The treatment of commercial real estate mortgages under the Basel II guidelines has many institutions considering the best way to differentiate their mortgage portfolios. The appropriate reserve level for this asset class has long been the source of debate, and the latest Basel accords have increased the intensity of such discussions. The elusiveness of a meaningful risk metric has primarily been driven by a lack of robust performance information, and investors are left struggling with the quantification of their downside exposure on a go-forward basis.
The September 2002 "Real Estate Quantified" discussed an alternative approach to the issue of assessing commercial mortgage default risk. By calibrating a model based on systematic factors influential in determining the performance of the underlying real estate collateral, it appears that meaningful PDs (probability of default) can be assessed on a proactive basis. Basically, an informed opinion regarding default risk can be obtained by examining the mathematical integration of three key risk factors: 1) volatility, 2) expected growth, and 3) loan structure protection.
With this approach, meaningful default expectations can be assessed with a very limited amount of loan information. By direct modeling of individual mortgages, the impact of each loan can be determined in assessing a portfolio. As a result, important differences in the composition of a portfolio are reflected in the quantification of default risk. Critical considerations such as various levels of debt-service-coverage ratios (DSCR) and loan-to-value ratios (LTV), assorted changes in the ratios through time, and a diverse range in the uncertainty of such changes all play an essential role in the analysis of a commercial mortgage portfolio.
Using this approach to compare two portfolios reveals the impact of possible differences that regularly occur. The samples happen to be mortgage pools underlying two CMBS issuances currently in today's marketplace, but the analysis is routinely performed for private portfolio differentiation as well. The first pool is older, as it was put together in 1995, while the second pool was issued this past year. Both portfolios have geographic diversity and a range of property types. At issuance, the underlying portfolios were similar with both weighted average DSCRs of around 1.35 and each LTV averaging approximately 71%. They were also somewhat similar in size, each having between 120 to 150 loans at origination and balances of $650 million and $800 million. Currently, Portfolio 1 has 81 outstanding loans with an average DSCR and LTV of 1.34 and 71%, respectively, while the 121 loans in Portfolio 2 have an average DSCR of 1.40 and an average LTV of 71%. As of February 2003, neither portfolio had any current delinqu encies, although Portfolio 1 has 2.7% in REO (bank-owned/foreclosed).
Since we are running the analysis on a go-forward basis beginning with the fourth quarter of 2002, the summary statistics would suggest similar default risk characteristics. If anything, the somewhat lower current DSCR of Portfolio 1 would indicate marginally increased risk exposure. The analysis shows significant differences, however. First, cumulative PD for Portfolio 1 is expected to be less than half that of Portfolio 2. Although both pools have LGD (loss given default) numbers in the low 20% range, Portfolio 1's PD of 3.25% means that expected loss comes in at only 67 basis points, whereas because of its much greater PD of 6.88%, the loss expected for Portfolio 2 is 153 basis points. Obviously, simple summary statistics do nor tell the true story, as the systematic credit risk over the life of each portfolio is drastically different.
Part of the discrepancy may be explained by examining the riskiest loan in each portfolio. Portfolio 1 has an apartment loan in Kansas City that has a DSCR of 1.15, which is significantly lower than the average. The lower DSCR, combined with the relatively poor status of the Kansas City market, pushes the PD for this loan to a relatively high 10.78%, but the riskiest loan for Portfolio 2, a Charlotte apartment loan, comes in at an extremely forceful 18.69%. Although the 1.22 DSCR for this loan is not as poor as that of Portfolio l's riskiest, the dynamics in the Charlotte apartment market are extremely poor in an already volatile market. This is one of the most overbuilt markets in the country, and the severe deterioration in NOIs and values already occurring will continue over the next several years. This also means that the loss severity on the 85% LTV Charlotte loan will be in the 35% range, whereas the 25% LGD for the Portfolio 1 loan places it in a more normal range for apartment loans.
Along these same lines, examination of the 10 riskiest loans in each portfolio continues to support the outcome. For this subset, Portfolio 1 has less than half the expected loss of Portfolio 2. The 10 riskiest loans for Portfolio 1 and Portfolio 2 account for 162 basis points and 387 basis points of expected principal loss, respectively. As we have seen, this pattern of divergent default risk continues through each entire portfolio.
Another major part of this difference is the seasoning of each portfolio. Generally, CRE loans exhibit a typical default pattern that starts off at very low levels soon after origination, ramps up significantly over several years, reaches a peak somewhere between four and seven years into loan life, and then consistently declines to lower levels during the remaining term. Furthermore, the exact timing and level of this pattern are strongly influenced by economic conditions in specific property markets. Portfolio 1 has about eight years of loan seasoning. That means it should be past its peak levels of default, and its future trend in default frequency is likely to be downward. Loans underwritten in 1995 had a strong run-up in collateral values and NOI during the late 1990s. That means that they had larger equity and debt service cushions when they entered the recent downturn. As a result, many such loans are much better off than those underwritten a year or so ago.
Such is the case with our two portfolios. Figure 1 outlines the forward default frequency curve for the two portfolios. The red line of Portfolio 2 clearly shows the hump-shaped seasoning curve typical of mortgages over their entire lifespan. The forward-looking curve of Portfolio 1 is that of a pool beyond its peak likelihood of default. As can be seen by comparing the areas under each curve, the remaining cumulative defaults for Portfolio 1 are less than those of Portfolio 2.
It is very important to note, however, that over the next year or so the default risk for Portfolio 1 is actually higher than that for the second portfolio. Also note the "bump" in each curve. This chart outlines the PD for each portfolio, and, as loans mature, the picture changes. There are a large number of 10-year bullets in these portfolios. As a result, the portfolio characteristics change substantially 10 years after loan origination. Here we see that Portfolio 2 will benefit from the roll-off of some of its riskier loans (PD goes down after the change), while Portfolio 1 gets riskier after many of its less risky loans mature.
Basically, the combinations of property market and loan structure protection, and each portfolio's distribution among the same, are serving to make one portfolio much riskier, and that difference can be directly quantified across a number of important fronts. Obviously, such information improves proactive risk management of a portfolio and allows appropriate comparisons to help identify better risk-adjusted opportunities.
[c] 2003 by RMA. George J. Pappadopoulos, CFA, is director of Risk Management and Debt Research with Properly & Portfolio Research, Inc. (PPR), Boston, Massachusetts. PPR is an independent real estate, research, and consulting firm that works with fund managers, REITs, investment banks, commercial banks, pension funds, insurance companies, and private capital sources to help them meet their portfolio performance and risk management objectives. PPR can be contacted by phone (617-426-4446) or fax (617-426-4443).
COPYRIGHT 2003 The Risk Management Association
COPYRIGHT 2005 Gale Group