Bernanke’s (new) Conundrum – Negative Convexity

Bruce Krasting's picture

The Atlanta Fed put out a report
on the status of the Fed's purchases of MBS. The report confirms that
91% of the anticipated $1.25 Trillion of paper has been bought. This
leaves about $110b of buying power left for the Fed. There is only nine
weeks left until the anticipated time that this program will end. This
implies an average of only $10b of intervention per week.
The most recent purchase was for $16B. Look for that weekly number to
fall pretty quickly from now on.

The following graph clearly shows the STEADY accumulation of Agency paper.

Now look at the following graph. If you print this out and check with a
ruler (I did) you will see that the lowest point on the brown shaded
area is 1,200 and the upper band is at 2,400 (1,200 total). The legend
states that brown are both Agency Bonds and MBS. From the report you
get those numbers to be Bonds = $175B and MBS = $1.14T, for a total of
$1.31T. Significantly higher ($110b) than you might have expected
looking at the graph. There is a simple reason for the apparent
discrepancy. It is called Negative Convexity.

A portion of all long term mortgages pre-pay prior to the stated
maturity. There are many factors that influence this. People die, get
divorced, send the kids to college and downsize. But the most
significant factor in the amount of prepaid mortgages is prevailing
interest rates. If rates are low on a historical basis, people who have
good credit will refinance and achieve a lower monthly mortgage cost.
These are the conditions that we are in today and have been for the
past year. The MBS securities that the Fed has purchased are very
diverse. They include primarily older, higher coupon mortgages. Ones
more likely to prepay.

I asked a friend who does this type of thinking for a living to give me
some thoughts on how quickly the Feds portfolio was shrinking due to
natural prepayments that are occurring. I think his words are better
than mine:

"Tough to say exactly how much
paid down, but if we say GSEs have paid at 25% cpr in the past year
(all pay differently and are well correlated to coupon, or WAC of
borrower), then that implies 25% of principal is paid down over a year.
You’re pushing 20% over 9mos…

typical speeds are in the low
teens, if not high-single digits, except in times of high re-fi /
default activity… I’d say you could feel good about using a 20-30% cpr
weighted avg…. of course, they did not buy ALL their MBS 9mos ago –
would have to weight that accordingly.

off the cuff, I’d say 10% principal back to them would be a good number….

From this professional you get a pretty good estimate of the prepay as being 10%.
That would come to $110B. This estimate goes a long way toward
explaining the discrepancy between what the Fed has purchased and what
the principal balance is that they currently own.

Some thoughts on this phenomenon:

-My friend suggests that going forward the prepay could be as much as
20% PA. Well that would mean something in the order of $250B over the
next year. That would, by itself, be a very deflationary force. It is
too big a number. It would be happening at a very bad time. Pure
economics would suggest that the supply of available mortgage credit
would fall sharply as a result. The Fed does not need to do repo’s to
suck up excess reserves. They just have to collect the prepays that are

-If you buy into this you have to assume that the amount of prepays in
the current month will be approximately $18b (1.1T * 20% / 12). The Fed
is buying $16b a week or $64b a month. So in January the net is only
$46B. Follow this dotted line and you will see that by early March the
purchases net of prepays will be a negative number. This will be the
starting date of the true reversal of the QE process. March is much
sooner than people are thinking it will occur.

-The Fed will make Net purchases totaling $1.25T. But they will never
have a portfolio of that amount. It has to be less. By the numbers they
will end March with approximately 1.14 - 1.16 Trillion. And the
portfolio will be shrinking by $20B per month thereafter.

This is a scary thought. This could well be the basis of a back door,
Sneaky Pete “QE 2.0 Lite”. If the intention were to purchase and
maintain a portfolio size of $1.25T they would have to make additional
purchases of $100B and continue the buys on a monthly basis of
approximately $20B. This would not be a change of policy (ahem). It
would be refining and maintaining the existing policy. If the Fed tried
to do that (I doubt they will) there would be uproar. But once Bernanke
is back “in”, there really is no way to get him or his love of QE “out”.

The following is the language from the Fed on the Agency MBS purchase program. I draw your attention to two words. Total and Anticipating. I think there is wiggle room in this statement on some revisions to both the timing and amounts involved. Ugh!

“On September 23, 2009, the FOMC announced that the Federal Reserve will purchase a total of $1.25 trillion of agency mortgage-backed securities and gradually slow the pace of these purchases, anticipating they will be executed by the end of the first quarter of 2010.”

Side Story – We Get Humped

Please look again at the graph of assets from the previous discussion.
Look at the brown section that represents the Fed’s holdings of BOTH
Agency MBS and Agency Bonds. That line is not a steadily increasing
number. It seems to go up and down quite a bit from the inception of
the program to the end of chart.

A major reason for this is how the Fed conducts its activities and how
they report them. The Fed, on a weekly basis has accepted tenders for
MBS. Of the amount tendered they report two numbers, Gross and Net

The Gross number is the one recorded in this graph and accounts for a
substantial amount of the variable nature of that data. Most notable on
the graph are the two obvious ‘humps and valleys’ in the March - April
period. Behind these humps is this data. From March 19 – April 2 the
Fed bought gross $196B, sold forward $94B and retained $96B. From April
16 – April 30 the Fed bought gross $175B, sold forward $100B and
retained $75B.

As you can see the Humps are unique to the graph. There is a reason.
March was a terrible month. Global GDP was in freefall. The stock
markets were again hitting decade lows. We thought that Citibank and
BoA were maybe going out of business. Bernanke has admitted that during
this period he questioned whether the system would survive.

So anything that the Fed did during that terrible period that is out of
trend is worth noting. What is the impact of these forward sales? The
answer lies below the brown area. Look at the grey and dark blue
levels. You see the reverse Humps. The forward roll transactions had
the impact of temporarily absorbing the inventory of the primary
dealers and the money center banks. It had the impact of creating
liquidity for the PDs so that they could absorb some of the other
things that were falling in price. (Stocks and bonds). The amounts were
bite sized. My estimate is that these dollar rolls put $100b in PD
hands for about six weeks. My guess is that the street got fat on the

Does this matter? Not really. That they Fed put some dough in the hands
of the street when they needed it was probably the right strategy at
the time. It was one of the many steps taken back then to calm things
down. Any impact that it had, is long since gone.

I think this is just another of those ways that the Fed can maneuver
things. They have extraordinary power. We don’t even understand what
they can do. The Fed has full disclosure policy on most of what it
does. They provide information on the forward rolls. Take a look and
see if their explanation makes sense. To me, it as clear as mud.

"As the investment managers (NY Fed) conduct dollar rolls they
simultaneously buy and sell agency MBS securities for different forward
settlement dates. To date all sales for the SOMA have been associated
with dollar roll transactions. These transactions have not represented
any outright sales of agency MBS from the SOMA."

Yeah Ben. But why the big numbers in March? Answer: When the market needed a boost, you delivered.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
dead hobo's picture

I think you are describing the Ponzi scheme of the 2010 decade.

Assume the Fed is engaging in reciprocal MBS purchases such that the party who sells the MBS to the Fed must purchase UST debt. Thus, The Fed is monetizing the deficit. So far, this is nothing new. It has been commented on many times in the past.

Now, assume a banker's viewpoint. If you print money to purchase MBS debt paying a higher rate than the UST debt being sold in reciprocity, a banker would think he is making money on the spread.Interest income from MBS revenue is likely higher than the cost of UST debt offsetting it. Profits are given to the UST to cover the underlying UST interest expense and pay for government operations.

The point being ignored is that UST debt is basically a perpetuity since it is constantly refinanced. The MBS debt will be paid off within a few years. 

Therefore ,this scheme is just a cynical plan to force the next generation to pay interest and principal from today's financial shenanigans. But wait, there's more.

Thinking ahead, I suspect the Fed will see this operation as a money making plan to subsidize both housing and the deficit. Of course there will be a QE2 as it relates to subsidized MBS. Printed money will subsidize housing and the deficit as long as sovereign and other buyers capable of massive purchases of MBS debt purchases are available.

Potentially, this will be the Ponzi scheme of the 2010 decade. In order to make it work, housing must be revitalized such that a ready supply of MBS debt is available for Fed purchases using printed money. The UST will always have a mountain of debt for sale to those who sold the MBS to the Fed. The Fed will profit on the spread and return those profits to the UST for debt service costs plus profit. Long term rates will likely remain low as long as money flows through this system. Housing prices will revitalize. The cycle will repeat.

ZH, regarding your question of "Who is the secret bidder?" It might be the Fed, indirectly, using printed money and MBS purchases. The Chinese stated there is not enough money in the world to finance the US debt planned for the coming few years. I disagree. There is an unlimited amount of cash available providing that housing can be stimulated and normalized and the mortgages relating to this expansion are securitized and sold. The Fed and its printing press will be the backstop of eveything in that economy. MBS will be the laundering tool of choice for this printed cash.

If QE2 is announced and housing is again appropriately 'stimulated' by fiscal policy, I suspect this will be the start of the next bubble.




Anonymous's picture

Why would pre-pays rise in what is a rising interest rate environment? The author of this article has it completely backwards; 30-year rates are going higher (as evidenced by the rising yield on 30-year treasuries), which means that, rationally, people would rationally *not* refinance, nor prepay.

What might appear to be a security with an empirical historical duration of 3-5 years, may very well end up being a security with a duration of 10-15 years, which, of course, has significantly greater interest rate risk than previously experienced.

If you were to take the first chart...instead of being flatter, it should actually steepen because, as the Fed buys more MBS (with money conjured out of thin air), they're debasing the dollar, which definitely pushes interest rates up.

Higher interest rates are, of course, asset deflationary, but consumer price inflationary.

Bruce Krasting's picture

There are always prepays. The base level is in the high single digits. It is twice that now, primarily due to low fixed rate mortgages. The most recent FNM 30 year was a tad over 5% That is still pretty much an all time low. So as of today the beat for ReFis goes on and the negative convexity continues. If rates rise this will abate. But rates have not risen yet, and they won't for a good while to come.

Another factor in the prepay calculation is the default/extend and pretend stuff. These also result in prepays. There is a ton of that going around today. Check the RealtyTrac numbers out this AM. Defaults are everywhere in the country at this point. It just adds to the convexity problem.



Anonymous's picture

Sams Club - 11000
Verizon - 13000
Home Depot - 1000
UPS - 2800

Ford - +1200

Recovery going very well!! Thank you BB

Eternal Student's picture

Excellent article. Thank you. I agree that the Fed has extraordinary power. And that we don't understand what they are capable of doing. But it all boils down to a simple question: Can this extraordinary power keep alive the biggest Ponzi scheme in human history? That it helped to create?

I think not. If I'm wrong, it would be the first time that such a thing had ever happened.

Place your bets accordingly.

Anonymous's picture

The only MBS i demand is Mature Bald Snatch. you know, like the kind on a 24-year-old be-atch

Anonymous's picture

Good discussion going on here. I have a few questions.

Does the fed hedge their convexity? Assuming they do not, now that they are the largest holder of MBS out there, will the effects of duration hedging be muted due to the fact that the FED is not dynamically hedging their portfolio?

Also, along the lines of SDR2's post. When the prepays and the coupon payments come in, where do they go? By definition, that money is coming out of the system. It all depends on what happens to it after. Do they repo it out? If they "burn" it that would be deflationary, correct? This is different that coupon payments on their treasuries. Because that is government money paying off the government so the effect is a wash (i think, if not please correct me).

All of this is a big deal, because it deals with liquidity which is the driving force behind the rally in risk. Take it away, and you take away the rally (IMO)

Bruce Krasting's picture

When the Fed gets a $1 of prepay it reduces their balance sheet. That is a good thing at one level, but bad in another. Their huge balance sheet is the IV that has been keeping us going. When that ends it will have a consequence depending on what other steps are taken.

Simple question: is the prepay phenomenon pro cyclical or anti? It is an anti in my view.

rawsienna's picture

The Fed has needed to conduct dollar rolls in the 5 and 5.5 coupons because they have overbought and distorted the market.  It has been done out of necessity (there is not enough "TBA" float for dealers to make delivery) and despite having to roll 30-40bb a month, there is still massive fails in the system. Also, when they roll, there is no paydown, no reduction in principal.  Regarding paydowns, the number is probably closer to 10-12% annually on their settled position. At current levels of rates, those speeds should go up.

 There is plenty of demand for agency MBS away from the FED from banks due to the lack of C+I loan demand so dont expect a massive spread widening at current rate levels or if long rates back up. It will only happen in a rally due to supply from refi's or when/if the FED needs to raise short term rates and people take off carry trades

SDRII's picture

th outright assumption that prepays is deflationary misses the point. First why is it assumed that there isn;t adverse selection going on with the pools/MBS? If indeed this is the case, the PSA prob is trumped by the delinquency rates. Of course it is one big circle J with the GSE anyway. Does it really make a lot of sense given the stated goal of the Fed to be buying TBAs?

But that aside, if a refi occurs on a non performer (or something likely to go nonperformer) and the refi comes off at a lower rate (the reason for the higher pSA in the first place) then the spending multiplier may indeed be expansionary. Isn't that the whole point of driving the refi boom? Also look at the mortgage refi fee income of the banks over the past several quarters - it has been the second great coming after FICC trading.


As for the roll, perhaps you are correct in the timing, but it would be a better exercise to figure out what was flowing through the incredibly robust ficc lines of all the major investment banks. The wider spreads since LEH went away doesn;t pas the smell test. The government washed money into the brokers and banks via the capital markets business and it is no surprise the opaque fixed income line was ussd.

I have been asking the question for a while what the Fed is doing with the P&I on the performing mortgages monthly? Are they burning it?

Anonymous's picture

didnt they report a profit to the treasury based on interest *they have a special treasury supplemental account). The principal they must be burning.

Bruce Krasting's picture

Yes $ rolls are common, and the Fed has been doing them for awhile. If you are able to make this observation you know the market and how it works. So you tell me, was the magnitued of the forward roll in the period I mentioned normal? Has there been a period since where the magnitude of the roll comes close to that it was back in March?

March was an outlyer month. It stands out. It was not a coincidence. The amount done (and the pricing that generated it) was managed.


Hephasteus's picture

Sometimes it's not the magnitude of the roll but the ricketyness.

Looking like we might be having less vocal pimco pimphand problems.

Doc Brown's picture

Your sentence summed it up "they have

extraordinary power. We don't even understand what they can do." If you believe that why are you trying to guess

now? Searching for some smoking gun?

Stick with the AIG story probably will yield better results.

Mr Lennon Hendrix's picture

agency mbs and agency bonds looked like they only increased.  this article contains great journalism.  

I must add however, the system did not survive.  this is it's corpse.  weekend at bernie's.

ghostfaceinvestah's picture

Dollar rolls are common in the MBS market, you are confusing their balance sheet which are settlements with the purchase committments which are sometimes rolled.