Sunday, August 17, 2014

TheTreasury and the Fed are at Loggerheads over QE

In my last post on QE, I quoted a paper by James Hamilton and Cynthia Wu that provides some empirical evidence for the importance of the asset composition of the Fed's balance sheet and its effect on the term structure of interest rates. They have posted their data online and it makes for interesting bedtime reading. 

Hamilton and Wu combined their data with evidence from the yield curve. They found that qualitative easing can be effective at the lower bound and that
... buying $400 billion in long-term maturities outright with newly created reserves, ... could reduce the 10-year rate by 13 basis points without raising short-term yields.
To construct these estimates, they used a theoretical model developed by Vayanos and Vila which assumes that there are investors who have a 'preferred habitat'.

The Hamilton Wu results are important. I ran some regressions of term premiums on bond supply by maturity, using their data, and I found the same orders of magnitude in the response of interest rates that they found. But there is an interesting sub-text to their analysis discussed in Section 8 of their paper. The Fed and the Treasury have been following conflicting policies. David Beckworth on his blog in 2012 makes the same point.

Quantitative Easing took place in three phases. QE1 from 11/08 to 03/10, QE2 from 11/10 to 06/11 and QE3 which is ongoing. Along with monetary expansion, the Fed attempted to refinance its portfolio by selling at the short end and buying at the long end of the yield curve. But at the same time, the Treasury was refinancing its own portfolio. The end result was that the Treasury restructuring completely swamped any effect of Fed operations at the long end of the yield curve.  

Figure 1
In Figure 1 I have broken down the System Open Market Account (SOMA) of Fed holdings of Treasuries by maturity as a percentage of all outstanding Treasuries, using the Hamilton Wu data set. The two vertical red lines are the beginning and end of the last recession and the vertical black line marks the collapse of Lehman Brothers.

There are two takeaways from Figure 1. First, the constancy of Fed holdings by maturity in the period leading up to the recession, and second, the dramatic change in this portfolio after the collapse of Lehmann Brothers. The big increase in Fed holdings at the long end is the result of 'operation twist'.

How big a player is the Fed in the Treasury markets? Leading up to the Great Recession, the Fed held 12% of all Treasuries with a maturity of two years or less, 3% of two to five year maturities, 1.5% of five to ten year maturities and 2% of maturities from ten to thirty years. Once the recession hit, Fed holdings of maturities shorter than two years plummeted, and longer maturities increased. 

But Figure 1 gives a misleading picture of Fed actions in response to the crisis since it divides SOMA holdings, chosen by the Fed, by total outstanding Treasury debt. There were two important changes going on during the recession. First, the Treasury dramatically changed the way it finances its deficit, substituting two to ten year bonds for shorter maturities. And second, the proportion of bonds held by the Fed fell dramatically.
Figure 2
In Figure 2, I illustrate the importance of the first point. This figure shows the percentage of all outstanding Treasuries, by maturity, as a percentage of total outstanding Treasury debt. After the collapse of Lehmann Brothers, the percentage of short denomination bonds plummeted. In their place, the Treasury sharply increased its issuance of two to ten year bonds. It is important to note that this Figure has nothing to do with any action by the Fed. It is a consequence of decisions by the Treasury to refinance its debt at longer maturities in the low interest rate environment that followed the collapse of Lehmann Brothers.

Figure 3

Figure 1 divides SOMA holdings by total Treasury debt outstanding. In contrast, in Figure 3 I divide Fed SOMA holdings by the Fed's holding of all maturities. This figure DOES reflect decisions made by the open market committee of the Fed. The Fed's holdings of short term bonds fell from 65% of its portfolio in 2007 to 25% in 2010. In contrast, holdings of two to five year bonds increased from 17% to 30%, five to ten year bonds increased from 6% to 26% and ten to thirty year bonds went from 12% to 19%. These are all percentages of the Fed's total Treasury holdings.
Figure 4
How successful was operation twist at changing the maturity structure of Treasury securities held by the public? In Figure 4, I break down Treasuries held by the public as a fraction of total debt outstanding. This figure shows that although the Fed switched its holdings from yields of three months to two years to yields in the two to ten year range (Figure 3) this operation was swamped, after November of 2008, by Treasury operations that increased the supply of maturities in the two to ten year range (Figure 4).  The end result was that the public ended up holding more of these two to ten year bonds in 2010 than before the recession hit.

Could we have a little coordination here guys?
Footnote: The Hamilton Wu data have since been updated through January of 2011 but I haven't had time yet to update my figures using their revisions.


  1. Interesting, but I don't think these charts give you enough of a window onto the recent interaction of the Treasury and the Fed. When the charts cut off in 2010, the first round of QE was still underway and Operation Twist was still in the future (beginning in 2011). QE1 did not seem to be as tilted toward longer maturities as the Fed's later unconventional operations. See slide 20 of this presentation:

    This chart, by Stone and McCarthy, compares the to evolution of total outstanding (blue line) and privately-held (red line) Treasury debt, as measured by duration-weighted 10-year equivalents. You can see that the privately held treasurys rose nearly in parallel to total outstanding during QE1, but not so much during later operations, when the Fed steered hard towards longer maturities.

    That presentation, by the way, argued that the Treasury and the Fed were not impeding each other but in fact had entered into a kind of symbiosis in the bond market. By 2012-2013, the Treasury was selling very short paper, with median maturities of less than two years, keeping out of the Fed's way as it absorbed longer paper. The Treasury's front-loaded issuance has been obscured by increases in the average maturity of the outstanding debt, but it has been shown that increases in this measure do not necessarily indicate a pattern of longer issuance. See:

    1. Interesting stuff! Thanks for the links to the Stone and McCarthy Slides and to your own blog post on this.

      Something I still don't understand. You quote the Treasury as saying

      "WAM (weighted average maturity) extension is not due to extending WAM of new issuance. WAM extends as maturing securities are reissued as longer maturity notes/bonds."

      Surely the decision to reissue maturing securities as "longer maturity notes" is a conscious decision by Treasury Officials. Nothing stops the Treasury from changing the maturity structure of its debt as old debt is retired.

      And the sharp change in maturities that occurred in November of 2008 does not look like a passive decision to reissue retiring securities at longer horizons.

    2. Thank you very much. I agree that there was an active and sharp change in 2008 that affected the average maturity. More recently, however, the Treasury's issuance patterns have been fairly stable, and front-loaded, yet the average maturity continues to increase. I agree that their issuance policy is a conscious decision.

      My explanation of average maturity is not quite the same as the Treasury's (actually, I think it was the bond dealer's part of the TBAC presentation). It's always true that maturing bonds are reissued at longer maturities, but that's not enough to understand the dynamics of average maturity. There are three components to changes in average maturity over some period:

      1) The bonds at the beginning, that are about to mature. These have a very short remaining life, and tend to blend down the average maturity at the start of the period. By the end of the period, they are gone and their depressing effect is gone. If shorter maturities dominate the portfolio, as in the case of Treasury debt, this is a strong effect that tends to push up the average life over time.

      2) The bonds that remain outstanding over the entire period. The remaining time to maturity shortens for these bonds, tending to reduce the average maturity.

      3) New debt issued over the period. The impact of the new debt intuitively depends on how it compares with the portfolio as a whole.

      There is a "natural" rate of increase in average maturity for a front-loaded portfolio. This is counterintuitive, but it is what we would see if somehow maturities could just be paid off instead of refinanced. The Treasury's rate of increase is slower than the natural rate because it tends to issue shorter debt than the portfolio as a whole. I try to explain this here:

      Oh, and to be clear, that one slide used a Stone and McCarthy chart, by their permission, but the presentation was one of mine.

  2. I suspect that there is something wrong with the Hamilton and Wu data, which they admit comprises "rough estimates".

    Even with a back of the envelope calculation, the numbers don't look right. At the end of the 2007, the amount of outstanding US Treasury debt amounted to about $5tn ( ), while the SOMA portfolio was about $750bn in size, suggesting that on average across the curve, the SOMA held about 15% of treasuries, whereas you are reporting in Figure 1 that the SOMA held less than this even at short maturities around then.

    The obvious place to go for the SOMA holdings reports is the FRBNY website anyway. I don't understand why Hamilton and Wu were "unable to secure access to historical archives of these". I tried it ( ) just now, and I can. The reports even include the SOMA holding as a percentage of the outstanding stock. An unweighted average of this column in the spreadsheet for SOMA holdings of notes and bonds as at Jan 2nd 2008, the data I used in my blog post ( ) is 15.6%.

  3. Rebel Economist: You are misreading the first chart. As of the end of December 2007, we estimated that the Fed held 10.7% of total outstanding marketable nominal Treasury debt in the form of instruments less than 2 years, an additional 3% of the total outstanding in the form of instruments 2-5 years, an additional 1.9% in 5-10 years, and 2.0% in more than 10 years, for a total of 17.6% of the Treasury debt held by the Fed.

    Our definition of marketable nominal Treasury debt excludes TIPS. If you look at the Treasury document you linked to at you will find it reports marketable Treasury debt as of Dec 31, 2007 as $4.52 T. Of this $0.47 T is TIPS, leaving $4.05 T in marketable nominal debt. If you sum the entries for all weekly maturities for Dec 2007 from our database at you will find that it comes to exactly $4.05 T. Of this total, we calculate that the Fed held $0.43 T in nominal Treasury securities of less than 2 years duration, from which the figure 10.6% quoted above can be confirmed.

    Thanks for noting the link to At the time our research was completed, the ability to download historical Fed holdings by CUSIP was not available at this or any other site. And in point of fact, I personally contacted a half-dozen people within the Federal Reserve Bank of New York (some of whom I know pretty well) requesting access to these data when we were conducting the research. I was told it could not be obtained. Possibly my inquiries and our research prompted the additional features to now be provided by the site. But note in any case that our data set goes back to 1990, whereas the FRB NY is still only reporting the numbers back to 2003.

    Instead we were able to obtain Fed holdings by broad maturity categories for every month going back to 1990, imputed these to individual CUSIP assuming proportional holdings within categories. But this individual imputation would have zero effect once the weekly holdings are reaggregated, as Roger has done, back into broad maturity categories. I'm quite convinced that if you did attempt to redo these calculations at the individual CUSIP level, your graphs would look very similar, if not identical, to Roger's above, if you did the calculations correctly.

    James Hamilton

  4. Thanks for responding James. I was assuming that you were trying to replicate the SOMA report and its percentages further back than you could obtain the FRBNY reports. I now understand the graphs. My problem is really with Roger's "stylised" Fed balance sheet in the post where this came up: in which the Fed were supposed to hold no long-term bonds before the financial crisis. And I think the point I made there stands, which is that during the pre-crisis period, when the Fed was lamenting the low long-term interest rates as caused by a "saving glut" beyond their control, they were themselves holding about half their SOMA portfolio or $300-400bn in bonds longer than two years. At the time, notwithstanding publications like Warnock and Warnock presenting evidence that foreign reserves managers' purchase of treasuries was driving long rates lower, the Fed would cite the Woodford doctrine to argue that their holdings did not matter, but soon changed their tune during the financial crisis. Since Volcker, the Fed have always proved more imaginative in finding arguments and techniques for easing than they have for tightening, which if you ask me, is a big reason for the financial crisis and their continuing difficulties.

    Tim Young


Note: Only a member of this blog may post a comment.