from Follow the Money

The impact of central bank demand on US bond yields.

October 18, 2006

Blog Post

More on:

Monetary Policy

In late 2004 and early 2005, Nouriel and I estimated that the actions of foreign central banks might be holding US yields down by as much as 200 bp. 

That estimate found its way into Pam Woodall’s Economist Survey.  And it then caught former Treasury Secretary Rubin’s eye. 

Robert Rubin is a wise and cautious man.   He didn’t want to endorse the results of work that he hadn’t personally evaluated.   So he qualified his remarks by noting: 

"I read something the other day (probably from the Economist) that -- if it were not for the inflow of capital from abroad -- our long-term interest rates would be as much as 200 basis points higher.  I don't know who did the work or how accurate that is.  But you may have had a distortion of the bond market by these vast inflows." (emphasis added)

Rubin caught us.  The 200 bp call didn’t emerge from careful empirical work.    It represented our best judgment at the time.  Back then there hadn’t been all that much empirical work on the question.    And some of that work had found effects that seemed much too small, given the scale of central bank intervention.  

Subsequent empirical work, I think, has been reasonably kind to our initial judgment.  Let’s review the evidence.

Remember, the 200 bp estimate was made in the fall of 2004 and in early 2005 – back at the peak of the “conundrum.”    The US economy was recovering, the US fiscal deficit was quite large by any standard, US policy (short-term) rates were rising and, well, US long rates weren’t. 

That also was back at the peak of foreign central bank demand for Treasuries.  I think it is reasonable to conclude that an awful lot of that demand came from the Bank of Japan, acting on behalf of the Ministry of Finance.  Other central banks do buy Treasuries, of course, but many do so in ways that don’t register quite as cleanly in the US data (see below). 

Moreover, our estimate explicitly included both the direct and indirect effects of central bank intervention.

The direct effect is obvious.   Central banks have been intervening in the foreign exchange markets of the world at an unprecedented pace over the past few years – and as a result, they have a ton of dollars and euros that they have to invest.  And they generally invest those funds in relatively safe bonds – whether Treasuries, Agencies, of Euro-denominated bonds of comparable quality (one interesting side note: the Italians are the biggest source of supply of euro-denominated government bonds, and, well, they aren’t exactly the best of all euro-denominated sovereign credits – something that may contribute to the dollar’s enduring popularity). 

The indirect effects are more amorphous and harder to quantify.

Expectations of central bank intervention shape the actions of private market participants.   This is most obvious in Japan.   Japan has a history of intervening heavily to keep the yen from strengthening (too much).   If market participants expect Japan will act that way in the future, they will be more comfortable borrowing yen to buy higher-yielding assets abroad.   After all, the big risk of such a trade is a sharp appreciation of the yen.   And if the ministry of finance won’t let that happen, the carry trade becomes more attractive.   Think of it as government action that reduces the risk of a “fat tail” kind of outcome … 

That argument can be generalized.   The fact that central banks stand ready to buy dollars and dollar-denominated bonds if private investors aren’t willing to buy (enough) of them to finance the US deficit makes private investors more comfortable holding dollars and dollar-denominated bonds.  Ask PIMCO.   All their forecasts are premised on the continuation of the “Bretton Woods” system.

More generally, by keeping their exchange rate weak, the central banks of countries like China held down the price of their exports and thus the price of many imported goods, That helped keep inflation down in the big advanced – and increasingly housing and consumer-drive economies.    

In retrospect, though, the nature of this effect isn’t as clear as it seemed at the time. Intervention by China’s central bank has helped keep the RMB from rising and thus helped to hold down the price of many traded goods.   That I have no doubt.   And it thus contributed to holding core inflation down.     But China’s overall impact on inflation is a bit less clear – intervention contributed indirectly to China’s investment boom, and thus to rising commodity prices.   Philippe d'Arvisenet of BNP Paribas has a good summary of the debate -- he argues that rising commodity prices have offset about half the effect of rising imports from low wages countries.  And then there are all the issues around how to accout for the rising price of housing …

Quantifying these indirect effects is hard.  It isn’t obvious to me that there is any real way to gauge how expectations about central bank behavior have shaped private expectations. I don’t know of anyone who has seriously tried.    At the same time, I think many in the markets accept that central banks are now shaping their own activities (Ken Rogoff has a few colorful quotes). 

But there have been a series of studies that have tried to quantify the direct effect of central bank purchases on US bonds.     The Banque de France (Gilles Moec and Laure Frey) found the largest effects: they estimated that, at its peak in 2004, central bank demand for Treasuries subtracted 125 bp from Treasury yields.

That study, though, has a couple of problems.   

One shouldn't be a problem, but it probably is.   French economic analysis is (unjustly in my view) considered an oxymoron in some parts of the US.  Their study hasn't gotten the attention it deserved.

The other is a real problem.   The Banque de France worked off the US data on central bank purchases of Treasuries.   That isn’t a problem for 2003 and 2004 – back when the Japanese were intervening like mad and buying treasuries like mad and those purchases showed up as both Japanese and central bank purchases in the US data.   But it is a bit of a problem now.    Lots of other central banks prefer to buy their Treasuries in ways that don’t register as central bank purchases in the US data.    

That was likely the case in 2005.  Central bank demand for Treasuries fell sharply (no more Japanese intervention).  But total foreign demand for Treasuries didn’t fall (lots of petrodollars had to be invested somewhere … ).   The Banque de France methodology would record a big fall in central bank demand, when, in all probability, “true” central bank demand didn’t fall as fast as implied in the US data.

A recent study by Francis Warnock and Virginia Warnock deals with both problems.   Warnock teaches at Virginia, is affiliated with Trinity College Dublin and has done lots of works for the Fed as well as the IMF.  He can hardly be accused of being (god forbid) French.   And his approach looked at total foreign demand for Treasuries, not just measured central bank demand. 

The result: Warnock and Warnock estimate that foreign demand for US treasuries held yields down by about 90 bp in the spring of 2005 – the last data point in their study.

And an awful lot of that effect seems to be the product of the Asian demand.  See Figure 4b. 

But if you look at their Figure 3a and 4b, it is clear that foreign demand had a bigger impact on Treasuries in 2004.  At its peak, they seem to estimate an effect of around 130 bp.  

OK, that isn’t 200 bp.   But it isn’t 25 bp either.   And 130 bp is only the direct effect.   Warnock and Warnock do not try to do the impossible and estimate the extent Chinese intervention has helped hold down core inflation, the extent that Japanese intervention – and expectations of renewed Japanese intervention should the yen ever move too fair – have facilitated the yen carry trade or the fact that Bill Gross sleeps a lot sounder at night despite holding a lot of dollar-denominated claims with a low nominal yield (currently under 5%) on a country with a huge current deficit … 

True, a substantial debate remains.   But I still would argue the 200 bp estimate has aged relatively well – though I would tend to put the effect at closer to 150 bp today.

It is worth noting that foreign demand for Treasuries of all types dried up in the first half of the year.   That reflects a couple of things.  The Japanese authorities seemed to shift some of the Treasury portfolio into Agencies.   The Chinese are buying more Agencies and corporate bonds (most likely mortgage backed securities that are not guaranteed by an Agency) and fewer Treasuries.   And the oil exporters seem to have lost interest in Treasuries as well.

It isn’t clear, though, that the impact of central banks on US markets disappeared.   Agencies are a pretty close substitute for Treasuries.    Not a perfect substitute, but a close substitute.   

Still, as central banks snap up a wider range of debt – and, alas, often buy debt in ways that the data doesn’t fully capture – the challenge of accurately estimating their impact on the market will only become more difficult.

More importantly, it seems likely that central banks didn’t completely lose their appetite for Treasuries.  It seems the world's central banks were waiting for the Fed to end its tightening cycle before locking in long-term rates.   Foreign demand for Treasuries remerged in August.  I would bet September was similar.  

One last set of points for the true geeks.  Warnock and Warnock’s appendix has a very useful run-down of the limits of the US data on foreign inflows. 

Data from the custodial accounts of FRBNY (the weekly H4.1 release) provides the most timely data on foreign official holdings.   But Warnock and Warnock note that the FRBNY data is very partial: 

“FRBNY is just one of many custodians that foreign governments might use.  For reported US capital flows data, the FRBNY is the custodian of choice for many of the world’s central banks and finance ministries …. However, some foreign governments, notably Middle East oil exporters but also others, avoid the FRBNY and this source is best described as partial.” (emphasis in original)

Well said.  Obviously, as oil exporters reserve growth increased , the value of the data series from the FRBNY custodial holdings fell.    More recently, Russia seems to be running down its custodial balances in the US – though I am not 100% sure these are held at FRBNY.  China doesn’t entirely avoid FRBNY (best I can tell), but it also doesn’t seem to use it almost exclusively (unlike, say, the Japanese). 

The TIC data doesn't hinge on FRBNY custodial data.  But it too has a few problems.

“It is well known that the TIC system is not able to correctly differentiate between foreign official holdings and other foreign investors (such as pension funds, stabilization funds, insurance companies and others) when the trade is made through a third party intermediary, we will utilize overall TIC foreign flows into Treasury and Agency bonds in constructing our preferred measure of foreign official accumulation.”

They also note that flows into Agency bonds need to be adjusted for amortization payments. 

I agree.  That is why I am skeptical of studies that use very narrow measure of foreign official holdings (the FRBNY series, even the BEA’s data) as the basis for estimating the impact of central bank flows on the US.

Indeed, if the Chinese are buying private mortgage backed securities – that is, mortgage backed securities that lack an Agency credit guarantee, total inflows into Treasuries and Agencies may miss some central bank purchases … 

More on:

Monetary Policy