Hoocoodanode will be going offline for security related maintenance at 11 PM PST tonight, and may be down for several hours. Sorry for the inconvenience.
Great post. What happened to all the stated income buybacks though? If the fraud was THAT obvious (hawk alert!), then why didn't the investors go through the files and send the loans back to the issuer? Sure, some of those issuers are no longer with us, but a handful of major ones (Wamu, Countrywide, Indy) are still alive and kicking. Also, my understanding is that the time limit for buybacks is more generous in cases of misrepresentations.
This intolerable. The BS spin is criminally fraudulent.
October new home sales came in at 728,000
AND September's sales reported last month of 770,000 were revised down an incredible 54,000 to 716,000. Horrible numbers, Yet guess what the headline was.... CBS marketwatch reported as an improvement over last month.
Tanta, are you possibly suggesting the Golden Sacred Fatted Calf of "quick" loan approval is fundamentally inefficient? How dare you, madam!
But seriously (and, for once, un-sarcastically) - what's the best way forward?
20% down full-doc is the only way to get money? I'm not opposed to that, but surely there's some use to modern efficiencies.. installed the right way, of course. I am a process engineer by trade, so I have a couple biases in that direction.
For many subprime loans, the program guidelines allowed the originator to base qualification on features such as
stated income. Assuming that the originators underwriting standards did not require the verification through another means, or that a reasonableness test be conducted, the failure to perform these steps would not be an exception to their underwriting standards. Therefore, if the borrower or broker misrepresented the actual income, it is fraud on their part, but is it a breach of the reps and warranties? The same question would apply to borrowers who have artificially enhanced their FICO.
It "reads like a medical school textbook on using expensive fifth-generation antibiotics to treat staph infections caused by failure of doctors and nurses to wash their hands between patients."
Such an apt, well-placed analogy. What a pleasure it is to read this blog --
What happened to all the stated income buybacks though? If the fraud was THAT obvious (hawk alert!), then why didn't the investors go through the files and send the loans back to the issuer?
They probably did. Fitch doesn't say where it got these files.
The issue for Fitch here is, as far as I can tell, it started out by wanting to know if the data it got (FICO, say) was accurate. In other words, the tape said the borrower's FICO was 770. Was it "really" 770?
Once they looked at the files, they learned that yes, the FICO was really 770. That is not a misrep on the originator's or issuer's part.
But then they began to notice that there were facts in the file that should have lead an attentive underwriter to treat that 770 as if it were a 620. All of a sudden (for Fitch), it's not about data integrity any more, it's about whether a statistical indicator like a FICO is more meaningful than a comprehensive analysis of all the facts available.
November 29, 2007 -- AFTER a year and a half of stalling, the US Treasury finally complied with The Post's requests for information about The President's Working Group on Financial Markets - by delivering 177 pages of crap.
In essence, the Treasury's Freedom of Information officials said that the Working Group - affectionately nicknamed the Plunge Protection Team - doesn't keep records of its meetings.
How interesting and convenient!
Included in the 177 pages that the Treasury said responded to our request on the actions of The President's Working Group were 53 pages on which something was redacted - blacked out so that the discussion was unreadable.
Many of those 53 pages contained no words at all - just a big black blob.
Starting in June of 2006 The Post asked for an accounting of the actions of The President's Working Group, which was formed under President Reagan. The Group seems to have the ill-defined task of keeping an eye on the financial markets. We also asked for e-mails related to our request through the Freedom of Information Act (FOIA).
The Working Group operates out of the Treasury Department and includes the heads of the various exchanges in the US, as well as top-ranking government officials.
Hank Paulson, the Treasury Secretary, and Ben Bernanke, the head of the Federal Reserve, are the two most prominent members.
Back in August, Paulson said in a television interview that "we've reenergized The President's Working Group on Financial Markets."
The Wall Street Journal last year said that Paulson, upon becoming Treasury Secretary, was insisting that the Working Group meet every six weeks.
Whatever the schedule of meetings, one of those meetings occurred on Aug. 17 - the day the Federal Reserve surprised the financial markets with a cut in its discount rate.
According to records that someone else got from Bernanke's office through a FOIA request, there was an 11 a.m. conference call on Aug. 17 of the "PWG" - the President's Working Group.
Fed Governor Kevin Warsh and Patrick Parkinson, a Treasury staffer, took part in that call, according to Bernanke's phone log.
The day before - Aug. 16 - Bernanke and Paulson had lunch, but it isn't clear whether this was just two guys having a meal or if it, too, was related to The President's Working Group.
Hours after that lunch, word got around on Wall Street that the Fed was about to make a move and the stock market staged a tremendous rally.
The next day those rumors of Fed action proved accurate.
So what's the Working Group up to?
I suspect the group is ready to come to the rescue of the financial markets - even equities - in the case of a meltdown.
And as I've said in the past, that would be a completely acceptable task as long as it remains a limited power that is used infrequently.
But who decides when a rescue is needed?
And if no records are kept, who is held accountable if The
I am continually impressed with the thoroughness and intelligence put into the articles. Kudos.
However, I may have one discrepancy: Per my understanding a HAWK alert is an elected notice placed on the credit report because the individual borrower/consumer wished it so. For ex: I lost a credit last year. I cancelled everything but still called the three credit bureaus and placed an Alert on my Credit file.
Summation: I don't believe HAWK alerts are indicative of suspected fraud but rather protections placed by the consumer....of course I could be wrong!
Well, you're dealing with someone who picked up one of those antibiotic-resistant infections in the hospital, which got so bad it required another hospitalization.
So it cost thousands of dollars and lots of expensive drugs and IV fluids and days of my life spent watching CNN from a hospital bed to treat a problem that could have been prevented with 20 cents worth of soap.
To be fair about it, of course, it's not necessarily the case that I picked up this problem from the medical staff. Hospitals are full of visitors who carry microbes that don't hurt them, but that are deadly for people who are already sick. But rather than enforcing draconian rules for visitors (don't use patient bathrooms, don't touch the patient's food trays), we just invest in expensive drugs to treat the inevitable effects.
There's a moral in that story, too. Hospitals should be a place where you are subject to certain kinds of inconvenience. The demand on too many people's part is that visiting your friend in the hospital should be as easy and painless for you as visiting that person at his or her home. The mortgage counterpart is this idea that somehow getting a mortgage should be "painless," because most people aren't fraudsters, or if they are, it's the "harmless kind."
20% down full-doc is the only way to get money? I'm not opposed to that, but surely there's some use to modern efficiencies.. installed the right way, of course.
Full doc is probably the only way to get money. I don't think you have to go so far as to require 20% down across the board.
I like the idea of using AUS to "calibrate" the decisions of human underwriters. It provides an important corrective to, among other things, discriminatory practices. It's easy enough to do your own review of a loan, then run it through the machine to see what it says. If you and the machine keep diverging, you need to figure out who's nuts.
But that's a very different approach from running it through the machine first, and then only looking at it if the machine issues a warning message. The machine will only warn you if the data it gets is problematic; hence this widespread practice of making sure you never get a warning message by making sure you plug in "safe" data (stated income, inflated FICOs, etc.).
Fitch observes that the "underwriting" function has become "validating" the data that was originally fed into the AUS. The trouble is that we use the AUS results to determine the level of validation needed! It becomes a vicious circle, as long as the underlying assumption is that we "waste time" by lavishing attention on most loan files, because most loan files aren't fraudulent.
growler,you are wrong,a Hawk Alert can show any of a number of fraud and identity theft indicators,including but not limited to a stolen cc.also as far as AUS and AVM's they are very useful tools when used appropriately,and AVM's should IMO be used routinely in QC.Absent Fraud,they are about as accurate as a driveby appraisal,the only advantages a driveby has is to ensure there is actually a structure still there and to give notice of any visually obvious problem.
I don't believe HAWK alerts are indicative of suspected fraud but rather protections placed by the consumer....of course I could be wrong!
There are actually several different kind of "alerts" you can get on a credit report. If the alert is based on the consumer's request, then when you underwrite a loan for that consumer, you have to work harder to make sure that you are dealing with the real consumer--you were just told that the real consumer has reason to believe his credit record was hijacked!
There are other alerts that come out of the repository's anti-fraud software: things like multiple names used in conjunction with one SSN and so on that it really takes a computer to catch. It's only an "alert" because sometimes it's legit: you have old student loans under your maiden name, your mortgage under your married name, and you just opened a credit card under your newly-divorced name that isn't exactly the same as your old maiden name. So it falls on the underwriter to find out if it's something legit like this, or someone getting credit under someone else's SSN.
What bugged Fitch was not that the alerts were there, it was the lack of any notation in the file that they had been addressed. I see this a lot in due diligence: I look for a note by the underwriter on the Transmittal Summary that the alert was investigated and turned out to be a false alarm or adequately explained. The absence of such a note is a huge red flag not necessarily of borrower fraud, but of underwriter incompentence.
One would think that if fraud on the borrower's part can be proven, the lender would possibly have some recourse in the event of a foreclosure if the borrower has other assets.
Finally!! - how much of this crisis could have been averted had someone been doing this at the lender, the investment house or even at a so-called rating agencies? Today's financial instruments (mortgage/applications/appraisals) have become too complex to be underwritten without human intervention. All mortgage pool's should be sampled and reviewed.
I suppose that I will have to go and read tha report for the answer, but being lazy I'll ask first. How bad is the "negative selection" based upon payment history? Are these loans that were late on one payment or made no payments at all? What percentage of total loans have a payment history bad enough to be seleceted by their study? 1%? 10%?
When I visited a friend in the Burn ICU ~15-20 years ago, they DID have draconian rules for infection prevention: Handwashing, caps, disposable scrubs, and no touching.
"You do not have to believe that traditional human underwriting is perfect to wonder whether the cure is worse than the disease when it comes to compensating for automation."
There is a fallacy that sub prime underwriting was done in an automated, electronic decisioned manner. Nothing can be further from the truth. Outside of some mortgage players using a DU or LP decision in part to certify the credit portion of the underwrite, everything was manual.
The issue of the HAWK alerts had more to do with undocumented borrowers( in my experience here in So Cal) and less to do with fraud rings. Just because there is an alert doesn't mean there is a problem. It must be researched, and in my company, that was done by a auditor, not an underwriter. We auditors right on th eproduction floor because of the sheer volume of alerts and SS discrepancies.
And I would also say that the sub prime loan process was anything but cheap. Wages were sky high and the process was pretty much manual and was full of stops and starts and sideway paths.
That's the major reason why most ultimately failed, margins were too thin to absorb losses.
Jim a, it was actually a very interesting combination of adverse and non-adverse selection.
They were loans that had early defaults (mostly). But they were loans that would never have "looked" high-risk from a data tape in the beginning, before the defaults occurred.
They specifically picked out higher-FICO (650-770) purchase transactions.
One of the troubles with due diligence sampling during the boom was that everyone's "adverse sampling" protocols went out of their way to pick up, say, low-FICO cash-outs, because that, historically, was a much riskier transaction than a high-FICO purchase.
In the last year or two a lot of due diligence firms have been "reversing" their sampling techniques--going for overrepresenation of the "good loans" in their samples--and they've been finding more fraud than they did with the old adverse sampling.
There are two mechanisms in play: the good stuff isn't as good as it looks from quantitative measures, and also originators know perfectly well how sampling takes place. So if you do have a low-FICO cash-out, you know your odds of having it picked up in an audit are good, so you work it harder and document it better. On the other hand, you know that your 770 FICO purchase is probably going to blow through, so you don't work very hard on it.
I always enjoy and learn from your writing, and your conclusions in this case are right on. But I have to disagree with your use of these loan file results in your discussion. The population of loan files is so large, and adverse selecting a few clunkers so easy to do from a population of millions of files that Fitches results do not add weight to your conclusions. Better to wait for a statistically valid sample of the loan population, before including specifics IMO.
"What makes you think we're talking about subprime?"
In this case I am not sure what classification of loans they are, but based on my reading of prior posts on this blog, the case has certainly been pushed that sub prime has been guilty of too much reliance on speed induced technology.
When I visited a friend in the Burn ICU ~15-20 years ago, they DID have draconian rules for infection prevention: Handwashing, caps, disposable scrubs, and no touching.
ICUs are good about that kind of thing. So is the "surgical step-down unit," in which I once spent a week. That, actually, was the problem: most patients spend 24-48 hours in the step-down unit. Those of us who don't "step down" fast enough are in there long enough to get a string of visitors and exposure to half the hospital staff instead of the same two nurses.
But emergency rooms? God help us. The whole family, including snotty-nosed kids, goes to the ER with grampa and waits the ten hours it takes for gramps to get treated. They touch the equipment, they use the bathrooms, they flag down every passing nurse to come see gramps, assuring that every passing nurse is exposed to whatever microbes they're sharing. Plus they're loud, says someone who once tried to fall asleep in an ER bed. I'd say "pox on them" but that's the problem.
Based on my knowledge of processing practices of many of my former competitors, I would have to say this sampling probably is representative of the quality of most loan pools.
The population of loan files is so large, and adverse selecting a few clunkers so easy to do from a population of millions of files that Fitches results do not add weight to your conclusions.
Read the Fitch article. First, as I explained above, this isn't really "adverse selection." Yes, they were bad loans, but Fitch picked the bad loans that would not have been picked up in an "adverse sampling" method before they went bad.
Second, what "statistically valid" results do I need? Neither I nor Fitch is making a claim about the frequency that fraud occurs, or even the frequency that bad underwriting occurs. I am making the claim that it did not require extraordinary measures to see the problems in these loan files. Ya just had to open 'em up and look at them in the same way that a traditional human underwriter would have looked at them at the time they were made.
What is critical is to know how much of the overall total of the 2006 defaults are fraud related (not just high FICO). If the high fraud content permeates the whole group, then it is possible that the higher default rates have not even been impacted yet by the lowered housing values yet. It should also be noted that stated income makes the fraud easier for the amateur, but there are a number of other schemes that are possible (and have been done) that could be worked without it.
"Ya just had to open 'em up and look at them in the same way that a traditional human underwriter would have looked at them at the time they were made."
I know this point could be debated, but, not having a commentary in the file does not mean an investigation did not occur. As a lender, we underwrote to our guidelines which were reviewed and "approved" by either the conduit or rating agency that we were doing business with. Any due diligence, auditing deal-making and other underwriting discussions were kept confidential, in our systems conversation log or in a place outside the system. Of course, if a file made its way back to us because of a real or perceived problem, we would "share" our prevoius deliberations and findings with our securitization or whole loan partners if this was helpful.
What is critical is to know how much of the overall total of the 2006 defaults are fraud related (not just high FICO).
Well, my take on this is that they're just EPDs. We knew that.
The "Early Payment Delinquency" warranties that all the originators got caught up in last year and early this year--the forced buybacks--happened precisely because those loan sale contracts had EPD warranties in them, and had for decades. I mean, we have known for a very long time that EPD correlates very very strongly with fraud. That's why an EPD loan is such a hot potato, in the way that a loan with a similar delinquecy level that occurs after the first year is not.
Sure, the number of EPD loans rose dramatically starting in 2005-2006. But it isn't like the old EPD loans were any different; there were just fewer of them.
I know of exactly nobody who needs to have more statistical proof that EPD correlates with fraud. It does.
The question is, why did so many of them get past us?
Of course to say most EPDs are fraudulent is not to say that all fraud ends up EPD. A fair amount of fraud--the "fraud for housing" instead of the "fraud for profit"--is more durable than that. So you can expect fraud to keep resurfacing in these pools on "seasoned" loans.
Any due diligence, auditing deal-making and other underwriting discussions were kept confidential, in our systems conversation log or in a place outside the system.
You must be kidding.
You cannot sell a loan on the secondary market without representing that all material facts you have are in the loan file. Anyone who investigates a fact (such as a fraud alert) and doesn't put the results of that investigation in the loan file deserves to buy back loans.
I have done a lot of due diligence. Some of it was for the purpose of measuring the quality of a loan portfolio; some of it was for the purpose of measuring the quality of a loan origination business. I find omission of material fact, you take your loan back (or I withdraw my offer to do business with you). It doesn't have to be an adverse material fact. Your contract with me says you disclose all material facts that affect the investment quality of the loan.
Lenders whose underwriting logs are "offline" are waving a big red flag to me in terms of business due diligence. You are obligated to keep certain things confidential from the public. You have zero right to keep them confidential from your investors, funders, auditors, and examiners.
The fact is that in a well-run organization it is easier to keep all the information in the loan file than to segregate out parts of it for the "confidential" file. There is no operational upside for segregating information that you claim is supporting the loan decision.
Well the selection thing is one that the IRS does alot of work with. They have criteria that select returns for higher scrutiny and audit. THEY also pick some completely at random to verify their selection critera.
I am bemused that this came out just now. From the first time I heard, and indeed, everyone who hears about the innovative loans, wonders: What keeps fraud from happening? Apparently the answer is: nothing. When a friend thought that it might be a minor problem I said,"This is America damn it! It's already happening in every major city. If I can think of it, hundreds of people are already going at it."
"I am making the claim that it did not require extraordinary measures to see the problems in these loan files."
Sorry, I could not get access to the Fitch article. However, as one who has reviewed his share of loan files, I would venture to guess that I could go into any large home lending operation in the country and find 50 loan files with these kinds of problems. Just finding them tells us nothing meaningful about the frequency of occurence (ie. is this a serious problem, or a few outliers?).
What keeps fraud from happening? Apparently the answer is: nothing. yep. Of course the idea was that it was cheaper to insure (or assume that pooling and trancheing would deal with) the small number of defaults (fraud or otherwise) than screen for them. Of course this completely ignores the fact that if you're not looking for it, fraud will increase.
At the risk of sounding idiotic, this looks like the gritty, ground level view of systematic underpricing of risk.
George Bailey didn't need AUS or FICO scores to need an angel to bail him out.
The only reason to mention this is that if process breakdowns didn't cause the problem (but are a symptom), then better process improvement won't prevent the next problem. In addition, larding on a lot of additional controls after the fact is unnecessary if you survive. All you really need to do is follow the existing ones - which is perhaps your point.
When you have a relatively long cycle (multi year) there is this a huge tendency to mistake a cycle for a trend. You put a ruler on it and your forecasts are spectacularly wrong.
I've seen depressingly similar details in another financial business. If you are in a business like this, the only thing that you can do is quit. Quit early and get as much money as possible on your way out the door.
Hmm...on the face of it, the victims of predatory lending practices involved in this study would not seem to represent optimal material for "workouts."
When you have a relatively long cycle (multi year) there is this a huge tendency to mistake a cycle for a trend. You put a ruler on it and your forecasts are spectacularly wrong.
Gosh that's well put. I worked for one of these companies, and the tendency for the first half of the aughts to look at last year's loosened guidelines and last year's great performance (in a housing bubble) and to extrapolate that we could keep expanding guidelines and cutting underwriting and appraisal review ad infinitum with no increase in risk was simply irresistable to top management (Look, Ma, no defaults! What have we learned? Well, income doesn't matter! DTI doesn't matter! verifying assets doesn't matter! Owner occupancy doesn't matter!). Huge bonuses were "earned" based on tremendous earnings growth. Now the company is gone.
If you are in a business like this, the only thing that you can do is quit.
Also sad, but true. As those big bonuses keep getting earned, those at the top start (continue?) to believe their own BS. The see a "trend" and come up with what later are exposed as patently ridiculous explanations, but in the moment, they accept no dissension. (And the business press constantly harping on "optimism" being the key to business success doesn't help). Eventually, you get labeled a Nay-sayer.
" better process improvement won't prevent the next problem "
The fundamental problem is incomplete underwriting at the BEGINNING of the process - everything else is lipstick on a pig. Quality at the source is the cause of, and solution to, all of our issues. Therefore, metric of process = FTQ at whoever the original underwriter hands the completed file to.
Multi year trends? The only trends relevant to this discussion are the political semantics of "what loans do we want to approve" .. not "what are solid, time-tested guidelines for loan approval."
Process improvement isn't about adding controls, unless there aren't any controls on the system in question.. and then of course you have to take measurements to see what is actually happening. The slip of paper in your new khakis saying it was inspected by number 17? Gratuitous and unnecessary - definitely not a process improvement.
Of course a process breakdown is a symptom -- the breakdown was caused by something, and that something is exactly what improvements.. improve.
Also, it should be noted.. there are several process faults illuminated by the current investigations of the mortgage industry. There's multiple schools of thought as to where to start your improvements if there are giant kaizen opportunities in all steps of a process. One says that starting with the final step is your answer, and then work upstream so that each project will measurably improve the daily functionality of the work area and the overall output quality/volume. The other, and the one I'd go with in this case, would start with the source and work downstream to eliminate error accumulation throughout the process.. probably back to 100% manual underwriting, nothing's quick any more.
Although you thought your staph infection was an aside, Tanta.. this mortgage processing system, if you can call it that, sounds more like a hospital process than a financial process -- the entire mortgage chain is more a giant network of workarounds than a single path.. just like following the flow of a patient from ED admit to discharge.
"You cannot sell a loan on the secondary market without representing that all material facts you have are in the loan file. Anyone who investigates a fact (such as a fraud alert) and doesn't put the results of that investigation in the loan file deserves to buy back loans."
Representing and documenting are two different animals.
Perhaps you would require all the documentation, but most investors do not...unless of course you are having performance problems I suppose.
That is the whole point of representations and implied and explicit warranties. If you wanted to see our appraisal review process, I can show it to you, my audit process?, here is the flow and procedure. QC, that too. But I never have been asked to supply a blow by blow account of researching, detailing, investigating and resolving a HAWK alert. We put our third party audit report in the file as it shows the appropriate score with all issues cleared. That is about the extent of it.
The NYT article someone linked to on how the credit crunch is hurting the economy gave as an example a small business owner who manufactures carpets. She got her last $100,000 loan by mailing back a come-on flyer from a bank. 'This time it won't be so easy "They want to know abut me and my business, " she said. ' Horrors! The dreaded credit crunch means banks insist on knowing something about borrowers before approving loans; how will the economy survive?
We put our third party audit report in the file as it shows the appropriate score with all issues cleared. That is about the extent of it.
If you mean that there's a report in the file that looks like a checklist that has "HAWK Alert" checked off, with a comment saying it was investigated and cleared, then that would exactly be the kind of "notation" of this that Fitch wanted to see and did not.
Mind you, most of the loans they looked at actually had some evidence of fraud, but it was obviously ignored.
So in those cases, it's possible that the "audit report" wasn't included in the loan file because it would, um, show that the "issues" were not "cleared."
This whole business about not putting your documents in your loan files is about not making representations, that's all. In other words, there are lenders out there who think that if they just don't address an issue, they can't be busted for fraud. They'd rather not put the "audit report" in the file, because if it shows "cleared" and it turns out that there really was fraud, that looks like material misrep. Better to strip processing documents out of the files first, so nobody can "hang you" with them.
The operational issue for me, though, is that there is simply no upside to segregating file docs. Which is to say, it's a lot of work to maintain two or three files, or to keep documents associated with a specific loan anywhere other than in the loan's file. I therefore assume that anyone doing this is up to no particular good. That's a rebuttable presumption, but it's the presumption.
No dirty hands here...
i think some awesomely cool dood linked to this yesterday...
i think some awesomely cool dood linked to this yesterday...
Corrected
merci, T.
Tanta,
Great post. What happened to all the stated income buybacks though? If the fraud was THAT obvious (hawk alert!), then why didn't the investors go through the files and send the loans back to the issuer? Sure, some of those issuers are no longer with us, but a handful of major ones (Wamu, Countrywide, Indy) are still alive and kicking. Also, my understanding is that the time limit for buybacks is more generous in cases of misrepresentations.
This intolerable. The BS spin is criminally fraudulent.
October new home sales came in at 728,000
AND September's sales reported last month of 770,000 were revised down an incredible 54,000 to 716,000. Horrible numbers, Yet guess what the headline was.... CBS marketwatch reported as an improvement over last month.
Tanta, are you possibly suggesting the Golden Sacred Fatted Calf of "quick" loan approval is fundamentally inefficient? How dare you, madam!
But seriously (and, for once, un-sarcastically) - what's the best way forward?
20% down full-doc is the only way to get money? I'm not opposed to that, but surely there's some use to modern efficiencies.. installed the right way, of course. I am a process engineer by trade, so I have a couple biases in that direction.
Regardless, what do you think?
So the term "transaction friendly" applies to u/w and not just appraisers?
So much for operational excellence.
Tanta, thx for another great post. I like the medical analogy. Keb Mo' has some pertinent words of wisdom in his song "Keep it Simple".
I'm usually pretty good with the abbreviations, but am lost with AUS and AVM. Google returns me to this very post you made this morning. Please help.
This is a paradigm for all American businesses. Do it faster, cheaper and with half the personnel you did the same ting with last year.
How much more efficiency can the economy stand?
AUS= automated underwriting system,AVM =automated Valuation Model,or appraisal system.But Gosh,NOBODY KNEW that due diligence could detect obvious fraud!
Tom, thx.
David Pearson, Fitch also talks a bit about R&Ws:
For many subprime loans, the program guidelines allowed the originator to base qualification on features such as
stated income. Assuming that the originators underwriting standards did not require the verification through another means, or that a reasonableness test be conducted, the failure to perform these steps would not be an exception to their underwriting standards. Therefore, if the borrower or broker misrepresented the actual income, it is fraud on their part, but is it a breach of the reps and warranties? The same question would apply to borrowers who have artificially enhanced their FICO.
It "reads like a medical school textbook on using expensive fifth-generation antibiotics to treat staph infections caused by failure of doctors and nurses to wash their hands between patients."
Such an apt, well-placed analogy. What a pleasure it is to read this blog --
What happened to all the stated income buybacks though? If the fraud was THAT obvious (hawk alert!), then why didn't the investors go through the files and send the loans back to the issuer?
They probably did. Fitch doesn't say where it got these files.
The issue for Fitch here is, as far as I can tell, it started out by wanting to know if the data it got (FICO, say) was accurate. In other words, the tape said the borrower's FICO was 770. Was it "really" 770?
Once they looked at the files, they learned that yes, the FICO was really 770. That is not a misrep on the originator's or issuer's part.
But then they began to notice that there were facts in the file that should have lead an attentive underwriter to treat that 770 as if it were a 620. All of a sudden (for Fitch), it's not about data integrity any more, it's about whether a statistical indicator like a FICO is more meaningful than a comprehensive analysis of all the facts available.
LIGHTBULB!
JOHN CRUDELE
NEW YORK POST
November 29, 2007 -- AFTER a year and a half of stalling, the US Treasury finally complied with The Post's requests for information about The President's Working Group on Financial Markets - by delivering 177 pages of crap.
In essence, the Treasury's Freedom of Information officials said that the Working Group - affectionately nicknamed the Plunge Protection Team - doesn't keep records of its meetings.
How interesting and convenient!
Included in the 177 pages that the Treasury said responded to our request on the actions of The President's Working Group were 53 pages on which something was redacted - blacked out so that the discussion was unreadable.
Many of those 53 pages contained no words at all - just a big black blob.
Starting in June of 2006 The Post asked for an accounting of the actions of The President's Working Group, which was formed under President Reagan. The Group seems to have the ill-defined task of keeping an eye on the financial markets. We also asked for e-mails related to our request through the Freedom of Information Act (FOIA).
The Working Group operates out of the Treasury Department and includes the heads of the various exchanges in the US, as well as top-ranking government officials.
Hank Paulson, the Treasury Secretary, and Ben Bernanke, the head of the Federal Reserve, are the two most prominent members.
Back in August, Paulson said in a television interview that "we've reenergized The President's Working Group on Financial Markets."
The Wall Street Journal last year said that Paulson, upon becoming Treasury Secretary, was insisting that the Working Group meet every six weeks.
Whatever the schedule of meetings, one of those meetings occurred on Aug. 17 - the day the Federal Reserve surprised the financial markets with a cut in its discount rate.
According to records that someone else got from Bernanke's office through a FOIA request, there was an 11 a.m. conference call on Aug. 17 of the "PWG" - the President's Working Group.
Fed Governor Kevin Warsh and Patrick Parkinson, a Treasury staffer, took part in that call, according to Bernanke's phone log.
The day before - Aug. 16 - Bernanke and Paulson had lunch, but it isn't clear whether this was just two guys having a meal or if it, too, was related to The President's Working Group.
Hours after that lunch, word got around on Wall Street that the Fed was about to make a move and the stock market staged a tremendous rally.
The next day those rumors of Fed action proved accurate.
So what's the Working Group up to?
I suspect the group is ready to come to the rescue of the financial markets - even equities - in the case of a meltdown.
And as I've said in the past, that would be a completely acceptable task as long as it remains a limited power that is used infrequently.
But who decides when a rescue is needed?
And if no records are kept, who is held accountable if The
I am continually impressed with the thoroughness and intelligence put into the articles. Kudos.
However, I may have one discrepancy: Per my understanding a HAWK alert is an elected notice placed on the credit report because the individual borrower/consumer wished it so. For ex: I lost a credit last year. I cancelled everything but still called the three credit bureaus and placed an Alert on my Credit file.
Summation: I don't believe HAWK alerts are indicative of suspected fraud but rather protections placed by the consumer....of course I could be wrong!
Well then - I suppose we now know why this is taking place
, don't we?
Such an apt, well-placed analogy.
Well, you're dealing with someone who picked up one of those antibiotic-resistant infections in the hospital, which got so bad it required another hospitalization.
So it cost thousands of dollars and lots of expensive drugs and IV fluids and days of my life spent watching CNN from a hospital bed to treat a problem that could have been prevented with 20 cents worth of soap.
To be fair about it, of course, it's not necessarily the case that I picked up this problem from the medical staff. Hospitals are full of visitors who carry microbes that don't hurt them, but that are deadly for people who are already sick. But rather than enforcing draconian rules for visitors (don't use patient bathrooms, don't touch the patient's food trays), we just invest in expensive drugs to treat the inevitable effects.
There's a moral in that story, too. Hospitals should be a place where you are subject to certain kinds of inconvenience. The demand on too many people's part is that visiting your friend in the hospital should be as easy and painless for you as visiting that person at his or her home. The mortgage counterpart is this idea that somehow getting a mortgage should be "painless," because most people aren't fraudsters, or if they are, it's the "harmless kind."
Maybe, just maybe, that LIGHTBULB will be turned into a big freaking spotlight and turned on this issue:
The same question would apply to borrowers who have artificially enhanced their FICO.
bacon dreamz
Great post and good paper as well, read that last night and was really hoping you would do a post on it, Tanta-Thanks!!
Act V Scene 1
Enter Lady Macbeth, with a taper.
"All the perfumes of Arabia will not sweeten this little hand. Oh, oh, oh!"
LIGHTBULB! The emperor has no clothes. All it takes is one small child.
20% down full-doc is the only way to get money? I'm not opposed to that, but surely there's some use to modern efficiencies.. installed the right way, of course.
Full doc is probably the only way to get money. I don't think you have to go so far as to require 20% down across the board.
I like the idea of using AUS to "calibrate" the decisions of human underwriters. It provides an important corrective to, among other things, discriminatory practices. It's easy enough to do your own review of a loan, then run it through the machine to see what it says. If you and the machine keep diverging, you need to figure out who's nuts.
But that's a very different approach from running it through the machine first, and then only looking at it if the machine issues a warning message. The machine will only warn you if the data it gets is problematic; hence this widespread practice of making sure you never get a warning message by making sure you plug in "safe" data (stated income, inflated FICOs, etc.).
Fitch observes that the "underwriting" function has become "validating" the data that was originally fed into the AUS. The trouble is that we use the AUS results to determine the level of validation needed! It becomes a vicious circle, as long as the underlying assumption is that we "waste time" by lavishing attention on most loan files, because most loan files aren't fraudulent.
growler,you are wrong,a Hawk Alert can show any of a number of fraud and identity theft indicators,including but not limited to a stolen cc.also as far as AUS and AVM's they are very useful tools when used appropriately,and AVM's should IMO be used routinely in QC.Absent Fraud,they are about as accurate as a driveby appraisal,the only advantages a driveby has is to ensure there is actually a structure still there and to give notice of any visually obvious problem.
I don't believe HAWK alerts are indicative of suspected fraud but rather protections placed by the consumer....of course I could be wrong!
There are actually several different kind of "alerts" you can get on a credit report. If the alert is based on the consumer's request, then when you underwrite a loan for that consumer, you have to work harder to make sure that you are dealing with the real consumer--you were just told that the real consumer has reason to believe his credit record was hijacked!
There are other alerts that come out of the repository's anti-fraud software: things like multiple names used in conjunction with one SSN and so on that it really takes a computer to catch. It's only an "alert" because sometimes it's legit: you have old student loans under your maiden name, your mortgage under your married name, and you just opened a credit card under your newly-divorced name that isn't exactly the same as your old maiden name. So it falls on the underwriter to find out if it's something legit like this, or someone getting credit under someone else's SSN.
What bugged Fitch was not that the alerts were there, it was the lack of any notation in the file that they had been addressed. I see this a lot in due diligence: I look for a note by the underwriter on the Transmittal Summary that the alert was investigated and turned out to be a false alarm or adequately explained. The absence of such a note is a huge red flag not necessarily of borrower fraud, but of underwriter incompentence.
One would think that if fraud on the borrower's part can be proven, the lender would possibly have some recourse in the event of a foreclosure if the borrower has other assets.
Tanta -
I wonder if most realize how apropos your handwashing analogy is. Did you know that next week is National Handwashing Week? Dec 2-8.
Now, that's what the industry needs. An awareness campaign, replete with cartoon character. "National Diligent Underwriting Week", anyone?
the lender would possibly have some recourse in the event of a foreclosure if the borrower has other assets.
If the borrower had any other assets, they'd have been used to make a legit underwriting decision.
Recourse is worthless in 99% of these cases. It all hinged on the liquidation value of the property, and it still hinges thereon.
Finally!! - how much of this crisis could have been averted had someone been doing this at the lender, the investment house or even at a so-called rating agencies? Today's financial instruments (mortgage/applications/appraisals) have become too complex to be underwritten without human intervention. All mortgage pool's should be sampled and reviewed.
Wonder if they get it yet??
I suppose that I will have to go and read tha report for the answer, but being lazy I'll ask first. How bad is the "negative selection" based upon payment history? Are these loans that were late on one payment or made no payments at all? What percentage of total loans have a payment history bad enough to be seleceted by their study? 1%? 10%?
When I visited a friend in the Burn ICU ~15-20 years ago, they DID have draconian rules for infection prevention: Handwashing, caps, disposable scrubs, and no touching.
"You do not have to believe that traditional human underwriting is perfect to wonder whether the cure is worse than the disease when it comes to compensating for automation."
There is a fallacy that sub prime underwriting was done in an automated, electronic decisioned manner. Nothing can be further from the truth. Outside of some mortgage players using a DU or LP decision in part to certify the credit portion of the underwrite, everything was manual.
The issue of the HAWK alerts had more to do with undocumented borrowers( in my experience here in So Cal) and less to do with fraud rings. Just because there is an alert doesn't mean there is a problem. It must be researched, and in my company, that was done by a auditor, not an underwriter. We auditors right on th eproduction floor because of the sheer volume of alerts and SS discrepancies.
And I would also say that the sub prime loan process was anything but cheap. Wages were sky high and the process was pretty much manual and was full of stops and starts and sideway paths.
That's the major reason why most ultimately failed, margins were too thin to absorb losses.
Jim a, it was actually a very interesting combination of adverse and non-adverse selection.
They were loans that had early defaults (mostly). But they were loans that would never have "looked" high-risk from a data tape in the beginning, before the defaults occurred.
They specifically picked out higher-FICO (650-770) purchase transactions.
One of the troubles with due diligence sampling during the boom was that everyone's "adverse sampling" protocols went out of their way to pick up, say, low-FICO cash-outs, because that, historically, was a much riskier transaction than a high-FICO purchase.
In the last year or two a lot of due diligence firms have been "reversing" their sampling techniques--going for overrepresenation of the "good loans" in their samples--and they've been finding more fraud than they did with the old adverse sampling.
There are two mechanisms in play: the good stuff isn't as good as it looks from quantitative measures, and also originators know perfectly well how sampling takes place. So if you do have a low-FICO cash-out, you know your odds of having it picked up in an audit are good, so you work it harder and document it better. On the other hand, you know that your 770 FICO purchase is probably going to blow through, so you don't work very hard on it.
There is a fallacy that sub prime underwriting was done in an automated, electronic decisioned manner.
What makes you think we're talking about subprime?
Tanta,
I always enjoy and learn from your writing, and your conclusions in this case are right on. But I have to disagree with your use of these loan file results in your discussion. The population of loan files is so large, and adverse selecting a few clunkers so easy to do from a population of millions of files that Fitches results do not add weight to your conclusions. Better to wait for a statistically valid sample of the loan population, before including specifics IMO.
"What makes you think we're talking about subprime?"
In this case I am not sure what classification of loans they are, but based on my reading of prior posts on this blog, the case has certainly been pushed that sub prime has been guilty of too much reliance on speed induced technology.
When I visited a friend in the Burn ICU ~15-20 years ago, they DID have draconian rules for infection prevention: Handwashing, caps, disposable scrubs, and no touching.
ICUs are good about that kind of thing. So is the "surgical step-down unit," in which I once spent a week. That, actually, was the problem: most patients spend 24-48 hours in the step-down unit. Those of us who don't "step down" fast enough are in there long enough to get a string of visitors and exposure to half the hospital staff instead of the same two nurses.
But emergency rooms? God help us. The whole family, including snotty-nosed kids, goes to the ER with grampa and waits the ten hours it takes for gramps to get treated. They touch the equipment, they use the bathrooms, they flag down every passing nurse to come see gramps, assuring that every passing nurse is exposed to whatever microbes they're sharing. Plus they're loud, says someone who once tried to fall asleep in an ER bed. I'd say "pox on them" but that's the problem.
Based on my knowledge of processing practices of many of my former competitors, I would have to say this sampling probably is representative of the quality of most loan pools.
"representative of the quality of most loan pools."
Sub prime that is.
The population of loan files is so large, and adverse selecting a few clunkers so easy to do from a population of millions of files that Fitches results do not add weight to your conclusions.
Read the Fitch article. First, as I explained above, this isn't really "adverse selection." Yes, they were bad loans, but Fitch picked the bad loans that would not have been picked up in an "adverse sampling" method before they went bad.
Second, what "statistically valid" results do I need? Neither I nor Fitch is making a claim about the frequency that fraud occurs, or even the frequency that bad underwriting occurs. I am making the claim that it did not require extraordinary measures to see the problems in these loan files. Ya just had to open 'em up and look at them in the same way that a traditional human underwriter would have looked at them at the time they were made.
That's the real point here.
Those of us who don't "step down" fast enough
back to the ICU with you, ABS.
Fitch shocked, shocked! at the possibility of fraud.
My, what mewling babes they were to be led astray by the bad men!
They are backing and filling to protect what is left of their reputation.
So many years and billions into this and now they pick up a miniscule portion of the crap and probe it for worms.
Due diligence, anyone?
Anyone at all?
What is critical is to know how much of the overall total of the 2006 defaults are fraud related (not just high FICO). If the high fraud content permeates the whole group, then it is possible that the higher default rates have not even been impacted yet by the lowered housing values yet. It should also be noted that stated income makes the fraud easier for the amateur, but there are a number of other schemes that are possible (and have been done) that could be worked without it.
"Ya just had to open 'em up and look at them in the same way that a traditional human underwriter would have looked at them at the time they were made."
I know this point could be debated, but, not having a commentary in the file does not mean an investigation did not occur. As a lender, we underwrote to our guidelines which were reviewed and "approved" by either the conduit or rating agency that we were doing business with. Any due diligence, auditing deal-making and other underwriting discussions were kept confidential, in our systems conversation log or in a place outside the system. Of course, if a file made its way back to us because of a real or perceived problem, we would "share" our prevoius deliberations and findings with our securitization or whole loan partners if this was helpful.
"So many years and billions into this and now they pick up a miniscule portion of the crap and probe it for worms."
This would be a complicated topic to explain, but I think the rating agencies are/were truly surprised about the packaging quality of these loans.
What is critical is to know how much of the overall total of the 2006 defaults are fraud related (not just high FICO).
Well, my take on this is that they're just EPDs. We knew that.
The "Early Payment Delinquency" warranties that all the originators got caught up in last year and early this year--the forced buybacks--happened precisely because those loan sale contracts had EPD warranties in them, and had for decades. I mean, we have known for a very long time that EPD correlates very very strongly with fraud. That's why an EPD loan is such a hot potato, in the way that a loan with a similar delinquecy level that occurs after the first year is not.
Sure, the number of EPD loans rose dramatically starting in 2005-2006. But it isn't like the old EPD loans were any different; there were just fewer of them.
I know of exactly nobody who needs to have more statistical proof that EPD correlates with fraud. It does.
The question is, why did so many of them get past us?
Of course to say most EPDs are fraudulent is not to say that all fraud ends up EPD. A fair amount of fraud--the "fraud for housing" instead of the "fraud for profit"--is more durable than that. So you can expect fraud to keep resurfacing in these pools on "seasoned" loans.
Any due diligence, auditing deal-making and other underwriting discussions were kept confidential, in our systems conversation log or in a place outside the system.
You must be kidding.
You cannot sell a loan on the secondary market without representing that all material facts you have are in the loan file. Anyone who investigates a fact (such as a fraud alert) and doesn't put the results of that investigation in the loan file deserves to buy back loans.
I have done a lot of due diligence. Some of it was for the purpose of measuring the quality of a loan portfolio; some of it was for the purpose of measuring the quality of a loan origination business. I find omission of material fact, you take your loan back (or I withdraw my offer to do business with you). It doesn't have to be an adverse material fact. Your contract with me says you disclose all material facts that affect the investment quality of the loan.
Lenders whose underwriting logs are "offline" are waving a big red flag to me in terms of business due diligence. You are obligated to keep certain things confidential from the public. You have zero right to keep them confidential from your investors, funders, auditors, and examiners.
The fact is that in a well-run organization it is easier to keep all the information in the loan file than to segregate out parts of it for the "confidential" file. There is no operational upside for segregating information that you claim is supporting the loan decision.
Well the selection thing is one that the IRS does alot of work with. They have criteria that select returns for higher scrutiny and audit. THEY also pick some completely at random to verify their selection critera.
http://voiceofsandiego.com/articles/2007/11/27/news/mortgagefraud112707.txt
I am bemused that this came out just now. From the first time I heard, and indeed, everyone who hears about the innovative loans, wonders: What keeps fraud from happening? Apparently the answer is: nothing. When a friend thought that it might be a minor problem I said,"This is America damn it! It's already happening in every major city. If I can think of it, hundreds of people are already going at it."
"archana"? Tanta, Are you a current or former ISKCON-er? ha ha
I think you meant arcana?
Are you a current or former ISKCON-er?
I don't know what that is.
I think I was just typing with my nose.
Tanta,
"I am making the claim that it did not require extraordinary measures to see the problems in these loan files."
Sorry, I could not get access to the Fitch article. However, as one who has reviewed his share of loan files, I would venture to guess that I could go into any large home lending operation in the country and find 50 loan files with these kinds of problems. Just finding them tells us nothing meaningful about the frequency of occurence (ie. is this a serious problem, or a few outliers?).
What keeps fraud from happening? Apparently the answer is: nothing. yep. Of course the idea was that it was cheaper to insure (or assume that pooling and trancheing would deal with) the small number of defaults (fraud or otherwise) than screen for them. Of course this completely ignores the fact that if you're not looking for it, fraud will increase.
ISKCON - International Society for Krishna Consciousness... The Hare Krishna
Hawk Alert from Trans Union:
http://aaacreditinc.com/hawkalert.htm
At the risk of sounding idiotic, this looks like the gritty, ground level view of systematic underpricing of risk.
George Bailey didn't need AUS or FICO scores to need an angel to bail him out.
The only reason to mention this is that if process breakdowns didn't cause the problem (but are a symptom), then better process improvement won't prevent the next problem. In addition, larding on a lot of additional controls after the fact is unnecessary if you survive. All you really need to do is follow the existing ones - which is perhaps your point.
When you have a relatively long cycle (multi year) there is this a huge tendency to mistake a cycle for a trend. You put a ruler on it and your forecasts are spectacularly wrong.
I've seen depressingly similar details in another financial business. If you are in a business like this, the only thing that you can do is quit. Quit early and get as much money as possible on your way out the door.
Hmm...on the face of it, the victims of predatory lending practices involved in this study would not seem to represent optimal material for "workouts."
When you have a relatively long cycle (multi year) there is this a huge tendency to mistake a cycle for a trend. You put a ruler on it and your forecasts are spectacularly wrong.
Gosh that's well put. I worked for one of these companies, and the tendency for the first half of the aughts to look at last year's loosened guidelines and last year's great performance (in a housing bubble) and to extrapolate that we could keep expanding guidelines and cutting underwriting and appraisal review ad infinitum with no increase in risk was simply irresistable to top management (Look, Ma, no defaults! What have we learned? Well, income doesn't matter! DTI doesn't matter! verifying assets doesn't matter! Owner occupancy doesn't matter!). Huge bonuses were "earned" based on tremendous earnings growth. Now the company is gone.
If you are in a business like this, the only thing that you can do is quit.
Also sad, but true. As those big bonuses keep getting earned, those at the top start (continue?) to believe their own BS. The see a "trend" and come up with what later are exposed as patently ridiculous explanations, but in the moment, they accept no dissension. (And the business press constantly harping on "optimism" being the key to business success doesn't help). Eventually, you get labeled a Nay-sayer.
" better process improvement won't prevent the next problem "
The fundamental problem is incomplete underwriting at the BEGINNING of the process - everything else is lipstick on a pig. Quality at the source is the cause of, and solution to, all of our issues. Therefore, metric of process = FTQ at whoever the original underwriter hands the completed file to.
Multi year trends? The only trends relevant to this discussion are the political semantics of "what loans do we want to approve" .. not "what are solid, time-tested guidelines for loan approval."
Process improvement isn't about adding controls, unless there aren't any controls on the system in question.. and then of course you have to take measurements to see what is actually happening. The slip of paper in your new khakis saying it was inspected by number 17? Gratuitous and unnecessary - definitely not a process improvement.
Of course a process breakdown is a symptom -- the breakdown was caused by something, and that something is exactly what improvements.. improve.
Also, it should be noted.. there are several process faults illuminated by the current investigations of the mortgage industry. There's multiple schools of thought as to where to start your improvements if there are giant kaizen opportunities in all steps of a process. One says that starting with the final step is your answer, and then work upstream so that each project will measurably improve the daily functionality of the work area and the overall output quality/volume. The other, and the one I'd go with in this case, would start with the source and work downstream to eliminate error accumulation throughout the process.. probably back to 100% manual underwriting, nothing's quick any more.
Although you thought your staph infection was an aside, Tanta.. this mortgage processing system, if you can call it that, sounds more like a hospital process than a financial process -- the entire mortgage chain is more a giant network of workarounds than a single path.. just like following the flow of a patient from ED admit to discharge.
"You cannot sell a loan on the secondary market without representing that all material facts you have are in the loan file. Anyone who investigates a fact (such as a fraud alert) and doesn't put the results of that investigation in the loan file deserves to buy back loans."
Representing and documenting are two different animals.
Perhaps you would require all the documentation, but most investors do not...unless of course you are having performance problems I suppose.
That is the whole point of representations and implied and explicit warranties. If you wanted to see our appraisal review process, I can show it to you, my audit process?, here is the flow and procedure. QC, that too. But I never have been asked to supply a blow by blow account of researching, detailing, investigating and resolving a HAWK alert. We put our third party audit report in the file as it shows the appropriate score with all issues cleared. That is about the extent of it.
The NYT article someone linked to on how the credit crunch is hurting the economy gave as an example a small business owner who manufactures carpets. She got her last $100,000 loan by mailing back a come-on flyer from a bank. 'This time it won't be so easy "They want to know abut me and my business, " she said. ' Horrors! The dreaded credit crunch means banks insist on knowing something about borrowers before approving loans; how will the economy survive?
We put our third party audit report in the file as it shows the appropriate score with all issues cleared. That is about the extent of it.
If you mean that there's a report in the file that looks like a checklist that has "HAWK Alert" checked off, with a comment saying it was investigated and cleared, then that would exactly be the kind of "notation" of this that Fitch wanted to see and did not.
Mind you, most of the loans they looked at actually had some evidence of fraud, but it was obviously ignored.
So in those cases, it's possible that the "audit report" wasn't included in the loan file because it would, um, show that the "issues" were not "cleared."
This whole business about not putting your documents in your loan files is about not making representations, that's all. In other words, there are lenders out there who think that if they just don't address an issue, they can't be busted for fraud. They'd rather not put the "audit report" in the file, because if it shows "cleared" and it turns out that there really was fraud, that looks like material misrep. Better to strip processing documents out of the files first, so nobody can "hang you" with them.
The operational issue for me, though, is that there is simply no upside to segregating file docs. Which is to say, it's a lot of work to maintain two or three files, or to keep documents associated with a specific loan anywhere other than in the loan's file. I therefore assume that anyone doing this is up to no particular good. That's a rebuttable presumption, but it's the presumption.