The Credit Risk Transfer (CRT) program began with three deals in 2013. The deals referenced a total of $85 billion of rather pristine agency fixed-rate collateral, and were formulaic severity deals. That original vintage had a factor of 0.67 as of our April CRT monitor. Subsequent issuance has resulted in reference collateral totaling close to $1.4 trillion, making the CRT market reference population significantly larger than the remaining population of legacy non-agency deals. As we gather in New York for our second annual CRT conference with co-host IMN, we release this special CRT focused edition of the Pipeline.

In our first article, Harry Lijia Qin looks at the CRT universe by applying CPC and its underlying metrics as a lens. We first introduced Credit Profile Category (CPC) as an analytically driven, timely ratings alternative for non-agency securities. Because the underlying analytical measures generalize to any credit-sensitive structured-asset class, we were able to seamlessly apply it to the CRT market and use our online Credit Profile (CP) tool and database to compare CPC distributions against rating-agency ratings over time. Harry’s major results are as follows:

• Tail risk drops quickly as M1 CRT bonds run to maturity, which CPC captures much more quickly than agency ratings.
• CPC is able to catch the variability in credit quality for bonds which are assigned the same agency ratings.
• CPC increases precede agency rating upgrades, which tend to be slightly smaller in magnitude.

The evolution of the CRT market has important implications for the GSE reform policy discussion. Our second article, by Richard Cooperstein, takes an analytical look at an important intersection of policy and analytics; the question of how much risk-based capital the GSEs need for the credit risk of the reference pools and potential sources for that capital. Richard applies our credit analytics and uses a 97% Expected Shortfall (ES) together with 10-12% ROE to determine the level of g-fees required. His assumptions include g-fee income and CRT issuance providing protection to the level of losses indicated by the 97% ES and a fee-based Treasury backstop on the catastrophic portion beyond this level. His analysis shows that with continuing growth in this market, what looked like a rather large amount of capital required becomes a much more tractable problem.

Finally, in our Graphs of the Month section, we focus on model vs actual performance for CRT deal collateral. We show graphs by coupon, time period, credit score and current credit score. In general, defaults are running considerably below model expectations on our base case path, while prepayment incentive slopes are showing up as somewhat steeper than modeled.

CreditProfile Category (CPC) is a proprietary dynamic rating measure developed by Andrew Davidson & Co., Inc. (AD&Co). It serves as an alternative to ratings by major agencies. Traditional ratings have been widely criticized following the crisis for their overly favorable evaluations of mortgage-related bonds and their failure to adjust to changing market conditions in a timely manner. CPC is a quantitative measure designed to evaluate the credit risk driven by loan-level, multi-scenario analysis to capture the structural features inherent in securitized transactions.

In this article, we empirically compare CPC with agency ratings, using Credit Risk Transfer (CRT) deals, a booming sector for investors to gain access to residential mortgage credit. For a more detailed description of the CPC analytical framework, please refer to Anne Ching’s publication, Credit Profile Category: A Brave New World.

The major findings of this research are summarized below:

1. As a quantitative multi-dimensional credit risk measure, CPC has advantages over traditional ratings in particular for a highly levered structure with a high-quality reference pool like CRT bonds, whose credit quality is driven by tail risk.
2. Tail risks drop quickly as CRT bonds run to maturity, which CPC is able to capture on an ongoing basis. In contrast, credit ratings are slow in responding to such dynamics of deal-specific structural features and tail risks.
3. CPC is able to catch the variability in credit quality for bonds which are assigned the same agency ratings.
4. CPC helps overcome the stickiness of ratings, which is arguably one of the major shortcomings of agency credit ratings. For CRT bonds, CPC increases precede agency rating upgrades. While CPC starts rising soon after bond origination and keeps evolving dynamically, agency ratings stay sticky for quite a long time and eventually begins to slowly catch up with CPC.
5. Due to the desire to keep ratings stable, agencies tend to be conservative with the magnitude of upgrading as well. The magnitude of agency rating upgrades are found to be smaller than CPC increases.

If the GSEs were SIFIs (Systemically Important Financial Institutions), how much capital would they need? How do we assure that Fannie Mae and Freddie Mac continue providing their sizable benefits to the mortgage market but don’t need taxpayer money again? This article lays out a capital framework for the GSEs and applies our analytics and insights from the CRT market to quantify how much capital the US Treasury is already on the hook for.

The GSEs provide unmatched stability and standards for the mortgage market, but an outgrowth of conservatorship is that they have no capital. Scenario modeling and CRT pricing imply the GSEs need combined loss protection of up to 5% for new vintages to adequately protect taxpayers; the sources are revenue, risk transfer and capital. Exposure beyond 5% is best absorbed by Treasury in exchange for a risk fee. This gap is currently about $150B and suggests that “privatizing” the GSEs in a conventional way is probably a non-starter.

In this special CRT edition of the Pipeline, we focus on model vs actual performance for CRT deal collateral. We show graphs by date, coupon, vintage, credit score and current loan-to-value. In general, defaults are running considerably below model expectations on our base case path, while prepayment incentive slopes are showing up as somewhat steeper than modeled.