It was over two months ago that Zero Hedge first described why we though QE3 would ultimately appear in the form a reincarnation of Operation Twist first utilized by the Fed in the 1960s to prevent the gold exodus from the US into Europe (read more here and here), also known as Operation Twist 2. In essence what the Fed would do would be a curve patterning exercise in which the Fed would lower long-term rates by changing the average maturity of Fed holdings, in the process removing substantial duration from the markets, once again pushing investors into far more risky assets such as stocks (but certainly not gold: the CME will see to that). Today, Morgan Stanley's Jim Caron (who has yet to be right about one thing in his prior 3 year forecasts so take this with a grain of salt) explains why Morgan Stanley is a big supporter (read lobbying heavily on behalf of) of Operation Twist 2. Quote Caron: "As outlined in the recent congressional testimony, the Fed could consider several ways to ease financial conditions further. One of the options mentioned by Fed Chairman Bernanke was to provide explicit language to keep the fed funds rate and the Fed’s balance sheet unchanged for an extended period. Another approach would be to keep the balance sheet unchanged but increase the average maturity of its holdings. If this path is chosen, we believe that a significant amount of duration could be removed from the markets – to the effect of $90-150bn 10y equivalents which could lower 10y yields by 20-35bps. Let us explain." Explain away Jim.
A modified ‘operation twist’. In the past the Fed has employed operation twist (in the 1960s) with the goal of raising short-term rates in order to support the dollar while keeping long end rates lower to support the economy. Today, we think the Fed may use a modified version of operation twist with the goal of lowering rates for the entire term structure with particular focus on longer term rates. They could use methods such as lowering interest on reserves that would have the effect of keeping short term rates lower while also increasing SOMA portfolio duration and keeping long end rates low.
In order to put the $90-$150Bn 10y equivalents size into context, the current SOMA portfolio has a total duration of about $990bn 10y equivalents, and the tradable coupon Treasury market has $3tr 10y equivalents. Using SOMA duration extension, the Fed could effectively take duration out of the market equal to about 15% of the current SOMA portfolio – this is over 30% of the reduction in duration that QE2 achieved.
Evaluating the impact on yields from a SOMA duration extension. We can use a framework borrowed from the mortgage market to evaluate what impact a reduction in duration in the SOMA portfolio would have on 10y yields. As we know, Mortgages are a negatively convex instrument, meaning that, unlike traditional bonds, investors accumulate losses at an increasing rate when interest rates rise, and accumulate gains at a decreasing rate when they fall. When mortgage portfolio managers hedge their holdings with Treasury bonds, they must make up for this difference in bond characteristics by buying additional bonds when rates fall and selling bonds as rates rise. The total market effect is that duration is either added or subtracted from the market, and the total amount is due to not only interest rate factors but also factors specific to the mortgage market.
We can turn this model on its head and instead solve for a yield move relative to a change in duration, as we do in the case of a possible increase in the duration of the SOMA portfolio. The situation is comparable to mortgage hedgers buying in a rally. According to our mortgage model, a market outflow of $90bn 10y equivalents from hedging is currently commensurate with a 20-25bp decrease in the 10y yield. In the more extreme case of an outflow of $150bn 10y equivalents, the implied 10y yield decrease is 30-35bp.
SOMA duration-extension scenarios. The Fed currently holds $2.6T of Treasury, Agency and Mortgage bonds in its SOMA portfolio (Exhibit 1). Of the $1.54Tr in Treasuries, a majority of the holdings are in the belly of the curve (Exhibit 2). In terms of duration, the belly and long-end sectors all have approximately equal amounts at about $200bn of 10y equivalents in each of the 4-6y, 6-8y, 8-15y, and 15-30y sectors. Of the $1.54Tr in SOMA par value, the total portfolio duration is approximately equal to $992bn 10y equivalents.
There are plenty of available short-dated securities that the Fed could use to extend duration in the SOMA. Coupon bonds with 2.5y to maturity and less in the SOMA total about $390bn. Adhering to the current Fed purchase regulations, there is approximately $350bn of par available in the 4-6y and higher sectors.
Although it is highly unlikely that the Fed would purchase all USTs available to it with greater than 4y to maturity, we nevertheless use this scenario as an upper bound on the amount of SOMA extension and thus duration that can come out of the market. We choose two other, more realistic scenarios, with the Fed using $100bn and $200bn of short-term notes to fund longer-dated purchases. In each of these latter two scenarios, we attempt to spread the duration evenly across the back end of the curve in order to match the current SOMA duration profile (Exhibits 3 and 4).
Scenario 1: $100bn of short-duration par extended. In the first and most conservative of our three scenarios, we assume that the Fed sells $100bn of short-end par amount, and extends into longer-duration debt. Specifically, we assume that the Fed takes the total extension from this sector ($191bn is available), and allocates proceeds by 35% into 6-8y, 47% into 8-15y, and 14% into 15-30y.
This extension takes the maximum eligible remaining amount from the 6-8y sector and overweights the 8-15y sector versus the 15-30y to compensate. This operation removes a total of about $90bn in 10y UST duration equivalents from the market.
Scenario 2: $200bn of short-duration par extended. Doubling our scenario 1 extension par amount provides a significantly more aggressive impact on market duration. We assume that the Fed sells all of the 0-1.5y sector and 4% of the 1.5-2.5y sector and reallocates the proceeds 40% into 4-6y, 17% into 6-8y, 30% into 8-15y, and 13% into 15-30y.
Again, all eligible bonds are purchased in the 6-8y sector, and we slightly overweight the 8-15y sector in duration. In this scenario, we do target a lot of bonds in the 4-6y part of the curve, which provides a more even duration profile, but does not maximize extension. Nevertheless, we believe that this more closely represents what the Fed would actually do. Total duration removed from the market is $150bn in 10y equivalents.
Scenario 3: $350bn of short-duration par extended. This scenario is meant to represent the maximum meaningful duration extension that could be accomplished – an upper bound on the SOMA extension. With the current Fed purchase limits, any additional short-duration bonds sold above the $350bn would simply be re-allocated into the 4y and under sectors. $359bn in proceeds are allocated 30% to the 4-6y, 10% to the 6-8y, 30% to the 8-15y, and 30% to the 15-30y sectors. Taking out all the available 30y USTs, this scenario drastically overweights the 30y sector. At $200bn in 10y equivalents, the 15-30y sector duration removed doubles the second-highest sector, the 8-15y.
So there you have it: it's coming. And it's almost here.