in 2005, and $252b to their portfolio of euros, pounds, yen and other currencies.   That compares to an estimated $537b increase in dollar reserves in 2004, and an estimated $142b increase in non-dollar reserves in 2004. 



 Central banks pretty clearly bought more euros and fewer dollars in 2005 than in 2004.    That reflects two things.  First, some central banks try to keep the share of euros (and currencies like the pound that have moved with the euro) and dollars in their portfolio constant.   That means buying dollars when the dollar is under pressure (2004) and euros when the euro is under pressure (2005).   Second, as Ted Truman has emphasized, central banks try not to disrupt the market, and they know that if they sell a lot of dollars when the dollar is falling, they could easily move the market.    Central banks ended 2004 with more dollars than they wanted, and were looking for opportunities to sell.   Interest rate differentials that favored the dollar, the failed European referendum and the Homeland Investment Act provided them the chance they were looking for.

Dollar reserve growth, by the way, both overstates and understates total official financing of the US.   It overstates total inflows because some countries hold the dollar debt issued by other countries (Yeah, I am talking about you China) and institutions like the World Bank as part of their reserves.  It understates total official financing because it leaves out inflows from oil investment funds (at least those oil investment funds that are not managed by the central bank).   Abu Dhabi, Dubai, Kuwait and Bahrain all had a lot of cash to invest in 2005, and they generally place their windfall directly in their investment accounts, not with their central bank.    Most of the Saudi current account surplus appears on SAMA’s books, but not all of it – so various other Saudi accounts also had funds to play with. 

To provide some comparison between the data I derived from the IMF COFER series (augmented with the reserves China transferred to its state banks and the non-reserve foreign assets of the Saudi Monetary Agency) and the US data, I did a quick comparison between my data and the dollar reserve growth total that emerges by combining the US data with the international banking data, following the methodology employed by Robert McCauley of the BIS.

Comparing different estimates of central bank financing of the US

$ billion







Official inflows (BEA data)







Japanese dollar/ euro deposits in domestic Japanese banks (Japanese data, reported to IMF)







Dollar deposits in international banks (BIS table 5c)







Total of these three














Setser estimate for increase in  dollar reserves







Difference between US and BIS data and Setser estimate for dollar reserve increase







The “gap” between my estimate for dollar reserve growth and the sum of recorded central bank inflows to the US and the growth in central bank offshore dollar accounts in the BIS data rose significantly in 2005.   

That could mean that my estimate for dollar reserve growth is way off.   Or it could mean that many central banks used custodians to purchase US debt; official purchases from the Middle East consequently end up appearing in the US data as private purchases by London custodians.   I tend to think the latter explanation makes more sense.   Japan accounted for a large share of dollar reserve growth in 2004, and it tends to use FRBNY for its purchases of US debt.   The oil exporters accounted for a large share of my estimated 2005 dollar reserve growth, and they clearly don’t use FRBNY.

Does this all matter --  I think so.  Consider the following two graphs. 

The first shows the breakdown between official and private financing of the US current account deficit that emerges from the official US data.  The 2006 data point is derived from the q1 data.



The second shows the breakdown between official and private financing implied by my data on dollar reserve growth, along with an assumption that 60% of the official (non-reserve) outflows from the Middle East in the IMF’s data are invested in dollars.



If you believe the US data, there was a big surge in private demand for US debt in 2005.   If you believe my numbers, the surge was a lot smaller.     

The truth is probably somewhere in between – I suspect that my estimate sets an upper bound for official inflows to the US.   And some of what I label official flows are in reality a bit of a hybrid.  A central bank deposits a dollar in an account in a London-based bank, and the London bank lends the dollar to hedge fund that wants to buy US debt -- but doesn’t want the currency risk.  

But I also suspect real picture – if it ever could be determined – would look more like my picture than the BEA’s picture.

There is more behind the RGE’s firewall.    If you have access to RGE premium content though, I do (immodestly) think it is worth looking at .. .

More on:

Monetary Policy