China probably holds over 70% of its reserves in dollars

Or, rather, it almost certainly held at least 70% of its reserves in in June 2005.   At the end of June 2005, China’s reserves – counting reserves transferred to the state banks – were around $770b.   China’s holdings of US securities totaled $527b (68% of the total) at the time, according to the latest US survey data.  

And the survey only covers China’s holdings of US securities.  It leaves out China’s dollar bank accounts – so presumably China had slightly higher dollar holdings at the time.   Consequently, it represents a lower limit on China’s dollar holdings.  So I would say that China held at least 68% of its reserves in dollars – and probably more -- in the middle of 2005.

More importantly, the survey data tells a very different story about the scale of Chines financing of the US than the story than emerges just from the TIC data.   Since 2004, the TIC  "flow" data has consistently suggested that a very low fraction of China’s reserve increase – maybe 40% -- was going into US securities.   

For example, between June 2004 and June 2005, according to work that Casson Rosenblatt of RGE has done, China’s reserves increased by around $240b.   During that period, Chinese purchases in the TIC data totaled $97b (exactly 40% of the increase in US reserves).   But if you look at the change in Chinese holdings between the End-June 2004 survey and the End-June 2005 survey, a very different story emerges.  China’s holdings of US securities – according to the survey data – increased by $186b (78% of the increase in China’s reserves).

The TIC data tells one story.  The survey data quite another.  

That isn’t atypical.  The data are collected in different ways.  And both data sets have problems The TIC data comes from data on transactions, and thus is a good measure of flows.  But it sometimes gets the buyer wrong.  The survey data is a measure of the stock of US debt held abroad.  It sometimes does a better job of tracking the ultimate owner of US debt than the TIC data.   The TIC wouldn’t, for example pick up the People’s Bank of China’s purchase of a US treasury bond held by a German pension fund, while the survey data, in theory, would.  But, as Daniel Gros has noted, the survey data also seems to miss some foreign holdings of US debt – not everyone seems to want to report everything. 

In this case, though, the story in the survey data makes a bit more sense.  China is big and it pegs to the dollar, so if it really shifted away from dollar reserves, it presumably would put quite a bit of pressure on the euro/ dollar – and thus on the RMB/ euro.

It also leads me to wonder about some recent studies that failed to find much impact of central bank purchases on US rates – for example, this new Fed study (Hat tip, the Capital Spectator).    The Fed study found evidence of a conundrum – bond yields were lower than predicted by the model.   That is consistent with most work.   But it didn’t find evidence that central bank buying – estimated using the Fed’s custodial data – explained low yields. 

Maybe.  Other studies, like this one from the Banque de France, have come to a different conclusion.  I haven’t had the time needed to really understand the sources of the difference.   

One reason may be that the Banque de France study looked at 2003 and 2004 but not 2005 data – and in 2003 and 2004, Japanse intervention dominated the data.   Japanese buying showed up in the Fed data, in the TIC data and in every other data set.   Not so for the People’ Bank of China.  Or the Saudi Monetary Authority.  I suspect that the Fed custodial data – like the TIC data – understates central bank demand right now.   

That, I strongly suspect, complicates efforts to find evidence that central bank demand directly influences the Treasury market.  

The Fed study certainly goes against my priors.  Which makes me suspicious.  But that doesn’t mean it is wrong.  It just means I am suspicious.    Applying statistical techniques to imperfect data can generate less than ideal results.  And I would be rather surprised if record levels of central bank intervention -- over $600b a year since 2003 -- and China's efforts to hold its real exchange rate down in the face of pressures for appreciation (The Economist doth protest too much; Frankel has shown that China's real exchange rate is quite low even compared to other relatively poor countries) has had no impact on US rates.

Bernanke should be worried too.  Official intervention -- not private flows -- have been the vector that has brought the emerging world's savings surplus to the US.  Read the data in the WEO's statistical abstract if you doubt me.   So if official action has no impact on US rates, it is hard to see how the emerging world's savings glut induces the US to save less ...

That said, it is also true that foreign central bank demand for US treasuries has been relatively strong in other periods of relatively high reserve growth when US yields weren’t unusually low.  Periods in the 1990s for example.   And that does suggest that central bank demand isn’t the only factor at work.