The mortgage industry used to care.
We cared about all sorts of things… your landlord, your rent, where you got your down payment, your old W-2s, how long you had been on your job, how many jobs, what kind of jobs….
We don’t care about a lot of that stuff any more. The brave new world of underwriting by artificial intelligence, neural network, and website has dispensed with many, many things that human bankers used to think had something to do with your willingness and ability to pay back a loan.
Bankers were (yes, past tense) human, and in their incarnation as underwriters susceptible to asking for documents, evidence, proof of this or that which ought to be important in a credit decision. Like, how reliable a rent-payer you were before you wanted a home loan, what your job title is, or the continuity of your career path. Many, many boxes to fill in.
Humans being human, those vested with the authority to approve or deny a loan too often found it comfortable to say, “Because it’s the rule” — and that line was far too often a thinly-veiled version of “Because I say so.”
A common exchange under the old system began with an exasperated client: “I have a quarter of a million dollars in my Merrill Lynch account, and you want me to prove that my down payment has come out of that account?” Yes; and I won’t accept a copy of a check drawn on your money market account — wire or certified check only. “Even though you can’t possibly lose any money on my loan?” Yes; it’s the rule.
Enter the computer-as-underwriter.
There is a set of rules inside the computer, for sure, but it’s a very, very different set. This new rule book is secret, as the inventors at Fannie and Freddie fear that mortgage bankers like me will try to figure out how to “game” the new system. We are, of course, gaming the bejabbers out of it, one file at a time all day long every day, learning more about what it wants, what it will do and won’t do.
The new rule book is a vast improvement: when the factors crucial to repayment are present — large down payment and good credit — the computer dismisses the trivial. The computer gets no enjoyment from inflicting a bureaucratic insult, and so doesn’t bother.
There are exceptions: the FHA now underwrites by computer along with everyone else, but the FHA is engaged in a Great Leap Backwards, intensifying its demands for irrelevant information; and the Veterans Administration continues to transform the wonderful, old “G.I. Loan” into an impenetrable morass no longer worth the trouble.
If you have a large down payment, say 20% or more, and high credit scores, in or above the 680-720 range, here is a short list of the things we no longer care about:
- Your job title, job description, or line of work; where you worked before your current job, how long you worked there, or if you worked at all.
- Old W-2 forms: usually none; sometimes we don’t even need a current pay stub; we verify employment with a phone call, and take your word for your salary.
- If self-employed, or if total income includes a high percentage of incentive payments, one year’s tax return is sufficient… a big down payment, 30% or more, and we don’t even need that. (Exception, probably temporary: Jumbo “lenders” still want the whole tree-killing pile of 1040’s.)
- We no longer need proof that your down payment left the account where you have it; and the old 90-day account statement history is reduced to a single month.
- Your rent-payment history. We do not investigate your performance as a tenant at all.
- A residence history: we don’t care where you lived last month, or last year, or ever.
- The appraiser doesn’t need to go inside the house you’re going to buy: save $75. With 25% or more down, the appraiser doesn’t need to establish the value of the house, just drive by and report that there is a house: save $175.
- You no longer have to write us those dopey letters explaining why you made a late payment to Sears in August, 1995.
- We don’t need your divorce decree(s) unless you are obligated to make payments to an ex — and then only the page describing the payment.
Astounding. A machine — a calculating device — knows what too many human bankers never figured out.
If the deal is so strong that the bank can’t possibly lose money, why inflict bureaucracy on a good customer? Just say “yes,” ask how you can help, and ask how quickly the customer would like to close.
If artificial intelligence is capable of a breakthrough like this, it doesn’t bode well for the original model.