LendingClub's Troubles Bring Back Bad Memories
There are two main ways to look at LendingClub's problems this week:
- LendingClub got in trouble for being too much like a fintech -- a "financial technology" company -- and too little like a bank, focusing on algorithms and speed and coolness rather than the plodding legalistic work that is the actual business of finance.
- LendingClub got in trouble for being too much like a bank and too little like a fintech, with mission creep, conflicts of interest and "a complicated network of middlemen" instead of a pure technology-driven neutral platform.
But maybe it's both? Yesterday I wondered what "non-credit and non-pricing element" of its loans had gotten LendingClub into such hot water that it needed to kick Chief Executive Officer Renaud Laplanche out of the club. Bloomberg's Matt Scully went and found out:
Jefferies, which planned to bundle LendingClub loans into bonds, wanted the firm to improve its notifications to ensure that loan applicants understood they were handing over rights that could allow the lender to close loan applications faster, potentially by accessing information about customers’ assets and income. Jefferies demanded that LendingClub improve the way it discloses to borrowers what those rights, known as "power of attorney," allow the company to do, said two of the people, who asked not to be identified because they were not authorized to discuss it publicly.
Here, for reference, is the power of attorney language from what seems to be the current (March 2016 form) LendingClub Borrower Agreement:
As a condition to receiving a loan from us, you hereby grant to Lending Club a limited power of attorney and appoint them and/or their designees as your true and lawful attorney-in-fact and agent, with full power of substitution and re-substitution, for you and in your name, place and stead, in any and all capacities, to complete and execute the Loan Agreement and Promissory Note(s) in the form attached as Exhibit A that reflect- the accepted terms set forth in each of your final Truth in Lending Disclosure(s) as such may be posted from time to time in response to your loan request(s) in the on-line account you have established with Lending Club where documents are stored and with full power and authority to do and perform each and every act and thing requisite and necessary to be done in connection with such power as fully to all intents and purposes as you might or could do in person ("Power of Attorney" ).
I don't know what has changed between that and the last version. It is tempting to imagine this as a conflict between freewheeling financial technology people at LendingClub and buttoned-up bankers at Jefferies. In the brave new fintech world, who wants to be bogged down by explaining to borrowers that their lender will be signing their promissory notes for them? Signing your own loan agreements is so quaint and 20th-century! In the brave new world of financial technology, lenders will get your tax returns through an API, fair lending standards will be superseded by algorithms, and, you know, blockchain. Blockchain!
But probably it was just a pretty boring disagreement between LendingClub's banking lawyers and Jefferies' banking lawyers. "LendingClub believed its existing language was just as clear," reports Scully, "but changed its notifications to meet Jefferies’ request at the end of March." But then it still had some loans made with the old language, and just ... used those anyway:
LendingClub found its staff changed the dates on those loan applications, to make them appear to have been made after the company agreed to fix its notifications, the people said. That change would have given Jefferies the impression that the loans were made under the new terms, and were therefore eligible to be bundled into bonds.
So! Yesterday I was kind of meh on this scandal: Selling Jefferies some loans that didn't meet its "non-credit and non-pricing" needs in a way described as "fairly minor" doesn't, in the abstract, sound all that bad.
But forging loan application data to make a securitizer think that you met its standards for borrower disclosure, while not actually meeting those standards, is just the reddest of red flags. Jefferies was relying on the disclosure to borrowers being accurate, so that the loans would be enforceable, so that its notes would perform as expected, so that it could make representations about them to securitization buyers, so that those buyers could etc., etc., etc., etc. Loan securitization is a horrible Jenga tower of legalities that have to be perfectly balanced so that the whole thing can work as expected. If the borrower disclosure, at the very start of the process, is insufficient -- even if Jefferies was being overly conservative about the sufficiency of that disclosure -- then everything collapses. It isn't good. Jefferies and Goldman Sachs have stopped buying LendingClub loans and are reviewing their securitization plans.
There is something very, you know, tech-ish about this situation. "Tech is an industry of moving fast and breaking things," I once wrote. "Finance is an industry of moving fast, breaking things, being mired in years of litigation, paying 10-digit fines, and ruefully promising to move slower and break fewer things in the future." The technology startup world is, stereotypically, an ask-forgiveness-not-permission sort of place, and after all LendingClub did eventually ask Jefferies for forgiveness. And the particular thing that it got wrong really is so boring and archaic. Disclosure about who will physically sign certain forms? Come on. When all our contracts are smart and our borrowing is on the blockchain, none of this stuff will matter. It is an already fading archaism. You can see how a financial technology startup might care less about it than an investment bank.
But surely the real reason that LendingClub's troubles freak people out is not that its actions were so alien to the banking industry, but that they were so familiar. We have really, really, really been down this path before with securitizations. We have had poor borrower disclosure, and inadequate documentation, and misrepresentations from originators to securitizers that were passed unchecked to investors, and loan documents that were modified so they seemed to meet investors' standards. We have even had a convenient technology-enabled way to avoid the dumb archaism of signing too many pieces of paper. It is called MERS, and it replaces the old process of registering transfers of mortgages on paper in local county land offices with a sleek centralized computer registry. A blockchain for mortgages, if you will. It led to deep suspicion, legal uncertainty, forged mortgage transfer documents and court decisions declaring some of those mortgages unenforceable.
The fintech industry, after all, didn't invent fintech. Mortgage securitization was itself in some ways a precursor of "marketplace lending": Instead of investors funding banks, and banks lending to homeowners, securitization was a way for banks to provide a platform for investors to lend to homeowners. It modernized mortgage lending, abstracted it, de-risked the middleman and put the focus on algorithms -- or, at least, mathematical correlation models -- rather than community bankers' gut instinct and eye contact. It was clean and rational and sensible and scalable and modern and high-tech. And it was really great. And then it turned out that a few documents weren't quite right, a few legal interpretations were too aggressive, a few corners were cut. You know the rest. Certainly LendingClub's investors do.
From yesterday's Treasury Department white paper on "Opportunities and Challenges in Online Marketplace Lending":
Finally, many RFI responses highlighted the efficiency benefits of automated data sources replacing paper sources. It was recommended that the Internal Revenue Service (IRS) replace the existing Income Verification Express Service (IVES) taxpayer transcript request process with a more modern system known as an application programming interface (API). This API could allow any lender to build an automated way for borrowers to voluntarily share their tax data in a simple, fast, secure way. Online marketplace lenders argued that this change would make a meaningful impact on their ability to offer lower cost, faster, and safer loans.
This is surely a good idea, while nonetheless being adorably techno-utopian. The solution to everything is opening up the API.
That's also in the Treasury white paper:
While data-driven algorithms may expedite credit assessments and reduce costs, they also carry the risk of disparate impact in credit outcomes and the potential for fair lending violations. Importantly, applicants do not have the opportunity to check and correct data potentially being used in underwriting decisions.
By the way! Calling it a "non-credit and non-pricing element" is maybe a touch optimistic? Like, if Jefferies thought there was a real risk that these disclosures were inadequate, and might give borrowers claims or defenses against their lenders, then is that not, at least indirectly, a credit issue?
Ooh, this reminds me, my Bloomberg View colleague Noah Smith has a delightful post about Yuliy Sannikov's work on what Smith calls "high-frequency lawyers," referring to contracts that can be modified in continuous time. This is not practical now, but soon, soon:
Picture supply chains managed by algorithms, with purchasing contracts adjusted at high frequencies in response to constant flows of data about sales, shipping times, manufacturing costs, and so on. Imagine demand shifting back and forth across continents at close to the speed of light, switching small-batch production and shipping orders from one bidder to the next. Instead of high-frequency traders, imagine high-frequency lawyers, adjusting contracts to reward contractors appropriately for tiny improvements in efficiency, using sophisticated statistical analysis to figure out whether performance is being driven by outside conditions or by the contractor’s skill and effort.
Consider also the notion of blockchain courts for smart contracts. The future will be so goofy.
And this is not even to mention the other big issue that came to light in LendingClub's investigation, the thing where Laplanche, and director John Mack, and LendingClub itself, were investing in a fund that was buying loans from LendingClub (and other marketplace lenders). Bloomberg's Tracy Alloway:
"Their intention was to try to position themselves as a technology company and the more assets that they hold that they fund themselves the less they look like a technology company and the more they look like a bank or asset manager," said Peter Atwater, president at Financial Insyghts and a former banker at JPMorgan Chase & Co. "Those two industries have very different valuations."
Alphaville's Kadhim Shubber:
One reason for that coyness might be that such an investment contradicts the core proposition of Lending Club, as per Laplanche’s comments in January. The business is supposed to be a pure marketplace, which takes no risk on to its balance sheet and merely sells loans through to retail and institutional buyers. We now discover that the business does indeed have direct exposure to some of the loans that it makes.
My own view is that there is nothing necessarily terrible about one company both offering a trading platform and doing some trading on it. But it's an obvious conflict of interest that needs to be managed. If the LendingClub-affiliated fund gets inside information about the best loans, or gets to cherry-pick loans before they are offered to outside investors, or whatever, then that is obviously bad. There don't seem to be any allegations like that, though; it's mostly just the theoretical conflict at this point. But after years of theoretical conflicts at banks that turned out to be real conflicts, there is a certain level of cynicism.
In the olden days, a mortgage had to be registered in an actual piece of paper in a county land office, and each transfer of the mortgage required a new signature in that county office. This is still true in many places, and going to every county land office in the country just to transfer a pool of mortgages would take forever. So the mortgage industry invented MERS -- it more or less stands for "Mortgage Electronic Registration Systems" -- and gives MERS title to mortgages, in the county land offices, "as assignee." (Here are a diagram and FAQ.) Then MERS gives the relevant bank an entitlement to the mortgage in its database, and when the bank wants to transfer the mortgage, it transfers the MERS entitlement, not the actual underlying scrap of paper in a county office. It is like what the Depository Trust Co. system does for stocks, or I guess like what the blockchain will do for everything else.
I mean, indirectly, through lots of legal structure and underwriting controlled by the banks. But that is true of LendingClub too; it's not like LendingClub's lenders are actually lending directly to its borrowers.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Matt Levine at firstname.lastname@example.org
To contact the editor responsible for this story:
James Greiff at email@example.com