Forums  > Basics  > CMO issuance data  
     
Page 1 of 1
Display using:  

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-08-29 19:02
I'm python illiterate, and so maybe my answer is to learn to read, but what are the best ways to go about accumulating historical CMO issuance data? I know Freddie Mac has a bunch on their website, and so maybe I should hire someone to web scrape.

I was wondering if someone out there knew more about how to collect such data. BBG worth a look?


"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-08-29 19:33
Re website scraping: you might want to have a look here: http://brooklyninvestor.blogspot.de/2016/08/more-13f-fun.html. If this guy can do it... Smiley

Re data: Bbg might be helpful if you can get the prospectus. If you have someone with access to S&P or Moody's you can try to get the prospectuses there (non-rated stuff is most likely out of scope anyway).

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-08-29 19:52
I can get the prospectuses (prospecti?) from freddie mac's website directly. Problem is getting them one by one going back to 1992 or whatever...

"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-08-30 08:43
Dunno honestly. Back then when I was looking at that stuff I used to download them one by one... besides those are usually pdfs with no standard format, so you have to think about a clever way to analyse them.

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-09-30 16:42
Ok, so I have (almost) the raw data, but a few questions for those with experience in CMOs.

The big picture thing I'm looking at is some quantifiable reasons for why a given issuance would be X% PAC vs Y% support, or Z% sequential, or yadda yadda...

But before I do that I need to make sure my understanding is sound. Is this how it actually works?

- I buy a passthrough for $400MM face
- I chop that up into $400MM worth of face in various tranches
- At least potentially, the proceeds from selling these tranches is something greater than $400MM.

If that's true, the weighted average price of my tranches should be greater than par, right?

Also, is there any way I can look at the OAS for the resulting bonds vs underlying collateral and imply a similar measure?

What I mean is, it's easy enough to say that

Price_CMO - Price_passthrough >0 implies some structuring/arbitrage profit

can i do something like this in OAS though? obviously I cant say

OAS_CMO - OAS_passthrough
It's early. This may make no sense.

"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-10-08 18:32
Just to get the grammar straight: passthrough describes a kind of SPV mechanism, namely what goes in goes out as-is. There is another mechanism whose name I have forgotten.

So you buy $400MM mortgages, put them in the SPV and issue tranches to refinance those mortgages. You could issues tranches above or below par (iirc equity tranches in CLOs are sometimes issued below par lately to give the equity investor a small upside kicker). So the wavg price could be above or below par. If you sell above there must be some specification where the surplus ends up (the SPV has several accounts, it has to be somewhere).

What is the OAS of the trance? You don't have a fixed maturity... of course you can run Monte Carlo and determine a maturity for a given path and then average over all paths. But I haven't seen this being done.

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-10-10 02:32
I'm a little less tired than when I tried my last post, so let me rephrase a bit.

I want to test a hypothesis regarding a possible explanation of the absolute and relative supply of different types of CMO tranches. Is it possible to use as a proxy for this a measure of arbitrage profits for the structurer?

Let's say today I buy $400MM face in mortgages and there is sufficient demand such that I can issue a big PAC tranche and a support tranche to accompany. If the market is dying for this type of security, does it stand to reason that as the structurer I may be able to issue my CMO at a weighted average price of 101 (assume buying the mortgages at par)? Where that extra goes is not technically my concern, but let's say the going market price for sequential tranches is 99.

Don't I rationally sell the PAC/support this time and "keep" the extra in some sense?

If next month the price of PAC is 99 and sequential is 101, is it reasonable that I would choose to do another $400MM deal with more sequential to supply that market and have more "arbitrage profit" in some sense?

Now, speaking to the OAS issue, if I look at something like the ML CMO indices, they quote an OAS each month for the various sub-indices. I dont know what type of term structure model they have and your point is well-taken on path-dependency; however, that number comes from "somewhere", right?

THe reason I bring up the OAS issue at all is that there is a data issue with trying to price the tranches at all. I have all the CMO issuance from Freddie Mac since 1988 or whenever it started, and I'd like to get some semblance of how much excess can be generated by buying mortgages at market prices and issuing CMOs. I dont know how to get prices (model or otherwise) from every CMO tranche through time. So, as a first pass I am using the price on these indices which try and mirror the market for different types of CMOs. But my problem is that each month the characteristics of the underlying mortgages may be sufficiently different and not a good proxy. The coupon characteristics may be changing over time, for example.

So I thought maybe I could use OAS in some way since I have OAS and price of these indices each month. The problem is that unlike the price arithmetic, I can't take OAS out of the risk-neutral expectation in the same way. It's not exactly like

Price_inputs OAS_inputs > OAS_outputs

Does that make any more sense?

"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-10-10 11:27
I'm a little less tired than when I tried my last post, so let me rephrase a bit.
So everything is fine at home Smiley.

I want to test a hypothesis regarding a possible explanation of the absolute and relative supply of different types of CMO tranches. Is it possible to use as a proxy for this a measure of arbitrage profits for the structurer?
Honestly, I am not sure whether there is resp. was demand for specific tranches. My impression was that certain classes of collateral were in demand or not (like US subprime at one point in time) and tranches were built such that a certain rating could be achieved (with a huge AAA tranche of course).
Caveat: I didn’t spent a lot of time on the cash side of things, so take this with a grain of salt.

Don't I rationally sell the PAC/support this time and "keep" the extra in some sense?
Usually issuers would sell down the tranche and not keep some in stock, waiting for better times to come. Rather go ahead and structure the next transaction and sell at a lower spread (but still at par).

Now, speaking to the OAS issue, if I look at something like the ML CMO indices, they quote an OAS each month for the various sub-indices. I dont know what type of term structure model they have and your point is well-taken on path-dependency; however, that number comes from "somewhere", right?
As far as I understand PAC tranches have scheduled amortisation to the extent that prepayments are within a certrain range. Based on that assumption (!) you can of course calculate an OAS. But if prepayments come in differently (e.g. because the housing market is going belly up) your OAS changes, maybe massively. There should be some statement regarding that assumption. Hopefully…

I dont know how to get prices (model or otherwise) from every CMO tranche through time. So, as a first pass I am using the price on these indices which try and mirror the market for different types of CMOs. But my problem is that each month the characteristics of the underlying mortgages may be sufficiently different and not a good proxy. The coupon characteristics may be changing over time, for example.

If you assume that characteristics might change the OAS approach won’t work. The very moment you plug in different prepayment assumptions (which might vary depending on your collateral and probably will) your OAS says something different. So I think this won’t solve your problem.

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-10-10 17:40
-So everything is fine at home....

Mostly yes...all things considered, quite well.

-Usually issuers would sell down the tranche and not keep some in stock, waiting for better times to come. Rather go ahead and structure the next transaction and sell at a lower spread (but still at par).

Ok, so how do I quantify how much an issuer makes on a given CMO? In my mind I had an arbitrage-type argument in my head. I turn inputs (mortgages) into outputs (CMO tranches), and if my inputs cost less than my outputs, on a weighted average basis, my business is profitable....I want to do more of this whenever there is a big demand (higher price) for my outputs, relative to my inputs.

Is this line of thinking "wrong"?

"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-10-11 12:04
Ok, so how do I quantify how much an issuer makes on a given CMO? In my mind I had an arbitrage-type argument in my head.

This argument is correct imo. I don't know the economics for CMOs, but there is definitely an arbitrage in CLOs. LCD even produced charts showing how big it is.

Besides the pure cash gain reducing your risk weighted assets also was part of the game. Securitizations had better risk weightings that the underlying loans. So even if you didn't make a few Dollar Bill you freed up capital which is also worth something (albeit less tangible).

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-10-12 15:22
How might one find the data (or even the chart itself) about CLOs? I tried some googling but no dice.

Hadnt thought of the risk reduction angle, but it makes sense as well.

"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-10-12 17:13
I was just looking through old documents but couldn't find anything. Lemme look a little more... maybe the arb is gone in the meanwhile.

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)

Rookie_Quant


Total Posts: 729
Joined: Jun 2004
 
Posted: 2016-10-14 17:35
Dont waste too much time.

As an academic, the current existence of a thing is rarely important ;)

We like to look back 10 years and go "ooooo wasnt that neat!"


"These metaphors and similes aint similar to them, not at all." -Eminem

Cheng


Total Posts: 2798
Joined: Feb 2005
 
Posted: 2016-10-18 13:52
Nothing so far. The reason could be that such concrete statements are rather rare these days... Wink

"He's man, he's a kid / Wanna bang with you / Headbanging man" (Grave Digger, Headbanging Man)
Previous Thread :: Next Thread 
Page 1 of 1