Most banks are just starting to explore possible avenues of addressing FASB’s Current Expected Credit Loss (CECL) reserving model. The guidance so far has left much discretion in the banks’ hands when it comes to determining proper life-of-loan reserve calculation methods and models. Due to stress testing and heightened regulatory scrutiny, one of the more complex methods that most larger banks already have experience with is econometric modeling, namely Probability of Default (PD) and Loss Given Default (LGD) modeling.

Banks will most likely be expected to use similar modeling and forecasting techniques across the regulatory and accounting functions, and those that have built regulatory models will use some or all of that infrastructure to tackle CECL.

In order to get a sense of the magnitude of prospective CECL loss calculations, Trepp ran the entire CMBS loan universe through the Trepp Default Model for CECL. Despite the fact that they are not balance sheet loans, the aggregate results give a good idea as to how much a bank may need to reserve for loans in given risk cohorts (DSCR, LTV, vintage, state and region).

The portfolio includes all non-delinquent, single-property loans with at least one year of remaining term. Portfolio loans analyzed here are backed by the five major property types: industrial, lodging, multifamily, office, and retail. Finally, the loans only come from standard/public conduit, single-asset, and large loan CMBS deals. The portfolio excludes agency, mezz, short-term, single family rental, cell tower and other esoteric deal types.

Unsurprisingly, recent vintages perform better than older ones, partly due to adverse selection of what remains in older vintages and the higher liquidity underwriting environment of 2005, 2006, and 2007 loans that leads to higher PDs (all else equal).

CECL – ALLL% by Loan Vintage

Debt service coverage ratio (DSCR) and loan-to-value (LTV) ratio are two major drivers of loan performance, and may also be a way banks stratify their loans into broad risk cohorts for CECL disclosures. In general, loans with lower DSCR and higher LTV will have higher CECL ALLL forecasts.


The question that will arise is whether these numbers are reasonable and supportable within the CECL framework. Trepp’s previous CECL research – Looking at Historical CRE Losses for CECL – shows that total losses for similar CMBS loans so far are well higher than the model output we used for this analysis. However, those losses came during a time which included historic collapses in real estate values and liquidity.



Banks Pull Back From Large Multifamily Loans

While banks are still the largest financier and loan originator in the apartment construction market, they’re unwilling to lend as much now as they have been in the past compared with the value of the property, according to National Real Estate Investor’s Bendix Anderson.

Banks that once made loans that covered up to 75% of the cost of a development may now only cover 65%, and interest rates have risen to 275 to 325 basis points (bps) over LIBOR, up from the low 200 bps point range earlier in the recovery.

New regulations, such as Bassel II and Dodd-Frank, account for some of this hesitation to lend, Anderson notes, given that banks are now required to keep cash in reserve to offset their investments.

Lenders of all types are also growing more cautious as more new apartments open and vacancy rates begin to climb in many markets. Many lenders already have construction loans that are not performing as well as expected.

“Absorption is more challenging,” says Mitchell Kiffe, senior managing director with CBRE Capital Markets. “Pro forma estimates are maybe not being achieved. Lenders are looking carefully at all their new loan applications.”

Alternative funding sources include life company lenders, private equity funds, and FHA programs.