And the full report:
What Happens After the Fed Stops Buying?
The Fed’s QE2 program could be keeping intermediate maturity US Treasury yields around 40-50bp lower than would otherwise be the case. The effects of the central bank’s purchase are most visible in the 5-yr sector of the yield curve, which still looks ‘rich’ relative to adjacent maturities. When the flow of purchases ends in June, and on the assumption that the stock of bonds held by the Fed does not change, there should be little immediate effect on bond yields. Nonetheless, the ‘term premium’ should gradually rise as the Fed’s bond holdings passively run off as securities mature. Moreover, expectations of the Fed selling securities back into the market—which is not our baseline case but could admittedly pick up as the economy continues to recover—could amplify the increase in yields we expect based on our macro forecasts.
Fed’s Bond Purchases Soon Drawing to an End
Through its asset purchase program (‘QE2’), the Fed has delivered a stimulus to the economy by-passing the nominal zero policy rate constraint and influencing directly longer-dated discount factors. In the December issue of our Fixed Income Monthly, we estimated the Fed had broadly succeeded in bringing intermediate and long Treasury yields close to a level consistent with negative policy rates as implied by a ‘Taylor Rule’.
As the purchase program draws to an end in June, we are frequently asked whether there will be an adverse impact on the bond market. We tackle this issue with an ad hoc regression analysis, cross-checking our findings through the suite of valuation tools we regularly employ in the formulation of bond strategy.
Our main conclusion is that the announcement of the total size of Treasury purchases, rather than their implementation, could have lowered intermediate maturity Treasury yields by as much as 40-50bp. As long as the Fed does not announce its intention to sell bonds back into the market any time soon (which continues to be our baseline case), this effect is likely to fade only gradually, assuming no intervening change in the macro landscape.
Our central forecasts, however, are consistent with a progressive increase in real bond yields over coming quarters. And, as the recovery takes hold, it is conceivable that market participants’ expectation of asset sales could increase, thus amplifying the sell-off in yields. With this in mind, we continue to recommend short positions in the 5-yr area of the Treasury curve, which looks to have been the most influenced by the Fed’s interventions.
The ‘Announcement Effect’
Assuming financial markets are liquid and forward-looking, bond prices should be affected by the announced total amount of purchases the central bank intends to conduct, rather than the subsequent flow of purchases. Chairman Bernanke has been among the proponents of this approach, which researchers often refer to as the ‘portfolio balance channel’ or the ‘stock view’.
If this theory is correct, when the flow of purchases is discontinued there should be little effect on yields provided expectations are that the Fed will not sell securities back into the market any time soon.
Empirical evidence appears to support the notion that the ‘stock’ effect dominates the ‘flow’. For example, 10-year US Treasury yields fell sharply following the surprise announcement of the ‘QE1’ program on November 25, 2008 and March 18, 2009. However, there is little evidence that yields increased following the termination of those purchases at the end of October 2009 (when the Fed stopped buying Treasuries) and March 2010 (the end of the mortgage-backed security purchase program).
The chart at the bottom of the previous page compares actual 10-yr US Treasury yields with the ‘fair value’ implied by our Bond Sudoku model. The latter describes US bond yields as a function of 1-yr-ahead consensus expectations on short rates, real GDP growth and CPI inflation, both domestically and in the other major advanced economies. The effects of the asset purchase program (or shifts in the net supply of government bonds) are captured by the model only indirectly, i.e., to the extent that these influence expectations on the future course of the relevant macro factors.
As can be seen, ‘QE1’ led to a fairly rapid move in bond yields right on the announcement date. In the case of ‘QE2’, the effect largely preceded the FOMC meeting in which the decision was taken. This can be attributed to the fact that, in a number of speeches through the Summer, Fed officials had hinted at a resumption of bond purchases to stimulate the economy.
By the time the Fed announced the intention to further expand its balance sheet on November 3, 2010, 10-yr government bonds were already trading around 1 standard deviation (or roughly 40bp) below their macro equilibrium. Our GS Curve model—which links the term structure of constant maturity Treasury yields to consensus macro expectations at different horizons—indicates that around the same period bond yields in the 5-to-7-yr maturity range (the main target area of purchases) stood at very depressed levels relative to their historical relation with the 2-yr and 10-yr sectors (see chart below).
Once again, most of the effect precedes the actual decision to conduct asset purchases, but expectations of such an outcome had been building ahead of the FOMC meeting. A parallel can be drawn to the Fed’s decision to cut policy rates to 1% on June 25, 2003. Intermediate maturity bonds rallied strongly in the 3 months before the policy meeting, only to sell off aggressively after the event.
Fed Holdings Keep Bond Premium Lower
In order to test the empirical validity of the ‘stock view’ more formally, in previous research Jari Stehn ran a regression between the nominal 10-year US Treasury yield against four factors: the stock of announced purchases, the actual weekly flow of these purchases, a number of economic variables (including payrolls, the ISM survey and the University of Michigan/Reuters 5-10 year inflation expectations) and measures of the Fed’s other unconventional monetary policies (such as its guidance that it would keep interest rates “exceptionally low for an extended period”). The results, summarised in the first column of the table above, indicate that the effect of the announced stock of purchases is negative and statistically highly significant.
One issue with this simple specification is that the coefficient on the flow of purchases takes the ‘wrong’ sign, suggesting that the flow of purchases raised bond yields. This counterintuitive finding has a simple explanation: just about as the Fed started to purchase assets last November, bond yields rose sharply because growth expectations improved and, partly as a result of this, market participants revised down their expectations of further easing. But because the statistical exercise controls for the contemporaneous rather than the expected macroeconomic landscape, the increase in yields is attributed to the flow of QE2 purchases.
To address this shortcoming, rather than focusing just on the 10-year yield, we explore how the Fed’s purchase program has affected the yield curve across maturities. This allows a differentiation between the impact of economic factors (which affect the entire term structure) and the Fed purchases (which could differ by maturity bucket). Making use of the relative movement of yields at different maturities provides more information and should therefore provide better identification.
The Box above outlines the approach we have taken and the main results are summarised in the second column of the table on the previous page. Once again we find a significantly negative and economically meaningful effect from the stock of purchases on the 2-10-year part of the yield curve, while the coefficient associated with flows is now insignificant. Specifically, the estimates suggest that yields in the 2-10-year maturity range have been reduced by around half a basis point for each US$1bn of announced purchases. This suggests that the Fed’s Treasury holdings could be currently holding down 10-year yields to the tune of 40-50bp. This numerical result is clearly subject to the usual caveats applicable to the outcome of statistical analysis, but is reassuringly not far from what other studies have found. In addition, similar regression analysis on 10-year yields for the UK also found a significant and meaningful effect from the stock of purchases but not from the flow of purchases (for more details, see “A Modest Impact on Markets from the End of QE2” Global Economics Weekly 11/14).
To Sell, Or Not to Sell?
The empirical analysis reviewed so far allows us to draw the following conclusions for bond strategy: