Using AI to make secondary market for online consumer loans

A San Francisco institutional investment management firm that acquires prime loans from online lenders is poised to securitize some of those loans for the first time.

What makes the 6-year-old company, Theorem, different from many players is that it uses proprietary machine learning technology to assess the likelihood of default of unsecured, online consumer loans available for purchase.

The asset-backed bond offering, Theorem Funding Trust 2020-1, is a $255 million transaction sponsored by the company, which counts on its big-data analytics platform to cherry-pick loans from the platforms of multiple online lenders.

The first collateral pool will consist solely of loans originated on LendingClub Corp.’s platform, according to Kroll Bond Rating Agency.

The Theorem 2020-1 collateral pool contains 24,725 loans with an average balance of $17,308. The weighted average coupon is 13.58% and WA borrower FICO is 708.

The notes offering includes a $199.95 million Class A tranche that carries a preliminary 'A' rating from Kroll. As is common for a first-time asset-backed securities issuer, the rating falls short of a higher investment-grade rating.

But also factoring into the deal are unique risks in the online personal-loan sector, such as the short ABS performance history of the asset class. Moreover, ratings agencies are worried about the higher risk of late payments now that so many consumers have lost their jobs during the coronavirus pandemic.

The issuer has included a 34.15% senior-note credit enhancement cushion (or surplus aggregate loan principal balance above the capital stack's notional value) as a buffer against any credit losses in the deal.

Theorem is also marketing a $28.35 million Class B tranche with an early BBB rating and a $27.6 million Class C tranche that Kroll expects will receive a final BB- rating.

The note proceeds from Theorem 2020-1 will be used to finance the loans’ acquisition. The notes themselves will pay investors through the receivables cash flow of the underlying loans.

The pool has a weighted average original term of 51 months on the loans, with remaining terms of 36 months. All of the loans are current and have never been modified.

In a presale report, Kroll added it has a higher-than-usual base-case default assumptions for the pool — a credit-loss range of 10.45%-12.45%). It cited its ongoing ratings review of several similar portfolios in the sector as the reason.

Those examinations include a review of LendingClub’s own deals. Ten of 17 tranche ratings from 11 of LendingClub’s securitization deals remain on watch for a possible downgrade by Kroll, owing to the economic downturn resulting from the coronavirus outbreak in the U.S.

Although 13% of Theorem’s total portfolio of loans purchased from LendingClub (NYSE: LC) are in a hardship repayment program, the loans culled for the current ABS deal “have demonstrated resilience thus far by remaining current on loan payments” and none have enrolled in any of LendingClub’s hardship programs, according to Kroll.

Theorem reduced its whole-loan purchases in March at the onset of the COVID-19 but resumed acquisitions in May.

The investment management firm oversees $1.2 billion in assets on behalf of institutional investors.

For reprint and licensing requests for this article, click here.
Digital banking Artificial intelligence Securitization Marketplace lending Machine learning Fintech Coronavirus
MORE FROM AMERICAN BANKER