Abridged from "An Update of the Credit Profile of Standard & Poor's Rated U.S. Commercial Mortgage Loans"
For the credit performance of commercial mortgage loans, 2003 was the most stressful year since the early 1990s. It witnessed a sharp increase in the incidences of default. Among commercial mortgage loans originated between 1993 and 2002 that were pooled for Standard & Poor's rated CMBS issued in the U.S. (hereafter, the S&P's mortgage loans), defaults in 2003 reached 399. They accounted for 30% of all S&P mortgage loans that defaulted as of Dec. 31, 2003.
* In a study published last July, "Credit Profile of S&P Rated U.S. Commercial Mortgage Loans" (hereafter, the first study), which analyzed the credit performance of S&P mortgage loans, as of March 31, 2003, there were 1,006 defaulted loans. The cumulative default rate for the study period, which covered a time span of a little over 10 years between Jan. 1, 1993, and March 31, 2003, was 3.52%. This update, which extends the study period to Dec. 31, 2003, shows that defaulted loans expanded to 1,346 and the revised cumulative default rate for an 11-year period rose to 4.51%.
* Based on the additional defaults in 2003, this update re-examines the credit behavior of the 1993-2002 originated loans in terms of year of loan origination (vintage), type of underlying property, time to default, and default year.
Important new findings of the update are the following:
* The profile of the updated 29,827 S&P mortgage loans (hereafter, the studied loans) was similar to that of the first study with the exception that the average loan age (seasoning) has grown to 54 months. The original characteristics of the updated loans included the averages of 123 months of maturity term, 320 months of amortization term, $8.376 million of original loan balance, 68.56% of loan-to-value (LTV) ratio, and 1.54x of debt service coverage (DSC). For fixed-rate loans (representing 95% of the studied loans), the average mortgage rate was 7.83%. It was on average 225 basis points (bps) higher than the yield on the 10-year Treasuries at the time when these loans were originated.
* Loans that did not default (good loans) had virtually the same profile as that of the studied loans. Defaulted loans, however, had a markedly shorter loan age, larger rate spread (fixed-rate loans only), longer maturity, shorter amortization, smaller principal balance, higher LTV, and lower DSC than good loans.
* The 1995-1997 vintages continued to be the worst performers, with their cumulative default rates ranging between 7.59% and 8.75%. Holding seasoning constant, however, the 2000 vintage, now in its fourth year of aging, had the worst cumulative default rate at 4.28%.
* Loans secured by health care and lodging properties, with cumulative default rates of 14.90% and 19.88%, respectively, performed substantially worse than other types of loans. They were followed by retail loans (cumulative default rate: 3.97%), multifamily loans (3.40%), industrial loans (2.85%), and office loans (2.57%).
* The pattern of default risk in terms of "time to default" remained relatively high between the second and the sixth years of loan aging with the third to the fifth years (25-60 months) being the most risky period.
* Of the 1,346 defaults of the studied loans, 20% occurred in 2001, 24% in 2002, and 30% in 2003. This steadily rising share of defaults followed the gradual worsening in the deterioration of the national real estate market.
* The update re-estimated the default model with the same functional form (logistic regression equation) and explanatory variables as those of the first study. Once again, all 12 explanatory variables were estimated with their regression coefficients having the expected signs (relationships to the probability of default). All coefficients were statistically significant at the 1% significance level (or 99% confidence level) that the variables can significantly explain the default behavior of the loans.
* Among the 1,346 defaulted loans, 523 were resolved. Of these, 319 incurred principal losses, and 204 had no loss. The average loss severity for those resolutions with losses was 46.8%. For all resolutions, with or without losses, the average loss severity was 28.5%.
* For resolutions, the two worst vintages were 1997 and 1998, with overall severity ratios of 37.8% and 38%, respectively. Loans secured by health care and lodging properties again exhibited the most severe losses, with severity ratios of 53.9% and 44.3%, respectively. The office sector, with a loss severity ratio of 33.6%, continued to rank high among major property sectors. In contrast, multifamily housing remained the sector with the lowest loss severity, with a ratio of only 13.2%.
* Based on the default experience of the 1993-2002 vintages, the update found evidence that credit support levels of existing CMBS transactions are at comfortable multiples of the expected base-case loss. Going forward, the study results imply that the credit performance of newly originated loans is expected to be more resilient and the base-case loss substantially lower than those of the older vintages.
Proliferation of Defaults
Defaulted loans increased markedly during the last three quarters of 2003. As of Dec. 31, 2003, the number of defaults rose to 399 from 76 at the end of the first quarter. Among the 28,593 studied loans in the first study, there were 1,006 defaults. The overall cumulative default rate for the study period between Jan. 1, 1993, and March 31, 2003, was 3.52%. Updated defaults for the studied loans now totaled 1,346. The cumulative default rate for the 1993-2003 period now stood at 4.51%. The additional incidences of default occurred among all vintages and across all underlying properties.
To reiterate the definition of default, a loan is considered defaulted if it became 60-day delinquent for the first time. A loan's 60-day delinquency status may be cured later in its life (with the current credit status being classified as "current/performing"), but it is still recorded for the study as being a defaulted loan.
The default rate is defined as a ratio of the number of defaulted loans to the total number of studied loans. This definition is "loan count" based, not weighted by the original principal amount of the loans. In fact, the principal weighted default rate is much lower, at 3.05% for the update and 2.25% for the first study. Also, it should be noted that since roughly 15% of the defaulted loans were eventually cured after being 60 days delinquent, the default rate, for ultimate expected loss calculation, could be markedly lower.
The updated default experience of the cohort of 1993-2002 vintages in 2003 demonstrated that the credit performance of commercial mortgage loans is heavily influenced by the real estate fundamentals and the conditions of the national economy. More important, it provided comfort to investors who, in view of the rising defaults in 2003, are concerned about the adequacy of credit support levels on existing CMBS transactions.
As of year-end 2003, the studied cohort had an average age of 54 months and a cumulative default rate of 4.5%. From the pattern of time to default, the update indicated that 64% of the total defaults occurred within 54 months after loan origination. Thus, it can be extrapolated that a cumulative 10-year default rate for the cohort is 7.05% (4.51%/64%). With an overall average loss severity ratio of 28.5% and factoring in a 15% eventual cure rate for the defaulted loans, the expected 10-year cumulative loss of the 1993-2002 cohort would be 170 bps (7.05% x 28.5% x 85%).
This loss level, generally viewed as the "base-case" scenario, is well within the average of roughly 2.5% credit support level of B' rated credit classes of conduit CMBS issued in the past three years. Moreover, the credit support levels of AAA' and BBB' rated credit classes in recent years averaged around 18% and 8%, respectively. They were over 10 times and four times that of the base case loss.
Going forward, the study results also imply that the credit performance of newly originated loans is expected to be more resilient and the base-case loss substantially lower than those of the 1993-2002 cohort. There are three reasons for this expectation. First, the study proved statistically that mortgage loans originated in a weak real estate market performed far more strongly that those originated in a robust market. Loans originated in 2003 and so far in 2004, which supposedly have been underwritten more stringently on conservative property cash flows than those in the late 1990s, are therefore expected to experience a substantially lower frequency of default. Second, newly originated health care and lodging loans do not resemble their earlier year counterparts that suffered exceedingly heavy defaults. In fact, had the health care and lodging loans been performing close to other major types of loans, the cumulative default rate of the studied loans would have been 125 bps lower at 3.25%. Third, loss severity in a weak real estate market was significantly higher than that in a healthy market. As the pace of the current real estate market recovery gradually synchronizes with that of the rapidly expanding economy, loss severity of resolved loans is likely to abate markedly.
Stronger credit performance, the absence of chronically high defaulting health care and lodging loans, and abated loss severity combined are going to substantially reduce the base-case loss of the currently originated loans. It is conceivable that new commercial mortgage loans originated today would share the default experience of the 1993-1994 vintages. These loans were originated right after a severe 1990-1991 real estate recession and went on to endure another stressful downturn in 2001-2003. For comparison, the 1993-1994 vintages had a cumulative 10-year default rate below 5%, which could be the ultimate cumulative default rate of today's newly originated mortgage loans. This modest default rate, coupled with the somewhat abated loss severity and adjustment factor for cured defaults, would yield a 10-year cumulative base-case loss as low as around 100 bps. A significantly lower expected loss on new loans helps validate the case for the reduced credit support for the new CMBS transactions.
In addition, as confirmed by the update, loans with a larger original principal balance and a lower LTV are considerably less likely to experience default. Recently, most new transactions have been structured to combine large-balance, low-leverage loans with normal conduit loans. Benefiting from the lower required subordination of large loans, the overall credit support levels of the "new fusion" transactions have been reduced markedly from those of older transactions. The credit support levels for AAA' and BBB' classes of new fusion transactions average around 15% and 5.5%, respectively. These levels are 15 times and 5.5 times, respectively, the expected base-case loss. The reduced support levels and greater multiples are more significant at speculative-grade credit classes relative to those at the investment-grade classes.
Full report available on S&P's RatingsDirect and at www.standardandpoors.com.