Thread by @patio11

Digitally Transforming the Mortgage Banking Industry is just a fantastic book. You should read it (critically, at points). It’s got one of the highest ratios of “novel insight” or “gobsmacking WTF” (sometimes simultaneous) per chapter I’ve seen.

amazon.com/Digitally-Tran…

You’ll get the most out of this book if you understand enough about mortgage industry to not need players in value chain explained to you. It’s a book by and for insiders; assumes good familiarity with e.g. securitization as a concept.

The central conceit for the book is that a mortgage is a manufactured product, a 700 page document written for the benefit of a buyer (who has authority to pick broker, who is you) but sold to an investor (via securitization).

Manufacturing process is 17th century.

A lot of the customer-visible pain in the process is that the front-end and back-end are managed by different people with different world views, imperatives, comp structures, etc.

Front-end: commissioned sales reps who turn leads into mortgage applications.

Back-end: operations professionals like underwriters who turn mortgage applications into a product which can be purchased by investors, while not violating the laws (of which there are many).

It’s “the back end is *}#*]ex” which makes for the best reading in the book, particularly because the author (who comes from the sales-y) side is resolute about the challenge being management and not IT systems, even when it’s manifestly IT systems.

A point the book makes numerous times: the human labor cost to produce a car at GM is less than that required to produce a saleable mortgage. (Both around $2k)

Previously, one sales professional generated sufficient paperwork to keep 0.66 operations professionals busy in producing the saleable artifact.

Dodd-Frank and post-crisis regulation increased that to 1.5 ops professionals, baking in about $2k regulatory expense per mortgage.

The industry clearly chafes at some of the refs, including Ability To Pay, which determines by fiat a relationship between documented income and the size of a mortgage payment one is allowed to take out.

Note that “documented” part because it is important.

Interestingly the primary cure for a defective loans (here meaning “it lacked a paper trail compliant with government standards” rather than “didn’t pay”) is that investors get to have you repurchase it, I assume as a free non-expiring put option.

Anyhow documented income:

Explicitly connected to millennials, blacks, and Hispanics having disproportionate troubles relative to true ability to pay, simply for setup of income streams. If you don’t have single established W-2, prepare for headaches.

Author profiles a hard-working immigrant who does 3 different gif economy things plus a bit of System D and would be almost immune to underwriting despite better income than many less industrious people.

Interestingly this *also* hit professional class early in careers.

Book claims that regulations no longer allow what were previously considered “common sense” loans or documentation of same, like “A graduating lawyer whose last 3 years show 0 income does not have a predicted income of $0” or “A med student with a signed offer of employment....”

Back to the IT. Oh God.

For want of workflow automation the war was lost, with the author’s explanation being a cultural impedance between front-end, back-end, and software vendors coming from one background or the other.

Manifest issues with software competence are handled with “human spackle” otherwise described as “checkers checking the checkers” (both times said with disdain by the author).

There’s an extended digression about Lean manufacturing and statistical process control, both written by and for an audience who understands neither.

I am aware that statement is uncharacteristically robust from me. I stand by it.

“Blockchain technology is the future, albeit not the immediate future.” <— Did I mention technical competence was not the industry’s defining strength?

Apparently the author had lunch at IBM which explained to him how Blockchain means he’ll never have to doubt the authenticity of a bank statement ever again, in case you wondered who pays IBM money or why.

Total cost of writing a mortgage averages $8.8k, which (if you express costs in basis points of the total loan value) allocates 6bps to software and 200 to employee compensation, with about half of that going to the originator (sales rep).

Quote:

One seasoned and cynical executive defined mortgage banking as “the systematic transfer of loan revenue and shareholder capital from shareholders to loan officers [sales reps], who bear no risk with respect to repurchase, prepayment, or regulatory actions.”

I could seriously quote this book to you all day, but the mortgage needs paying, so: just go read it.