This document discusses prepayment speeds (prepay speeds) for ARM mortgages from recent data. It provides prepay speed data for agency ARMs produced from September 2014 to present, with speeds by coupon, WAM, average prepayment, and price. For portfolio hedgers, it shows projections for seasoned Freddie Mac ARMs indicating speeds may be picking up. The document encourages readers to use available data to establish rational prepay estimates rather than relying solely on others' opinions, and offers the author's group's assistance in analyzing speeds.
Third-party sellers are kicking our first party butt. Badly.
ARM Speed and Duration Data for Hedging Purposes
1. ARM Speeds and Durations
January 21, 2015
(INSTITUTIONAL USE ONLY)
What in the heck is going on out there? In the last week, we have heard some
very, VERY odd comments on ARM durations.
No one is talking about speeds. No one seems to know where these risk numbers
are coming from. We’d like to offer some rational, hard-data-driven ideas back
into the melee.
Part of the problem, I suspect, is there are comments coming from traders with
mixed-vintage pools, portfolio managers with aged pools and pipeline hedgers
with phone numbers for both of the two. No matter which of the three you may be,
let’s talk about prepay speeds; not durations. Let’s talk about production age and
ARM class; not, “What are you using for 7/1’s…”
We’re going to cheat here and give the answers first. That said, a substantial
amount of data (on the attached spreadsheets) should illustrate where our speed-
guesses come from. No one is going to get speeds, dv’01s, durations and the like
correct. There should be, nonetheless, a reasonable set of boundaries for what
“may happen”—not just a prepay shout-out.
First up: Pipeline hedgers. The question (if my job is handling the pipeline) would
be along the lines of: “Where in the last few months has new production been
prepaying?” If I can get ballpark info on a HUGE range of new production, it
should be no problem to establish a starting point for hedge speeds.
Where to start? How about pulling every single agency pool produced from
September 2014 until now (well, 5-, 7- and 10-1s)? We did that and here’s what
came out:
2. The data to get all of the numbers in the pink boxes (above) is on the first three
pages of the (attached) spreadsheet. Once Coupon/WAM/AvgPPay/Avg Price are
established from very recent production, the risk numbers in the green boxes
may be calculated. DON’T LET SOMEONE GIVE YOU DV’01 AND DURATION! GET
THE INPUTS AND VERIFY THESE NUMBERS YOURSELF. USE OUR
MORTGAGEbuilder SOFTWARE OR BLOOMBERG™ “BC35” CALCULATOR. Our
program is easier and generates the proper hedge but, for the purpose of third-
party neutrality, here’s Bloomberg’s output:
So the risk numbers of dv’01 and duration may be calculated for all recent, like-
kind, pools. That’s great, but there are nuances YOU know about your pipe and
production that others may not. Perhaps your new issuance is coming from an
HPA surge area. Maybe your AEs just got a new expresso machine and they are
going to be pushing HARD for refi and recapture. Or maybe it’s the opposite of
both.
The first situation would likely push the pipeline hedge to re-run the risk at a 20%
faster speed. Or, maybe you’d like to see where your recent production has been
prepaying?-if you have pools trading from the last 4 months, we isolated each
3. servicer’s speed individually in the attached (look to the right of pool data for
each ARM class). You also know something about those pools—were they
“cherry picked” and the remainder sold or did your organization decide to go the
other way and fill those pools with slower retail origination?
The speed/performance data is out there. It will take a small effort to harness the
information. Calling around and asking for others’ opinions will not, in any
reasonable likelihood, get the correct answer.
Then, there is the portfolio hedger—be it trader on an available-for-sale desk or a
buy-hedge-hold manager. Portfolio speeds on seasoned ARMs do appear to be
picking up. There is also the seasonal kicker where prepays tend to make their
largest acceleration between January and July. In fact, speeds can jump by about
40% of their annual rate.
ARMs that were originated with wide gross-versus-net spreads look to be under
hard re-fi pressure.
Most of you that fall into the “portfolio/seasoned” hedger have “prepayment
specialists” on hand or who you pay as consultants. We absolutely have no
interest trying to interrupt that process.
For those hedgers that do not, we will offer a few projections on seasoned
Freddie ARMs.
4. If you are hedging any of these pools or like-kind pools/loans vastly faster or
slower, now would be a real good time to ask “am I sure?”
We, the Fixed Income Group at RJO, are not and never will be “the prepay guys.”
The last several business days have forced our hand. Successful hedging
requires a starting point of rational prepayment estimates.
Please look through the attached prepay data. Come to your own conclusions
based on that data and what you have specific to your product. The tendency will
be to guess faster; much faster. The data just isn’t there for new production.
The data appears to be there for faster speeds for older vintages.
More than anything, it is our hope that everybody challenges random prepayment
numbers. Accept the speed opinion of others with an understanding that you (or
ask us) will do the calculations. See the data—it’s all out there—and be
comfortable that your hedge numbers make sense to you.
Please see the attached workbook-
JC (for the Fixed Income Group)