I think its important to note; one home will be values in a MULTITUDE of tranches and different CDO’s through the splitting process that Strassia describes well then valued in the process that China Guy lays out. Essentially we would have to count 1/1000 of a home insured with money that doesn’t (and never) existed. Then try and find the other CDO’s and splits of that single home… This is where the rubber meets the road. Your one home is NOT held in one CDO.
Imagine your pie, split into a million pieces, and you and your brother (and all of your friends) are also going to split that pie piece based on the results of your Shroednger’s Cat Box, which he assures you, always splits things fairly. Now try to figure out how much pie you are going to get.
This is also the reason that some Mortgages are hard to “restructure” specifically if the principal is to be lowered. You need a lot of investors to agree to lower the value of their so called bond. Try and figure out if you will want less pie or not, before the box opens. Good Luck!
Umm… I’m an ex-mortgage broker. I have strange stories to tell from this whole shit wad and a little knowledge to add.
My GF and her coworkers were questioned by the SEC and NY DA’s office, however, there are various political reasons why the rating agencies aren’t going to be charged.
Well, looking at hundreds or thousands of anything is a job. Plus how would someone siting in their office in Manhattan go and review the validity of loan applications for 1000s of borrowers all over the country?
I’ve never seen one of their models, but I presume they take a number of factors like borrowers income, credit history, demographic info, location, etc throw in some stats on market conditions, default rates and whatnot and come out with an expected rate of return. That’s why they have RMBS in the first place. One homeowner, either he pays or he doesn’t. A thousand, you can put some statistics around them, figuing x% will be able to pay.
The following is just personal experience with mortgage-backed securities processing, and over a decade old. Things may have changed a bit, and it doesn’t concern CDOs specifically, but maybe it’ll provide some insight.
In the '90s, I worked as a computer guy for a mortgage portfolio analysis company. The valuation process went like this: we’d receive data tapes from the institution that was selling a portfolio, customize a data-entry system, go on site, set up a computer network, have a number of loan underwriters come in and go through the loan paperwork one by one and page by page, then generate custom reports for the client summarizing the data. As far as I understand it, not too much has changed as far as that process goes. Yes, it’s quite a job – and we were well paid for it due to our accuracy and speed.
One deal I handled came in on a Friday at 4pm. I had to pre-process the data and configure our software, drive from Manhattan to Boston to get the data sheets on site by 9am Saturday morning, get the data into the system as it came in, drive back to Manhattan on Sunday night, and have final reports to the client on Monday morning. I put in one 44 hour “day” and totaled over 100 billable hours for a 4-5 day deal. That was the worst one I ever did…the job made my later graduate school experience tame by comparison.
Yes, although demographic info isn’t really used (in my experience only, which did not include any additional manipulations or analysis the client did after we supplied our analysis). And the models, of course, can only process the data one has, which is effectively what’s in the loan file to begin with. If the loan files are inaccurate – say, the originator lied about income levels, or an appraisal was inflated – it might be a legal issue after the deal is done.
I’m sure the models got better over time, but the issue that exists now is that historical data patterns no longer apply at this time (to a large degree). Assumptions about default rates, payment history, etc. are radically different than they were.
<Soapbox>
And the scariest part back then was the people doing the buying. Shit – during the RTC heyday, I was privvy to some conversations concerning auction pricing, that went something like:
Client critter 1:Our model says 70cents on the dollar for pool 8. Cc2:But the LTV of the Florida loans in that pool sucks! Can’t be worth more than 40! Cc1:OK, put down 52. That’s as high as we’ll go. Cc2:Right. Now, for pool 9, it says 53cents on the dollar…
A 30cent swing on a multi-million portfolio…based on split-second gut feel! I doubt that’s changed. And that was one of the key points that made me realize two things: (1) for many, the trader culture is nothing more than high stakes gambling with other peoples’ money and (2) I despise the trader culture and will never be directly involved with it professionally. </Soapbox>
To what extent can the analysis be automated? Is there a paper trail from the original mortgage through to the ultimate instrument? For high-risk instruments that were insured to bring them to investment grade, how did the insurers not know they were over-exposed? Is it because their models were flawed? Were there premiums not high enough? Was it the case of the aforementioned “gut instinct” culture?
Also, can msmith537 elaborate on political reasons for not prosecuting ratings agencies? Also, is there a way to avoid bias by these agencies in the future?
http://www.youtube.com/watch?v=bFa6WhDaDJ8 Here’s an audio wrapup. The rating agencies made billions by over rating the deals. They were competing for the business and ,since they are paid by the banks, there were no counter pressures to over rating. They ate their ethics ,for big paydays. The bankers did, the rating agencies did, and mortgage companies didn’t have any to start with.
You have to go back and find everyone that has taken out a mortgage in the last , oh, 15 or 20 years or so and get the proper information that should have been on the mortgages when they applied, if they file it out fairly or not.
That is the data collection stage so you can begin to reanalyze, which is something else entirely.
This was essentially the issue that arose when dot-com burst. The analysts were working for the same companies promoting the stocks.
It is precisely why I chose to not have anything to do with the housing bubble - anyone who paid attention in 2000 to dot-com issues could have and should have seen this coming from 10 miles away.
And it will happen again, because their is money to be made in the short term. I don’t think you can regulate it away because it is human nature to want to believe, and until people are so afraid of complex investment instruments that they can’t be sold anymore (like that is ever gonna happen), then this bubble making will just move from one industry to the next, and eventually back to some of the older ones.
Ok…so if these CDO’s are comprised of little bits of many mortgages then how did things work going in the other direction?
What I mean is, rather than working backwards to unbundle them how in the normal course of events (before the bubble burst and it was business as usual) did someone defaulting on their loan work through the system to the investors? Presumably there was some way to track who defaulted and who didn’t up the pipeline (and then by extension it would seem you should be able to work backwards as well).
Hopefully, someone will correct me if I’m wrong, but…
Two words: servicing companies. Analogically, in computer programming, there’s the concept of abstraction barriers, which enforce separation of system components. A well-designed component can only be accessed through its interface.
A servicing company serves as the abstraction barrier between borrower and investor; their entire purpose is to manage the loan. As I understand it, it’s nigh impossible for a borrower to pierce that barrier (for instance, to find out who actually owns it) – part of the servicer’s function is specifically to not allow it.
In the original CDO (the one based directly on the mortgages) they would be set up so that if one goes into default (or pays off the loan early) it affects the different slices in predictable ways. Slice Z would lose first, and slice A last. The income earned by these bonds is what goes to pay off the next set of bonds. A second level CDO that was based in part on the Z slices of this would pay less to its Z level, and so on.
These things would have a certain expected rate of default and pre-payment and still be able to pay in full to everyone. Even multiple levels up are fine as long as the number of defaults are low. The whole point was that the risk was spread around. Holding a mortgage that has a 1 in 10 chance of failing is pretty risky if that is all your savings. If you owned a piece of 1000 mortgages that each had a 1 in 10 chance of failing, you can predict that it will pay out 90% of value and price accordingly.
The two big problems we have now is that the rate of failure is no longer known (due to bad lending practices) and which bonds are backed by which mortgages is now extremely complicated. It seems to me the simplest (ha!) solution is to reevaluate each mortgage in the lowest level CDOs and then work your way back up. This is only the simplest because it is the only way I see to do it. Maybe after doing a few of these a model can be created to provide an estimate for others, but I am not sure if that will work until the housing market stabilizes.
Jonathan
None of the models are “mathematically proven”. They all encompass some level of gut feel, although enhanced by data mining and statistics. The fallacy of induction applies – gut feel on top of gut feel just magnifies the gambling aspect of the setup, and is less than worthless if historical patterns do not apply.
In fact, they do not get paid up front. Bond issuers were able to shop their securities around to different ratings agencies, and get preliminary ratings from all these agencies. At that point, the bond issuers decided which rating agency to go with. Obviously, this created a race to the bottom in the rating standards of the preliminary reports, in order to entice a potential customer.
When did this race to the bottom start? Before the current crisis, I never heard much about overrated debt instruments, although admittedly, I don’t really keep my ear to the ground about this type of thing. However, big institutional investors need to be able to trust the ratings and I still don’t see how overrating a bond isn’t fraud, nor do I understand why overrating won’t be prosecuted. I do understand that if an agency is confronted with a bond that would ordinarily have junk status but is insured, then according to their rules. So, what happened here?
It’s not at all clear to me what you’re trying to get at. Your first reply was that gut feeling is “often much more correct than tyhe complex, mathmatically-proven model.” I took that as a criticism (and fundamental misunderstanding) of financial models, none of which are mathematically “proven”. You follow up with the statement above (especially the “lot more useful” bit), and I now think you’re championing “gut feel”. So let’s get at it…
Are you saying that sometimes a person’s guess may be better than trusting a model? Sure, sometimes – some people do win the lottery, after all, and anyone can win a random 2-outcome bet 50% of the time. But people are notoriously bad at rational decision making, and tend to lose (big) in repeated iterations.
Are you saying that someone’s gut feel is usually better than trusting a model? Perhaps, for an exceptional person and an exceptionally bad model. But the whole point of using a model is to manage overwhelming complexity – of a depth, breadth, and objectivity far exceeding what anyone can internalize.
Surely you’re not claiming that a person’s guess is invariably better than a robust and comprehensive model, are you? Without further prompting, I won’t even respond to that. It’s a prima facie crap position.
Or you may be saying that no financial model can ever be a legitimately accurate representation of the world in all its glory, and therefore “gut feel” is a more valid methodology. While I’d agree with that first part, you can clearly see the fallacy in the second, right?
Whatever it is that you’re trying to convey, none of that addresses my little soapbox speechifying, which was directed at the “swinging dicks” (and wanna-bes) of the financial world. As an extreme example, I was talking about swings of millions of dollars in valuation, made in almost no time and with even less thought. To me, the very existence of such a huge value variance indicates that these people essentially had no idea of the actual value of the assets and were simply making shit up. You see, my point is that it doesn’t matter whether they’re relying on their gut or a model, it’s still naught but high-stakes gambling, encouraged (no, celebrated!) by the “trader culture”.
I doubt that’s changed in any way (Taleb excepted), and that’s the problem.
Getting back to the OP, a recent article in TIME, One Bad Bond, highlighted the difficulty of assessing the value of a particular mortgage bond, the Jupiter V:
