Following its merger with a special purpose acquisition company (SPAC), the artificial intelligence loan platform Pagaya Technologies (PGY) has seen its stock explode higher thanks to an extremely small public float. But one of the main underlying drivers of Pagaya's actual business fundamentals is how well the company is able to leverage its proprietary artificial intelligence technology to underwrite loans for its bank and fintech partners and investors on its platform.
With that said, let's take a look at how effective the company's loan underwriting has been so far.
Cumulative net loss rates
Pagaya partners with banks and other fintech companies that send loan applications to the platform to underwrite with its artificial intelligence technology, which has more than 16 million data training points since inception. The company then takes those loans for its partners and either sells them to investors or securitizes the loans and sells notes to investors.
For the loans securitized, investors can gain insight into how credit quality is performing by looking at reports from the Kroll Bond Rating Agency (KBRA), which continuously evaluates various asset-backed securities. There are a lot of considerations that go into evaluating credit quality, but the easiest way to evaluate different vintages is to look at how actual cumulative net loss (CNL) rates are trending compared to the KBRA's expectations at a given point in time.
You can also see how the KBRA's base case CNL rate for the life of the securitization trends as the securitizations seasons. In this chart, I am only going to evaluate asset-backed securities that have been seasoned for at least six months.
|Vintage||Months Seasoned||Current CNL||KBRA Expected CNL||Initial KBRA Base Case Loss Expectation||Current KBRA Base Case Loss Expectation|
It's really important to focus less on the current CNL rate and more on how it compares to the KBRA's expectations. For instance, while the 2020-3 vintage has a seemingly high CNL rate of 6.93%, it is significantly outperforming the KBRA's expectations through 20 months seasoned, when it expected the CNL rate to be 12.75%. Furthermore, the KBRA has lowered its base case loss expectation for the life of 2020-3 from 16.5%, when Pagaya first issued the securitization to 12.75%.
On the flip side, the current CNL rates for all the 2021 vintages are currently trending worse than the KBRA had projected at this point in their life, which ranges from six months seasoned to 15 months. But the KBRA has still not adjusted its loss expectations for the life of these securities.
I didn't see a ton of information in the recent reports regarding the composition of loans in each securitization, but the KBRA did note that it increased its loss expectations for loan grades D and E, which are made to borrowers lower on the credit spectrum. The KBRA noted this may be due to the "expiration of government stimulus offered throughout the COVID-19 pandemic, inflationary pressures on consumer prices, and other macroeconomic factors."
The KBRA added that newer vintages issued by Pagaya in 2022 and that are hardly seasoned yet were allowed to include higher levels of grade D and E loans. The KBRA has also increased its base case CNL loss assumption for Pagaya's most recent 2022 vintage to roughly 16% at the midpoint of its range, which is about 3% higher than the previous 2022 vintage.
So how is the underwriting?
The bad news for Pagaya is that current CNL rates for many of its 2021 vintages are worse than the KBRA expected them to be at their various points of maturity. The good news is that the KBRA hasn't changed its overall loss expectations over the life of the vintages yet, meaning they expect loss trends to eventually revert back in line with the assumptions.
However, I do not think it's a good sign to see CNL rates worse than the KBRA expected, considering there were still a lot of government stimulus programs in effect in 2021. Those programs have largely run their course and economic conditions are deteriorating. Loss assumptions have already gone up on recent vintages, and these look to have more borrowers lower on the credit spectrum, which are showing more cracks right now.
Based on what's going on, I would expect CNL rates to get worse, not better. I have yet to be convinced that AI underwriting can materially outperform traditional underwriting in more difficult economic conditions, which is why I think Pagaya, along with other fintech companies in the industry, need to further prove their concept.