In Depth Analysis: CalculatedRisk Newsletter on Real Estate (Ad Free) Read it here.

Friday, March 28, 2008

More On Chase and the Zippy Tricks

by Tanta on 3/28/2008 10:19:00 AM

Which, as Dave Barry always says, would be a great name for a band.

CR posted this shocker yesterday, a memo with a Chase logo attached (which doesn't mean much in these copy & paste days), sent via e-mail to mortgage brokers in what appears to be a "package" of "training documents," that provides tips on how to "cheat" and "trick" Zippy, Chase's AUS (automated underwriting system), into approving loans it would not normally approve. (Or, possibly, allowing loans to be documented or priced in a way they would not have been had the loan been submitted properly. It's hard to say exactly.)

I am not especially interested in the debate over whether it was "official Chase policy" (undoubtedly it was not) or whether it was a "joke" started by some wag at Chase ("this is how we oughta be training the brokers, ha ha!"). My own hunch is that it did start out as a joke, but there was at least one Account Executive at Chase who either didn't "get" the joke--which is scary--or who didn't have the sense to realize that certain forms of satire shouldn't leave the building.

But that's the thing: it works for me as a "joke," of the black humor deadpan sort, because, well, it isn't that far off "official policy." "Official policy" is simply couched in ponderous language, hedged about with earnest exhortations not to "misuse" the system that everyone ignores, and mostly "functional" in the sense that it covers certain raw acres of corporate butt, not in the sense that it really communicates clearly to a worker-bee what you're actually supposed to do when certain things--cough, cough--cross your desk.

I say this with complete confidence even though I haven't read Chase policy documents for years, and I have never read internal use only Chase policy documents. I have read hundreds of documents produced by dozens of institutional lenders, wholesalers, and aggregators describing credit policy and acceptable origination practices. I have written quite a few of them, to tell the truth. For that reason I have been in plenty of arguments over the years about those documents, since I have this thing about using language people can actually understand, explaining things clearly, and making policy documents actually useful to everyday operations rather than merely having them around in a file cabinet in case the regulators show up, and entirely ignored by everyone in-house unless and until that day arrives. So I am not speaking with actual detailed knowledge of Chase policy or the likely tone or content of Chase policy documents in particular. I am generalizing based on years and years of having to wade through that crap because no one else will.

What I really got interested in was not where, exactly, this document came from, but why, precisely, it says what it says. I mean, you can see this as a Chase Account Executive (or whoever authored it) simply baldly encouraging fraudulent misrepresentation, and that of course is what it is. But you also have to see that it does appear to require some misrepresentation to get past Zippy in some respects. That much is to Chase's credit.

Also, the way these things tend to work is that the "cheats and tricks" that people come up with--and no, this isn't the only one out there by a long shot, it's just the only one I know of so far that got to a reporter--tend to focus on routine problems. Nobody writes "cheat sheets" to deal with obscure complex things that only arise on one out of 200 loan applications. People write "cheat sheets" to deal with the problems you are most likely to have. And yes, I am using this term "cheat sheet" in its commonplace sense of a "dumbed down" policy or procedure, a set of "short cuts." The term has always been ambiguous: is a "cheat sheet" directions for breaking the rules, or just directions for following them faster and more easily? "Efficiency" and "user friendliness" have always been in danger of converging on the unethical. This is not a new problem. Perhaps our apparently willful lack of attention to this problem over the last several years is what was, in fact, different this time.

Besides the "most common problems," I have often found "cheat sheets" appearing in contexts where it's not so much that the issue at hand is "common," it's that this particular lender has a "thing" about that particular issue. If you ask a bunch of industry participants who are willing to tell you the truth, I suspect you'll find a high degree of consensus on what the "preoccupations" or "hot buttons" for any given wholesaler are. "Everybody knows," the story will go, that X is touchy about gift funds and Y is the hardest to deal with on short-form appraisals and Z isn't the place to go if you want a high-rise condo loan approved. And so on.

That kind of "eccentricity" is less common than it used to be, given nearly universal securitization of loans (which tends to "homogenize" the industry and make lenders willing to write loans they wouldn't touch for their portfolio) and competitive pressures that spawn "races to the bottom" all over the place. In the early years of the boom (until 2005 or so) I used to actually keep spreadsheets in which I tracked policy of ten major correspondent/wholesalers on a couple of dozen selected issues. I did not necessarily select the "common ones"; I was precisely interested in the more offbeat. Like condos with less than four total units in the project, or the rarer forms of temporary buydowns, or non-arm's-length transactions, or foreign national borrowers. ("Foreign national" in this context does not mean undocumented immigrant. It means someone who is not only not a U.S. citizen, but also not even a U.S. resident.) In other words, "niche" stuff that was once simply not allowed at all in the "prime" world, but which increasingly crept into "expanded criteria" programs--the precursor of what you all know as "Alt-A"--and then even into so-called "prime jumbo" or "prime non-agency."

What my little spreadsheet showed was that, over time, we went from an environment in which "go to Lender X if you have a small condo project" was the way it worked, to "just about everyone does small condo projects these days, so why shouldn't we?" It also showed that certain things were "migrating" from "expanded criteria" programs into "mainstream" programs. The race was on. Back in 2005 I was arguing strenuously that it was a race to the bottom, but that was of course a minority view.

Way too many people were convinced that we had the technology that would allow us to "race to the top." The idea was that in the Days of the Dinosaurs, doing 2-flats turned into condos or non-arm's-length deals or what have you was really very risky, but not any more, because now we have all these computer models that can much more finely-tune our risk assessments. Plus there was always the supplementary argument that hey! if someone like Chase is willing to do it, that must mean it's "respectable." That last argument wasn't always stated in such an unvarnished fashion, but that was the drift.

So this brings us back to the Zippy tricks, and the specific content of the infamous memo. If you read it from a certain angle--just for the sake of analytical clarity, not in aid of "defending" what is obviously indefensible--you can see it as some evidence that Chase's system, Zippy, has been correctly programmed to weed out a couple of serious problems:

1. Unstable income. The "trick" of putting all income in "base," instead of breaking it out (as the application form is designed to get you to do) into base salary, bonus, commissions, etc., is partly about getting the AUS to "let you by" with only a paystub to verify current income, if that. This is because it has been recognized since about the end of the last Ice Age that base salaries tend to be fairly stable, but bonuses and commissions tend to be rather volatile. This "other" income is, traditionally, used to qualify borrowers only when they can verify a history of having received it regularly, typically for a minimum of two years, and prospects for continuing to receive it, typically for at least the next three years. It is difficult to verify prospects; in most cases lenders digged so deeply into the past history of the income because that was the best clue to its likely continuation. The theory, for instance, was that an employer who regularly paid bonuses for the last several years was more likely to continue to pay them in the future than an employer who only just paid its first bonus a few weeks ago. Of course, stability of income projecting into the future was important because, well, we expected loans to repay out of income, not refinance money or sale of the home.

You can, if you like, theorize that Zippy has some "rules" built into it that involve a different set of parameters or a higher degree of scrutiny or possibly even a different "rate sheet" when the loan has substantial qualifying income other than base salary. It should. So the instructions to defeat that purpose by lumping everything into "base" is not just saving some idiot some keystrokes. And it's not a trivial thing involving some fussbudget who wants the forms filled out properly for no real reason. There's a real reason here.

2. Gift funds. We have been banging on for years now about borrowers having no "skin in the game." Lenders need to know--have always required to know--what the source of the down payment money is. It isn't just that loans with gifted down payments default more often, although they do. It's also that a lot of fraudulent "straw borrower" deals require "gift funds" to work out. Zippy was programmed to require this information for very good reasons.

3. "Inching up" income. Well, that one's pretty obvious. It is, however, the point at which I for one conclude that Zippy was probably not programmed correctly.

The fact of the matter is that an AUS should simply ignore DTI or cash reserve calculations when income or assets are or can be unverified. In fact, using DTI or months of reserves in your underwriting decision will inevitably produce worse results on a "stated/stated" loan than ignoring them will. They mislead you. I have known good, experienced, savvy human underwriters to fall into this trap, in spite of themselves: they see those "nice-looking" numbers and can't stop themselves from using them as a "compensating factor" on the loan.

There are people in the industry who think that "NINAs" are "worse than" stated income/stated asset loans, but I have never been one of them. What distinguishes them is that in a NINA, you don't even state numbers. You literally leave those boxes on the application blank. The loan is qualified with a credit report and an appraisal.

Now, I'll agree that NINAs are stupid--no problem there. But they're less stupid than "stated" deals. They don't "distract" you with made-up numbers. Actually, they often result in much better analysis of the appraisal and the credit report than you get in a "stated" deal. After all, the appraisal and the credit report are all you got in a NINA: you work 'em over. In the "stated" world people--and apparently some computers--keep getting sidetracked by those unverified numbers.

So I got the impression, for what it's worth, that Zippy is a mixed bag: it seems to have some responsible programming and some stupid programming. As I remarked in the comments to yesterday's post, the fact that it doesn't flash red lights and immediately refer the loan to the fraud-detection squad when it notices that a file keeps getting "resubmitted" with "refreshed" income and asset numbers is a huge problem. If this AUS really does allow multiple resubmissions with increasing stated income each time without setting off the red fraud flags, this is a very big deal for Chase. I'm here to suggest that Chase's regulators need to look into that. As I said, given the chance that the memo is a "joke," it's possible that the thing handles re-runs better than it sounds like it does. But Chase should have to answer this question now that the memo is on the table.

The final question for me, then, is what if anything is likely to be "unique" to Chase here. My answer is "not much." This just isn't in the same class as small condo projects or foreign nationals. We're talking Underwriting 101 stuff that brokers can, apparently, defeat by just putting a number in the wrong data-field or putting a "no" in the gift funds field or getting the system to keep "recalculating" DTI or cash reserves until you have forced it to reveal to you where its cutoffs are.

We all know why people do those things. What somebody needs to explain to me is how, after all this time, it's still so easy to do it. Where are the internal plausibility checks? Where is the "behavioral" logic that notices not just the content of the datafile but the manner in which it was submitted? Where's the basic randomly-selected pre-closing QC that snatches files out of the AUS queue and matches up the data submitted to the AUS with a quick phone call to the borrower?

Where, in other words, is the "high" tech? We've had AUS since the green-screen mainframer days of the 80s, kids. Thirty years down the road and these things are as easy to fool as Barbie's My First Laptop? After all the money these lenders have spent over the years on IT? There's something else that doesn't add up here besides a borrower's paystubs.

Technology aside, where, we also have to ask, are the "real" writers of policy and procedural and training documents? My own sense is that you find "cheat sheets" in companies that don't provide "real sheets" that are usable or comprehensible or updated or easily available on the website. We are, you know, in the "cut & paste" days. It's just no longer difficult to provide your people--or your broker clients--with the "real thing." Anyone who used to prepare policy with a Selectric and an old slow photocopier has a right, I think, to ask just what we're getting out of the "information" part of the "IT revolution." We just shouldn't have to have "cheat sheets" any longer; the "search" button takes care of the difficulties of looking things up. It matters, and it matters because asking people to look at the "real" policy instead of some dumbed-down "cheat sheet" written up by an Account Executive is not too much to ask. You think you will ever control for unethical behavior when you don't even demand moderate amounts of effort?

The whole industry has some explaining to do.