By Arthur Q. Frank, director, and James M. Manzi, analyst, Mortgage Research, Nomura Securities International, Inc.
The relatively high defaults experienced by health care-related loans (financing nursing homes and assisted living facilities) in CMBS conduit deals in the late 1990s led CMBS conduits to avoid such loans in the last few years. In 1997, Congress changed Medicaid reimbursement policy for nursing homes from "cost plus" reimbursements, which guaranteed the nursing home owner a modest margin above actual patient care costs, to flat per diem payments regardless of provider costs. However, FHA-backed nursing homes have caps on unit construction costs, so they are built more cheaply, with fewer amenities than the nursing homes with higher fixed costs typically financed in the 1990s by CMBS conduits and other conventional financing. Consequently, the change in Medicaid reimbursement policy was less onerous for the health care projects in GNMA project loan pools. The graph below gives default rates by production year for health care loans (Section 232 and 232/223(f)) in GNMA pools versus the rest of the GNMA project loan universe, consisting almost entirely of apartment complexes of some kind. There is no clear pattern of health care facilities having consistently higher default rates than apartment complexes; it varies year-to-year, but overall default rates are comparable for health care and non-health care pools within the GNMA project loan market.
While we believe that Congress is more sensitive to the concerns of Medicaid providers now than it was in 1997, there remains some political risk embedded in projecting nursing home default rates. The above default history for Section 232 loans is reassuring, but hardly conclusive for the future. For the scenario where political decisions prevent Medicaid reimbursement rates from rising with the increase in operating costs (exacerbated at present by the national shortage of nurses), investors might examine the performance of REMIC tranches at default rates perhaps 20% above the normal default schedule used to price REMICs backed by GNMA project loans, the GN PLD curve, at least for deals containing a high percentage of Section 232 loans. The 120% default scenario is also useful for examining the impact of a moderately severe real estate recession, like that of 1990 to 1992. For all REMIC tranches other than Interest-Only securities, varying default rates from 75% to 120% of this GN PLD curve has little impact on the average life or yield.
Refi History of Health
While Section 232 and 232/223(f) health care-related GNMA project loans have not shown a historical propensity to default much differently than apartment complex loans, they have consistently refinanced over the past five years at much lower rates than similarly seasoned non-health care loans. We examine below the 5-year prepayment history of the universe of GNMA projects loans that finance health care facilities, compared to the history of similarly seasoned non-health care loans. Each GNMA pool from each production year from 1993 through 1997 is classified as health care or non-health care using Nomura's proprietary database of all GNMA pools by production year issued in 1993 or later.
The following graphs show the 5-year CPR history for the period January 1998 through December 2002 of health care versus non-health care GNMA project loans for each production year from 1993 through 1997. Pre-1993 pools are few and information needed to classify loans as health care or not is often missing, while from 1998 forward most loans were locked-out, and in any case they had not been in existence for long enough to have a 5-year prepayment history. Graph 2 gives the 5-year CPRs including loans that defaulted, counting defaults as prepayments, while Graph 3 excluded from the calculation all pools that terminated in default. By either calculation, the universe of health care loans has consistently refinanced at only about half the rate of non-health care loans.
Thus REMICs with a relatively high percentage of health care loans (40-50% in a few recent deals) clearly have less call risk to the investor than those with smaller percentages of such loans. For the investors purchasing long average life sequentials and last cash flow Z-bonds, as well as Interest Only (IO) tranches backed by GNMA project loans, there is clearly extra value in choosing REMIC tranches backed by collateral with a high percentage of health care loans. For buyers of short average life sequentials in a steep yield curve environment such as the present, a high percentage of health care loans slightly increases extension risk. However, short average life tranches in deals in which the sequentials receive accretions from the Z-bond as well as prepayments, have little extension risk in the first place. A typical 3.5 year average life sequential in a structure with a long Z-bond's accretions directed to the sequential tranches only extends to about a 4.5 year even at the unrealistically slow prepayment speed of 5% CPR. So the IO buyer, and to a lesser extent, the long tranche buyers benefit from a higher percentage of health care loans, while the short tranche buyer takes only slightly more extension risk.
Why have health care loans refinanced at so much slower rates than apartment complex loans? A major reason is likely the difficulty in operating FHA-financed nursing homes profitably in the environment of fixed per diem Medicaid reimbursement rates and rising compensation for skilled nurses due to their growing scarcity. Since a major factor in project loan refinancing is appreciation of property values leading to cash-out refinancing, even when prepayment penalty points are owed, the financial stress of the nursing home industry is preventing much price appreciation in this sector. While the financial stress is at present not severe enough to send default rates soaring, squeezed profit margins coupled with political uncertainty about the future of Medicaid reimbursement rates inhibit price appreciation for even successful nursing homes. Moreover, should nursing home operators succeed in obtaining higher per diem Medicaid reimbursements from Congress, the increasing competition from assisted living facilities for more affluent patients not dependent upon Medicaid will likely hold down nursing home values. Hence, we think that slower price appreciation and the associated slower refinancing activity compared to apartment complexes will continue for the next several years.
The growing percentage of health care loans in recent GNMA project loan issuance tends to make this market a bit better call-protected, since health care loans have consistently refinanced at lower rates than non-health care project loans. While the growing health care percentage of issuance perhaps increases the near-term uncertainty about default rates, given the dependence of FHA-financed health care operators on political decisions about Medicaid reimbursement rates, to date GNMA health care loans have not defaulted at significantly different rates than apartment complex loans.
The growth of the GNMA project loan market, with growing issuance and secondary market liquidity of GNMA REMICs backed by project loans, offers opportunities for structured finance investors to gain incremental yield compared to private label AAA-rated CMBS tranches, at the cost of slightly worse convexity but better credit (full faith and credit U.S. government guarantee instead of a AAA rating which can later be downgraded). These REMIC tranches offer agency CMO investors the opportunity to gain significantly better convexity with, at present, only a small yield concession. While none of the popular bond indices used as benchmarks for fixed income portfolio managers yet include GNMA project loans, that may change in the future as the project loan market grows in the years ahead from its current size of just under $23 billion.
* This is an excerpt from a much longer paper entitled "Some Investment Characteristics of GNMA Project Loan Securities." If you would like a copy, please contact Arthur Q. Frank (212-677-1477) or James Manzi (212-667-2231).