Unraveling the mysteries of the Treasury yield curve

If you cannot tell from the title, this is going to be a rather wonky post.

Nouriel and I are working on a paper on "Bretton Woods Two," so I have been delving into some data on the Treasury market. No surprise: the goal is to see what we can learn about foreign central bank buying of Treasuries at different points on the yield curve, and thus try to figure out whether the "central bank bid" is one reason why the 10 year bond ended the year at 4.25% -- just about where it started -- even though the Fed tightened and inflation picked up.

Since the supporting evidence is long, detailed and bit dull, I’ll lead with my conclusions.

a) Foreign central banks have been a significant source of support at the middle and long-end of the Treasury curve, not just the short-end.

b) Growing foreign holdings of marketable treasuries absorbed 80% of the increase in the outstanding stock of marketable treasuries in 2003, and over 100% of the increase in the first three quarters of 2004. In other words, foreigners could have bought all the new issuance of Treasuries needed to fund the budget deficit, and still had $50 billion or so left over -- money they used to buy existing Treasuries that started the year in American hands. US holdings of Treasuries went down, despite the increase in the overall stock of Treasuries.

c) Since dollar reserve accumulation exceeds recorded central bank buying of US assets, it is likely that some private foreign buying of Treasuries is disguised central bank buying. If that is not happening, then the growing dollar reserves of foreign central banks are necessarily lending support to other US fixed income markets.

d) There is good reason to think that the pace of reserve accumulation this year by the world’s central banks exceeded the pace of reserve accumulation last year. All signs suggest that emerging Asia and Japan added to their reserves at a slightly faster clip than in 2003, despite having to pay more for oil and other commodities. The pace of reserve accumulation by commodity exporters -- both formal reserve accumulation and informal reserve accumulation, as oil funds are a de facto form of reserves (though they typically are invested in a broader range of assets) -- no doubt accelerated.

Some terminology. Treasury bills are very short-term Treasury debt; they have a maturity of a year or less. Treasury notes are Treasuries with a (original) maturity of between two and ten years. The Treasury used to issue Treasury bonds, debt with a maturity of more than ten years, but it no longer does. The stock of outstanding bonds is slowly shrinking, as old bonds mature and no new bonds are issued.

To get a sense of how actively foreign central banks have been supporting the Treasury market, I compared data on the annual increase in foreign central bank holdings -- which are broken down into holdings of bills (short-term debt) and holdings of bonds and notes in the TIC data set -- with data on the Treasury’s annual net issuance of Treasury bills, Treasury notes and Treasury bonds, data that the Treasury publishes here.

There are some limitations to this methodology. We know that the Treasury data on official holdings likely understates central bank participation in the market. US data on official inflows does not match BIS data on the annual accumulation of dollar reserves, and the BIS data almost certainly provides a more accurate indicator of overall central bank financing of the US (though not all central bank dollar reserves are invested in treasuries; some central banks like agencies, for example). So this methodology would tend to understate central bank support for the market. Since the treasury data probably does a better job of tracking sales to foreigners than tracking which foreigners really are doing the buying, I also compared the increase in total foreign holdings of treasuries to the increase in treasuries outstanding.

And, of course, central banks do not have to buy new issues, they can also buy old issues in the secondary market. Still, I think comparing the increase in central bank holdings with the increase in treasury debt outstanding provides a reasonable way of getting at central bank support at the margin. Finally, I am using changes in stocks to estimate flows -- something that only works if Treasury bond prices are relatively stable (they generally were in 2004).

Here is what the data tells us:

In 2003, total Treasury issuance of bills and notes (net of redemptions of long-term bonds) was $366.7 billion. Net issuance of notes and bonds was 326.7 billion (Heavily concentrated in the 2 to under 5 years category. The treasury was aggressively reducing the average maturity of US debt, the average maturity of new issuance in 2003 was under 2.5 years, down from about 5 years in 2000); net issuance of bills was $40 billion The reported increase in central bank holdings of treasuries was enough to have financed 42.5% of the overall increase in marketable Treasuries, and 46% of the increase in Treasury notes.

This data no doubt understates central bank support for the Treasury market. The reported increase in foreign central bank holdings of treasuries was $171 billion, not a small sum, but far less than the $441 billion increase in dollar reserves reported by the BIS. I sincerely doubt that central banks invested $270 billion in other fixed income markets, and only $171 billion in treasuries. So let’s assume that a substantial fraction of the overall increase in foreign holdings of treasuries -- $289.6 billion -- represents disguised central bank purchases. Total foreign purchases of treasuries accounted for 78% of the increase in marketable treasuries in 2003.

Two additional notes: First, China gave away $45 billion in reserves in 2003 to recapitalize two state banks, reducing its reported reserves. The total increase in dollar reserves was probably closer to $480 billion than $441 billion. Second, foreign central banks clearly were buying something other than treasuries. Even if the entire $290 billion increase in foreign holdings of treasuries came from central banks, they still had at least another $150 billion to invest in other dollar assets.

What happened in 2004? We don’t have full data, but we do have data on Treasury issuance and the increase in central bank holdings for the first three quarters.

During the first three quarters of 2004, the treasury’s net issuance of new marketable securities totaled $266.2 billion, and its net issuance of notes and bonds totaled $233.3 billion (with more far notes in the 5-10 year range than in 2003 and far less in the 2 to under 5 range; the average maturity of new issuance this year is creeping back toward 3 years). The reported increase in overall central bank holdings of Treasuries over this time was enough to have bought 68% of the treasury’s total net issuance, and the reported in increase in holdings of bonds and notes was enough to have bought 69% of total note issuance.

So central banks lent more support to the Treasury market in 2004 than they did in 2003, at least looking at recorded central bank inflows. Obviously, Japan’s burst of intervention at the tail end of the 2003 and in early 2004 had something to do with this; some of the dollars Japan bought in 2003 were probably not invested in Treasuries until 2004.

The total increase in foreign holdings of Treasuries -- $315.6 billion during the first three quarters of the year -- was about $50 billion more than the $266.2 billion in new debt the Treasury issued.

Any wonder why large deficits have not had much of an impact on US borrowing rates?

The Treasury’s net new issuance of marketable treasuries in 2004 is likely to be about $393 billion, based on their published estimates for q4 borrowing.

Asian central banks -- based on my estimates, which include an assumption that China’s strong reserve accumulation continued in November and December -- increased their 2004 reserves by about $515.5 billion this year, a bit more than in 2003 ($465.2), though if you include the extra $45 billion in China’s 2003 total, the numbers are almost identical.

Some of the overall increase, in dollar terms, comes from appreciation of central banks euro assets. If 20% of emerging Asian central and Japan’s stock of reserves were in euros, they would have started the year holding $375 billion in euros (or for emerging Asia, a mix of euros and yen). Assuming 10% y/y euro appreciation, the real number was 8%, but there is something to be said for making the math easy, the "valuation increase" was roughly $40 billion. If Asia central banks had an "average" central bank portfolio with 70% dollars and 30% euros and other currencies, they would have started the year with about $575 billion euros, and maybe enjoyed a valuation gain of $60 billion. That is too high -- several Asian economies peg to the dollar and therefore would tend to hold most of their reserves in dollars (Hong Kong, for example) -- and several central and eastern European countries peg to the euro and therefore would tend to hold more euros. But to be safe, I split the difference and assume valuation gains of $50 billion.

That puts total purchases of new reserves at $465 billion; assume 80% went into dollars, and Asian dollar reserves increased by about $375 billion -- almost enough to have bought the total $393 increase in the stock of outstanding marketable treasuries.

Obviously, not all Asian dollar reserves went into the treasury market. At the same time, Asia was not the only region of the world building reserves this year. Oil and commodity producers saw their reserves go up too. Russia’s reserves are up $44 billion this year, and even if over 50% of its end 2003 reserves were in euro, changes in the euro/dollar only explain $4 billion of the increase.

All in all, I feel pretty safe predicting that global reserve accumulation picked up in 2004 from its already extraordinarily high pace in 2003, and that dollar reserve accumulation picked up as well. Total worldwide dollar reserve accumulation almost certainly exceeded the increase in stock of marketable treasuries by a substantial margin.

Combine that stylized fact with the data indicating that recorded purchases of Treasury notes -- Treasuries with a maturity of more than 2 years -- picked up relative to 2003, and I am pretty sure that central banks lent support to most of the yield curve. This data does not indicate whether foreign central banks are lending more support to the two year, the five year or the ten year bond, but they are lending so much overall support that it would be an enormous surprise if they were not lending substantial support to the ten year.

Comments on my argument are most appreciated -- particularly from those in the markets ...