[Federal Register Volume 60, Number 26 (Wednesday, February 8, 1995)]
[Proposed Rules]
[Pages 7468-7479]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 95-3076]



=======================================================================
-----------------------------------------------------------------------

DEPARTMENT OF HOUSING AND URBAN DEVELOPMENT

Office of Federal Housing Enterprise Oversight

12 CFR Chapter XVII

RIN 2550-AA02


Risk-Based Capital

AGENCY: Office of Federal Housing Enterprise Oversight, HUD.

ACTION: Advance Notice of Proposed Rulemaking.

-----------------------------------------------------------------------

SUMMARY: Title XIII of the Housing and Community Development Act of 
1992, known as the Federal Housing Enterprises Financial Safety and 
Soundness Act of 1992, gives the Office of Federal Housing Enterprise 
Oversight (OFHEO) the responsibility for developing a risk-based 
capital regulation for the Federal National Mortgage Association and 
the Federal Home Loan Mortgage Corporation (collectively, the 
Enterprises). To discharge this responsibility, OFHEO must develop and 
implement a risk-based capital ``stress test'' that, when applied to 
the Enterprises, determines the amount of capital that an Enterprise 
must hold initially to maintain positive capital throughout a ten-year 
period of economic stress.
    This Advance Notice of Proposed Rulemaking (ANPR) announces OFHEO's 
intention to develop and publish a risk-based capital regulation and 
solicits public comment on a variety of issues prior to the publication 
of a proposed rule. OFHEO requests comment from the public concerning 
issues set forth in the ``Solicitation of Public Comment'' subsection 
of the Supplementary Information section below.

DATES: Comments regarding the ANPR must be received in writing on or 
before May 9, 1995.

ADDRESSES: Send written comments to Anne E. Dewey, General Counsel, 
Office of General Counsel, Office of Federal Housing Enterprise 
Oversight, 1700 G Street, NW, Fourth Floor, Washington, D.C. 20552.

FOR FURTHER INFORMATION CONTACT: David J. Pearl, Director, Research, 
Analysis and Capital Standards; or Gary L. Norton, Deputy General 
Counsel, Office of Federal Housing Enterprise Oversight, 1700 G Street, 
NW, Fourth Floor, Washington, D.C. 20552, telephone (202) 414-3800 (not 
a toll-free number).

SUPPLEMENTARY INFORMATION:

Background

    Title XIII of the Housing and Community Development Act of 1992, 
Pub. L. No. 102-550, known as the Federal Housing Enterprises Financial 
Safety and Soundness Act of 1992, 12 U.S.C. 4501 et seq. (Act), 
established the Office of Federal Housing Enterprise Oversight (OFHEO) 
as an independent office within the Department of Housing and Urban 
Development. OFHEO's primary function is to ensure the financial safety 
and soundness and the capital adequacy of the nation's two largest 
housing finance institutions--the Federal National Mortgage Association 
(Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie 
Mac) (collectively, the Enterprises).
    Fannie Mae and Freddie Mac are Government-sponsored enterprises 
that serve important public purposes and receive significant financial 
benefits, including exemption from state and local income taxes and 
special treatment of their securities in a variety of regulatory and 
transactional situations. Although the securities that they issue or 
guarantee are not backed by the full faith and credit of the United 
States,\1\ their status as Government-sponsored enterprises creates, in 
the view of financial market participants, an implicit Federal 
guarantee of those securities. Furthermore, the failure of either of 
the Enterprises would have serious consequences for the performance of 
the nation's housing markets, with a potentially disproportionate 
effect on low- and moderate-income families.

    \1\See section 306(h)(2), Federal Home Loan Mortgage Corporation 
Act (12 U.S.C. 1455(h)(2)) and section 304(b), Federal National 
Mortgage Association Charter Act (12 U.S.C. 1719(b)).
---------------------------------------------------------------------------

    The Enterprises engage in two principal businesses. First, they 
maintain a portfolio of residential mortgages and, second, they issue 
and guarantee pools of residential mortgages--in the form of mortgage-
backed securities (MBS)--that are held by investors. One of the 
Enterprises' principal financial risks stems from losses associated 
with defaults on mortgages that they hold or guarantee. The other 
financial risk stems from losses associated with changes in interest 
rates. Because the effective maturities of the Enterprises' assets and 
liabilities are not the same, interest rate changes could cause the 
margin between the average yield on assets and the average yield on 
liabilities to narrow or even become negative.
    The Enterprises' capital serves as a cushion to absorb financial 
losses for a [[Page 7469]] period of time until the cause of the losses 
can be remedied, thereby reducing the risk of failure. The Act requires 
OFHEO to establish, by regulation, risk-based capital standards for the 
Enterprises. The regulation will describe a risk-based capital stress 
test (stress test) that OFHEO will develop and implement to determine 
for each Enterprise the amount of capital\2\ necessary to absorb losses 
throughout a hypothetical ten-year period marked by severely adverse 
circumstances (stress period).

    \2\For purposes of the ANPR, the term ``capital'' means ``total 
capital'' as defined under section 1303(18) of the Act (12 U.S.C. 
4502(18)) to mean the sum of the following:
    (A) The core capital of the [E]nterprise;
    (B) A general allowance for foreclosure losses, which--
    (i) shall include an allowance for portfolio mortgage losses, an 
allowance for nonreimbursable foreclosure costs on government 
claims, and an allowance for liabilities reflected on the balance 
sheet for the [E]nterprise for estimated foreclosure losses on 
mortgage-backed securities; and
    (ii) shall not include any reserves of the [E]nterprise made or 
held against specific assets.
    (C) Any other amounts from sources of funds available to absorb 
losses incurred by the [E]nterprise, that the [Director of OFHEO] by 
regulation determines are appropriate to include in determining 
total capital.
    The term ``core capital'' is defined under section 1303(4) of 
the Act (12 U.S.C. 4502(4)) to mean the sum of the following (as 
determined in accordance with generally accepted accounting 
principles):
    (A) The par or stated value of outstanding common stock.
    (B) The par or stated value of outstanding perpetual, 
noncumulative preferred stock.
    (C) Paid-in capital.
    (D) Retained earnings.
    The core capital of an [E]nterprise shall not include any 
amounts that the [E]nterprise could be required to pay, at the 
option of investors, to retire capital instruments.
---------------------------------------------------------------------------

    Use of a stress test will enable OFHEO to tailor carefully the 
Enterprises' capital standards to the specific risks of the 
Enterprises' businesses. It also will provide a structure for 
incorporating interrelationships among different types of risk 
(prepayments, for example, relate to both credit and interest rate 
risk).

Statutory Requirements

    The Act specifies a risk-based capital standard for each 
Enterprise. This standard establishes the amount of capital necessary 
to withstand simultaneously adverse credit and interest rate risk 
scenarios during the stress period plus an additional amount to cover 
management and operations risk, as follows:

Credit Risk

    The Act establishes a credit risk scenario based on a regional 
recession involving the highest rates of default and loss severity 
experienced during a period of at least two years in an area containing 
at least five percent of the total U.S. population. The stress test 
will apply these default and loss rates, with any appropriate 
adjustments, over the ten-year stress period on a nationwide basis to 
the Enterprises' books of business.\3\

    \3\Section 1361(a)(1) (12 U.S.C. 4611(a)(1)).
---------------------------------------------------------------------------

Interest Rate Risk

    The Act presents two interest rate risk scenarios, one with rates 
rising and the other with rates falling. The Act further describes the 
path of the ten-year Constant Maturity Treasury (CMT) yield for each 
scenario and directs OFHEO to establish the yields of other financial 
instruments during the stress period in a reasonably consistent manner. 
The stress test for each Enterprise incorporates the scenario with the 
most adverse impact.\4\

    \4\Section 1361(a)(2) (12 U.S.C. 4611(a)(2)).
---------------------------------------------------------------------------

    In the rising rate scenario, the ten-year CMT yield increases 
during the first year of the stress period and then remains constant at 
the greater of (a) 600 basis points above the average yield during the 
preceding nine months or (b) 160 percent of the average yield during 
the preceding three years. The Act further limits the increase in yield 
to a maximum of 175 percent of the average yield over the preceding 
nine months.\5\

    \5\Section 1361(a)(2)(C) (12 U.S.C. 4611(a)(2)(C)).
---------------------------------------------------------------------------

    In the falling rate scenario, the ten-year CMT yield decreases 
during the first year of the stress period and then remains constant at 
the lesser of (a) 600 basis points below the average yield during the 
preceding nine months or (b) 60 percent of the average yield during the 
preceding three years. The Act further limits the decrease in yield to 
not more than 50 percent of the average yield in the preceding nine 
months.\6\

    \6\Sections 1361(a)(2)(B) (12 U.S.C. 4611(a)(2)(B)).
---------------------------------------------------------------------------

New Business and Other Activities and Considerations

    Initially the stress test assumes that the Enterprises conduct no 
additional new business once the stress period begins, except for the 
fulfillment, in a manner consistent with recent experience and the 
economic characteristics of the stress period, of contractual 
commitments to purchase mortgages and issue securities.\7\

    \7\The Act states that OFHEO may consider the impact of new 
business conducted during the stress period after taking into 
consideration the results of studies conducted by the Congressional 
Budget Office and the Comptroller General on the advisability and 
appropriate forms of new business assumptions. The studies must be 
completed within the first year after the issuance of the final 
risk-based capital regulation. OFHEO may incorporate new business 
into the stress test four years after the regulation is issued. 
Section 1361(a)(3)(C) and (D), (12 U.S.C. 4611(a)(3)(C) and (D)).
---------------------------------------------------------------------------

    The stress test must take into account distinctions among mortgage 
product types, different loan-to-value ratios (LTVs), and any other 
appropriate factors.\8\ OFHEO determines the appropriate consideration 
and treatment of all other factors, activities, or characteristics of 
the stress period not explicitly identified and/or treated in the Act--
such as mortgage prepayments, hedging activities, operating expenses, 
dividend policies, etc.--on the basis of available information, in a 
manner consistent with the stress period.\9\

    \8\Sections 1361(b)(1) and (d) (12 U.S.C. 4611(b)(1) and (d)). 
The Act uses the phrase ``differences in seasoning of mortgages'' 
which is equivalent to differences in LTVs. The term ``seasoning'' 
is defined as the change over time in the ratio of the unpaid 
principal balance of a mortgage to the value of the property by 
which such mortgage loan is secured. Section 1361(d)(1) (12 U.S.C. 
4611(d)(1)).
    \9\Sections 1361(b) and (d)(2) (12 U.S.C. 4611(b) and (d)(2)).
---------------------------------------------------------------------------

Management and Operations Risk

    Finally, to provide for management and operations risk, after 
determining the amount of capital an Enterprise needs to survive the 
stress test, the Act requires OFHEO to increase that amount by 30 
percent to set the required risk-based capital level for each 
Enterprise.\10\

    \10\Section 1361(c)(2) (12 U.S.C. 4611(c)(2)).
---------------------------------------------------------------------------

Philosophy Guiding Stress Test Development

    The mission of OFHEO is to ensure that the Enterprises are 
adequately capitalized and operating in a safe and sound manner, 
consistent with the achievement of their public purposes. The principal 
objective of risk-based capital standards is protection of the taxpayer 
from potential Enterprise insolvency. However, effective capital 
standards should also permit the Enterprises to fulfill their public 
purposes while pursuing prudent business practices and strategies. 
Although the stress test produces a single capital requirement, it 
effectively creates marginal capital requirements--incremental 
requirements for each additional dollar of business--for every type of 
product the Enterprises guarantee or hold in portfolio. Marginal 
capital requirements for mortgages held in portfolio will vary 
depending on the risk, as reflected in the stress test, of an 
Enterprise's funding strategy. These marginal capital requirements will 
have significant bearing on how the Enterprises choose to conduct their 
businesses.
    OFHEO will seek to design the stress test so that the incentives it 
creates closely reflect the relative risks inherent [[Page 7470]] in 
the Enterprises' different activities. To this end, OFHEO will 
incorporate, to the extent feasible, consistent relationships between 
the economic environment of the stress period and the Enterprises' 
businesses. This will require modeling the Enterprises' assets, 
liabilities, and off-balance sheet positions at a sufficient level of 
detail to capture their various risk characteristics. Taking all this 
into consideration will require a balance between the complexity and 
realism of the stress test and its timeliness.

Solicitation of Public Comments

    OFHEO requests public comment on a number of subjects that must be 
addressed in its risk-based capital regulation. OFHEO will consider the 
comments received in response to this ANPR when developing a proposed 
rule. Following consideration of comments on the proposed rule, OFHEO 
will issue a final regulation. When addressing a specific question 
contained in this ANPR, OFHEO asks that commenters specifically note by 
number which question is being addressed.

I. Credit Risk

    The Enterprises face similar mortgage credit risk in their 
portfolio and securitization businesses. OFHEO defines mortgage credit 
risk as the risk of financial loss due to borrower default and 
subsequent foreclosure and liquidation of a mortgaged property. Losses 
are realized when the unpaid loan balance on a defaulted mortgage 
exceeds the net proceeds of a foreclosure sale, after deducting 
carrying and selling costs, less any recoveries from any private 
mortgage insurer, recourse agreement, or other credit enhancements.
    Loans with high current LTVs, where the borrowers have little to no 
equity in their homes, are the most likely to default.\11\ For any 
given set of mortgage loans, the probability of default is typically 
low in the first year after origination, rises to a peak somewhere 
between the third and seventh year, and declines thereafter. If 
declining interest rates induce prepayments on a group of mortgage 
loans due to borrower refinancing activity, defaults and losses on 
those mortgage loans likely will be reduced, because some of the 
prepaid loans would ultimately have defaulted. However, the remaining 
group of loans is likely to be at greater risk of default, because it 
includes all of the original loans where the borrower would not have 
qualified for refinancing, but only some of the loans where the 
borrower was eligible.

    \11\For example, see C. Foster and R. Van Order, ``An Option 
Based Model of Mortgage Default Rick,'' Housing Finance Review, 
3(4):351-372, 1984; C. Foster and R. Van Order, ``FHA Terminations: 
A Prelude to Rational Mortgage Pricing,'' AREUEA Journal, 13(3):273-
291, 1985; and R.L. Cooperstein, F.S. Redburn, and H.G. Meyers, 
``Modelling Mortgage Terminations in Turbulent Times,'' AREUEA 
Journal, 19(4):473-494. For a review of the literature in this area, 
see R.G. Quercia and M.A. Stegman, ``Residential Mortgage Default: A 
Review of the Literature,'' Journal of Housing Research, 3(2):341-
379, 1992.
---------------------------------------------------------------------------

    Economic downturns result in more frequent and severe losses in all 
categories of mortgage loans, especially in a period of house price 
declines. The stress test will incorporate changes in the economic 
environment and simulate the relationship of those changes to mortgage 
defaults.
A. Defining a Stress Benchmark
    The Act, in defining the risk-based capital stress test, refers to 
two time periods--a hypothetical ten-year ``stress period'' during 
which the Enterprises' capital should be sufficient to absorb losses 
and maintain a positive capital level while being subjected to adverse 
credit and interest rate risk scenarios, and the time period of ``not 
less than two years'' for which the ``highest rates of default and 
severity of mortgage losses'' occurred in a region containing at least 
five percent of the total population of the United States.\12\ For the 
purposes of this ANPR, OFHEO characterizes the latter time period and 
region as a ``stress benchmark.'' The stress benchmark will provide the 
basis for the development of the credit risk stress scenario that will 
be applied during the ten-year stress period.

    \12\Section 1361(a)(1) (12 U.S.C. 4611(a)(1)).
---------------------------------------------------------------------------

    The Act permits the identification of one or more stress 
benchmarks. A single benchmark is conceptually appealing but presents a 
number of difficult issues. A single benchmark may not include 
sufficient data on all Enterprise product types. Patterns of 
multifamily and single family mortgage losses differ (see ``Mortgage 
Types'' below) and a stress benchmark for multifamily mortgages 
representing the worst regional experience for those mortgages may not 
coincide with the benchmark for single family mortgages based on their 
worst experience. Finally, data limitations may prevent OFHEO from 
determining loss severities during the period of highest default rates; 
alternatively, highest loss severities may not coincide with highest 
rates of default by time period or region.
    Although the Act does not refer to a particular mortgage product in 
its reference to ``highest rates of default and severity,'' single 
family, 30-year, fixed-rate mortgages have long comprised the bulk of 
Enterprise mortgages. OFHEO expects to define a stress benchmark for 
these mortgages on the basis of a weighted average (by unpaid loan 
balance of various LTV groups) of default rates.
    Existing data on loss severities may be inadequate to contribute to 
establishing the timing or location of the worst regional experience. 
Systems for the storage and analysis of data on foreclosed properties 
are a relatively recent development. To overcome these data 
deficiencies, OFHEO will consider a number of approaches to determining 
loss severity rates during the stress benchmark. These approaches 
include the use of loss severity estimates obtained from different 
sources and for different time periods and regions than those used to 
estimate the benchmark default rates.
    OFHEO may use models (see ``Models of Default and Prepayment'' and 
``Models of Loss Severity'' below) to establish aspects of the 
benchmark for which data are insufficient or unavailable. These might 
include, in addition to loss severities for all products, default rates 
for mortgage products poorly represented or non-existent in the stress 
benchmark. Econometric models for default, mortgage prepayment, and 
loss severity would facilitate consideration of the simultaneous impact 
of many factors on default rates, such as changes in LTVs, the impact 
of contemporaneous prepayments, and the impact of factors associated 
with mortgage product types. Models would provide a link between the 
performance of mortgages owned or guaranteed by the Enterprises during 
the stress period and performance during the stress benchmark, with due 
consideration of the economic circumstances of the stress period, e.g., 
interest rates and house prices.

Data Issues

    OFHEO has received access to detailed information about the loss 
experience on mortgages that the Enterprises owned or guaranteed from 
the mid-1970s through the present. The type of information on mortgages 
that OFHEO needs to develop the stress test includes date of 
origination, original LTV ratio, type of mortgage, location, nature and 
degree of any credit enhancements, date of last paid installment, 
termination type, e.g., default or prepayment, and the amount of any 
ultimate loss (including holding and selling costs). However, there are 
serious gaps in the data on loss severity through the early 1980s 
resulting from the lack of systems for the storage and 
[[Page 7471]] analysis of data on foreclosed properties and the manner 
in which loan balances were reported by seller/servicers.
    In general, however, with the increase over time of the 
Enterprises' share of the overall mortgage market, the data grow 
increasingly rich. If necessary, OFHEO could supplement these data with 
data from the Federal Housing Administration or other sources such as 
TRW Redi and Mortgage Information Corporation.
    If the stress benchmark is wholly or primarily based on Enterprise 
data, the loan-level data could be aggregated across the two 
Enterprises in order to determine the worst historical experience. 
Preliminary analysis suggests that the worst historical experience may 
be different for the two Enterprises. An alternative would be to 
determine the worst historical experience for each Enterprise 
separately and then use a simple or weighted average of default rates.
    Question 1: What data and methodology should OFHEO use in its 
determination of the stress benchmark?

Benchmark Time Period and Region

    OFHEO has considered at least two approaches for defining the 
benchmark time period. It could be defined as the period in which the 
highest rates of default occurred, that is, an ``exposure year'' 
approach; or the period in which the loans with the highest cumulative 
or lifetime rates of default were originated, which can be termed an 
``origination year'' approach. At the start of the stress period, the 
Enterprises' books of business will include survivors from many loan 
origination years. An exposure year benchmark corresponds more closely 
to the manner in which the Enterprises' mortgage portfolios will 
experience the risk of credit losses as they move through the ten-year 
stress period. However, using exposure years may complicate adjustments 
for differences in LTVs and other factors (see ``Relating Stress Period 
Default Rates to Benchmark Default Rates'' below). Using origination 
years may require some adjustment for differences in mortgage age (see 
``Mortgage Age'' below) since virtually all of the Enterprise mortgages 
will have been originated prior to the start of the stress period.
    Alternative approaches to defining the stress benchmark (exposure 
year versus origination year) suggest alternative analyses of defaults. 
An exposure year approach requires the determination of default rates 
on loans of varying age at risk of failure within a specified period. 
The resulting time-period specific default rates for loans outstanding 
at the beginning of the period can be termed ``conditional rates.'' 
Because default rates vary with the age of a mortgage (see ``Mortgage 
Age'' below), OFHEO might define an age schedule of conditional default 
rates for loans outstanding at the start of the stress benchmark.\13\ 
For comparison across time periods and regions, synthetic cumulative 
default rates for the stress benchmark could be derived under a common 
set of prepayment assumptions. In an origination year approach, either 
cumulative or conditional default rates could be used.

    \13\Age is often a proxy for additional unobserved factors 
affecting the default probabilities of individual mortgages. 
Immediately after origination, default is unlikely for all 
borrowers. Default rates first rise over time as new information 
about properties and borrowers is revealed. Then as relatively 
weaker borrowers default, the average rate of default declines. See, 
for example, the discussion in C. Pestre, P. Richardson, and C. 
Webster, ``The Lehman Brothers Mortgage Default Model and Credit-
Adjusted Spread Framework,'' Mortgage Market Analysis, Lehman 
Brothers, Fixed Income Research, January 28, 1992. Other influential 
default studies that have included mortgage age as an explanatory 
factor include: T. Campbell and J. Dietrich, ``The Determinants of 
Default on Conventional Residential Mortgages,'' Journal of Finance, 
38(5):1569-1581, 1983; D. Cunningham and C. Capone, ``The Relative 
Termination Experience of Adjustable to Fixed-Rate Mortgages,'' The 
Journal of Finance, 45(5):1687-1703, 1990; and J.M. Quigley and R. 
Van Order, ``More on the Efficiency of the Market for Single Family 
Homes: Default,'' Center for Real Estate and Urban Economics, 
University of California, Berkeley, 1992.
---------------------------------------------------------------------------

    The Act requires that the benchmark region comprise a contiguous 
area containing at least five percent of the total United States 
population. Part or all of states such as Texas or California satisfy 
this population requirement; however, areas experiencing the highest 
rates of default may cross over one of these state's boundaries into 
adjoining states. As appropriate, OFHEO will use a definition of 
benchmark region that includes more than one state, part of one state, 
or parts of several states.
    Question 2: How should the benchmark time period be defined?

Measurement of Default

    Default can be defined in several ways: Defaults can be deemed to 
occur at the time a borrower ceases making payments, when a loan 
payment is past due by a contractually specified number of days, on the 
date of foreclosure, or on the date when losses are recognized. 
Defaults can be measured on a gross basis or net of any subsequent 
cures.
    Question 3: What are the relative merits of the alternative 
approaches for the measurement of mortgage defaults?
B. Relating Stress Period Default Rates to Benchmark Default Rates
    Default rates during the stress period may differ from the default 
rates associated with the stress benchmark. This difference may result 
from differences between the characteristics and composition of an 
Enterprise's mortgages at the start of the stress period relative to 
those of the mortgages identified with the stress benchmark. Stress 
period default rates may also differ from stress benchmark rates as a 
result of differences in the stress period environment, such as 
interest rates and inflation. OFHEO must also specify the timing of 
defaults and losses during the stress period.
    The Act requires that OFHEO, in establishing the stress test, take 
into account appropriate distinctions among types of mortgage products, 
differences in LTVs, and other factors that OFHEO's Director considers 
appropriate.\14\ Such factors include prepayment activity, mortgage 
age, and loan size. The Act also requires an adjustment for the effects 
of general inflation in the highest interest rate environment in the 
stress test.\15\

    \14\Section 1361(b)(1) (12 U.S.C. 4611(b)(1)).
    \15\Section 1361(a)(2)(E) (12 U.S.C. 4611(a)(2)(E)).
---------------------------------------------------------------------------

Loan-to-Value Ratios

    The payment of principal and changes in the value of the property 
securing a mortgage affect LTVs over time. Repayments of loan principal 
and rising property values lower LTVs, while falling property values 
raise LTVs. Because LTV is a common measure of borrower equity, and 
borrower equity is a major factor determining defaults and losses, the 
stress test must take into account changes in LTVs. If distributions of 
LTVs during the stress period differ from those for the same types of 
loans associated with the stress benchmark, defaults and losses during 
the stress period will likely differ from those of the benchmark.
    All loans owned or guaranteed by the Enterprises at the start of 
the stress period will have been originated prior to that time. 
Although relatively good estimates of property value are available at 
the time of loan origination, OFHEO will need to use house price 
indexes to obtain estimates of the LTVs for mortgages at the start of, 
and possibly throughout, the stress period.\16\ OFHEO 
[[Page 7472]] intends to use a repeat sales index based on sales (or 
appraisals undertaken by borrowers in conjunction with refinancing the 
mortgages) of the Enterprises' owned and guaranteed portfolios (see 
``House Price Indexes'' below).

    \16\For an origination year benchmark, OFHEO will likely have 
access to accurate information about the original LTVs for all 
benchmark loans. On the other hand, to develop an exposure year 
benchmark, OFHEO will have to estimate LTVs during the benchmark 
time period for all loans originated earlier. OFHEO would use house 
price indexes for this purpose.
---------------------------------------------------------------------------

    Models of mortgage default and prepayment (see ``Models of Default 
and Prepayment'' below) emphasize the importance of LTV because of its 
direct relationship to homeowner's equity, defined as the difference 
between the value of a property and the outstanding principal balance 
of the related mortgage. These models differ in their treatment of 
house price changes and with regard to how changes in equity affect 
default and prepayment. For example, one approach assumes that defaults 
occur only among loans with negative equity.\17\ House price indexes 
only provide estimates of the average change in property values between 
two dates. Because changes in individual property values are not 
continuously observed, simulation models have been used to characterize 
the distribution of changes in house prices relative to the market 
average. Estimates of the percentage of loans with negative equity and 
estimates of default rates can be derived from these distributions.

    \17\See Foster and Van Order, supra, (1984, 1985).
---------------------------------------------------------------------------

    This approach assumes that homeowner's equity includes not just the 
difference between property value and outstanding loan amount, but also 
the current value of the mortgage to the borrower. A below-market rate 
loan has positive value. The precise value of the mortgage depends on 
the loan interest rate relative to the current market rate and the 
borrower's expectations about future interest rates and mobility. A 
borrower whose loan has a fixed contract rate below current market 
yields has more to lose by defaulting than a borrower with a note rate 
above the current market rate.
    Question 4: What is the appropriate way in which to adjust the LTVs 
of mortgages in the stress test?
    Question 5: If estimates of the distribution of house price changes 
are used to adjust the LTVs of mortgages, what is an appropriate 
method, e.g., stochastic process?
    Question 6: In what manner, if at all, should OFHEO incorporate 
mortgage value as a factor affecting defaults?

Mortgage Types

Single Family

    The Act requires that the stress test consider differences in 
mortgage types (single family or multifamily, fixed or adjustable rate, 
first or second lien, owner-occupied or investor owned, positive or 
negative amortization, alternate term to maturity, etc.).\18\ Risk 
characteristics of different types of mortgages vary considerably. 
Because of the fundamental differences between single family and 
multifamily mortgage risk, we discuss the latter in a separate section 
below.

    \18\Sections 1361(b)(1) and (d)(2) (12 U.S.C. 4611(b)(1) and 
(d)(2)).
---------------------------------------------------------------------------

    Given that OFHEO plans to establish the stress benchmark based on 
single family, 30-year, fixed-rate mortgages, the Act calls for OFHEO 
to identify the worst rates of default and losses for any time period 
or region.\19\ The Enterprises may not have held certain types of 
single family mortgages in the stress benchmark OFHEO identifies. Other 
types of single family mortgages held during the stress benchmark may 
have experienced their worst defaults and losses at other times or in 
other regions.

    \19\Section 1361(a)(1) (12 U.S.C. 4611(a)(1)).
---------------------------------------------------------------------------

    Alternative approaches could include use of multivariate models to 
estimate separate equations for different mortgage products or 
different mortgage features, default rates representing some multiple 
of the standard single family mortgage, or some combination of these 
approaches (see ``Models of Default and Prepayment'' below).
    Question 7: How should OFHEO relate other types of mortgages to a 
single stress benchmark developed based on single family, 30-year, 
fixed-rate mortgages?

Multifamily

    While single family properties are both a source of shelter and, 
for most families, their most valuable financial asset, multifamily 
properties are primarily income-producing businesses for their owners. 
Multifamily loans are less homogeneous and subject to a more diverse 
set of risks than single family loans. The multifamily market has more 
pronounced business cycles and is heavily affected by tax and 
regulatory policy. Patterns of losses over time for multifamily loans 
have not tracked those of the single family market. The Enterprises 
operate several different types of multifamily programs, some of which 
rely heavily on lender recourse or other forms of credit enhancement 
with differing risk characteristics.
    Data needs in analyzing multifamily loans are greater than for 
single family loans and yet the quality of such data is poorer. Data 
are incomplete and cover a smaller portion of the multifamily market 
than the single family market. There is also a dearth of research on 
critical multifamily credit risk issues.
    For the owner of a multifamily property, net operating income (NOI) 
plays a more important role than equity in the decision to default. A 
property's debt service coverage, rather than LTV ratio, may be the 
most important indicator of multifamily credit risk, yet available data 
can only provide a short time-series for income. Multifamily value 
indexes are problematic because there are fewer transactions than in 
the single family market and property appraisals are less reliable. 
Appraisals are less reliable due to the varying methodologies used to 
calculate multifamily property income and the application of so-called 
``capitalization rates'' to NOI.\20\

    \20\Government Accounting Office, ``Federal Home Loan Mortgage 
Corporation: Abuses in Multifamily Program Increase Exposure to 
Financial Losses'' (Oct. 1991); J.M. Abraham, ``On the Use of a Cash 
Flow Time-Series to Measure Property Performance,'' forthcoming in 
Journal of Real Estate Research; and J.M. Abraham, ``Credit Risk in 
Commercial Real Estate Lending, ``Federal Home Loan Mortgage 
Corporation, 1994 presented at the 1994 meetings of the American 
Real Estate and Urban Economics Association (available from OFHEO).
---------------------------------------------------------------------------

    Prepayments play a far less significant role in the analysis of 
multifamily credit risk than single family credit risk because 
``lockouts'' and yield maintenance agreements effectively prevent most 
multifamily borrowers from refinancing to take advantage of declining 
interest rates. The Enterprises' activity in the multifamily market is 
expected to increase significantly in future years in order to meet the 
affordable housing goals established under the Act.\21\ Thus, the 
treatment of multifamily risks will be increasingly important.

    \21\Sections 1331-1336 (12 U.S.C. 4561-4566).
---------------------------------------------------------------------------

    Question 8: How should existing and emerging multifamily data 
sources be identified?
    Question 9: What are alternative empirical and theoretical 
approaches to the estimation of multifamily credit risk?
    Question 10: How should the projection of defaults and losses on 
the Enterprises' multifamily portfolio be related to a single family 
stress benchmark?

General Price Inflation

    The Act requires that OFHEO adjust credit losses in the stress test 
when large increases in interest rates imply higher rates of general 
price inflation.\22\ If the ten-year CMT yield is assumed to increase 
by more than 50 percent over the average yield during the preceding 
[[Page 7473]] nine months, inflation is presumed to be 
``correspondingly higher.'' If, for example, the ten-year CMT yield 
were to have averaged eight percent during the past nine months, a 50 
percent increase would raise it to 12 percent. The Act, however, would 
permit an increase to 14 percent.

    \22\Section 1361(a)(2)(E) (12 U.S.C. 4611(a)(2)(E)).
---------------------------------------------------------------------------

    OFHEO would first determine what annual percentage difference in 
general inflation rates best corresponds to the difference between a 12 
percent and a 14 percent ten-year CMT yield over a nine-year period. 
The difference in inflation rates could be assumed to be equal to the 
difference in interest rates or it could be based on an estimated 
historical relationship.
    OFHEO would then translate that higher inflation rate into 
individual house price changes. Again, the differences in house price 
changes could be assumed to be equal to the difference in general price 
inflation rates or could be based on an estimated relationship.
    As the last step, OFHEO would translate the difference in house 
price changes into differences in defaults. This could be done in the 
context of a multivariate default and prepayment model used for making 
many adjustments simultaneously (see ``Models of Default and 
Prepayment'' below), or it could be the subject of a separate analysis.
    Question 11: Should OFHEO assume a ``one-to-one'' relationship 
between long-term differences in interest rates, general price 
inflation rates, and house price inflation rates or should it estimate 
more complex, but potentially more realistic, relationships between 
these phenomena?
    Question 12: What is the best method of modeling the effects of 
higher house prices on defaults?

Mortgage Prepayments--Credit Risk

    Prepayments are a significant factor in interest rate risk, but 
they also affect credit losses. Interest rate changes have a 
significant influence on mortgage prepayments. Prepayment rates are 
sensitive to the differences between current market yields and the 
levels of mortgage rates among outstanding mortgages. A homeowner today 
will refinance (and prepay) when current mortgage rates fall as little 
as 50 basis points below the rate on his or her mortgage.
    Prepayment rates also depend on the time paths of interest rates. 
Homeowners who fail to refinance once mortgage rates become 
advantageous are relatively unlikely to do so in the future (many may 
not qualify for refinancing). Thus, prepayment rates for mortgages with 
a given coupon rate rise as interest rates fall below a particular 
threshold, but they eventually will slow, even if interest rates remain 
at the new lower levels or continue to decline. This phenomenon is 
commonly known as ``burn-out.''
    The expected pattern of prepayments in the stress period might be 
quite different from the pattern experienced during the benchmark 
period. The drastic yield curve shifts that will be experienced during 
the initial year of the stress period will almost certainly not be 
found during the benchmark period that OFHEO must identify. The greater 
number of mortgages that prepay, the fewer are the candidates for 
subsequent default. Conversely, the fewer mortgages that prepay, the 
greater the number remaining that might default. At the same time, the 
default risk of mortgages remaining after a refinancing wave may be 
higher than previously. Many homeowners who did not take advantage of 
attractive refinancing opportunities may have been unable to do so 
because of higher risk profiles. Given the widely divergent interest 
rate movements that the Enterprises may experience during the stress 
period, loss adjustments for differing prepayment behavior could be 
considerable.
    If OFHEO expresses mortgage default rates as conditional rates, 
defaults during any given time interval of the stress period will 
depend on the proportion of mortgages outstanding at the beginning of 
that time interval. Such an approach would, in effect, make a 
substantial adjustment for prepayments. A more complicated adjustment 
would take into account the generally higher quality of loans eligible 
for refinancing. In a stress scenario involving falling interest rates, 
for example, the stress test might take into account the generally 
higher quality of loans that qualify for refinancing and the 
potentially lower quality of surviving loans (see ``Models of Default 
and Prepayment'' below). Alternatively, if the stress test involves no 
interaction of the total amount of defaults and prepayments, OFHEO 
still might adjust the timing of defaults during the stress period to 
be consistent with prepayments expected in a particular interest rate 
scenario. Mortgage prepayments are discussed further under ``Interest 
Rate Risk'' below.
    Question 13: Should anticipated prepayments affect the volume or 
timing of defaults in the stress period?

Mortgage Age

    Holding homeowner's equity constant, a number of factors make the 
likelihood of borrower default vary over the life of a loan. On one 
hand, changes in a borrower's circumstances subsequent to the loan's 
origination, such as unemployment, marriage, divorce, childbearing, 
mortality, and residential mobility, affect the likelihood of default 
and prepayment, and the cumulative frequency of such events increases 
as a loan ages. On the other hand, a record of consistent payments by a 
borrower over time increases the probability of continued loan 
performance.
    Models that have included variables for both homeowner's equity and 
mortgage age have found the contribution of age to be statistically 
significant.\23\ This may be particularly important if an origination 
year approach is used in the benchmark. Using an origination year 
approach, loans in the stress benchmark would all be newly originated 
loans, while those at the beginning of the stress period would be a 
mixture of old and new loans.

    \23\For example, see the papers cited in footnote 11 above.
---------------------------------------------------------------------------

    Question 14: Is it appropriate for OFHEO to factor mortgage age 
into the stress test, and, if so, what is the best method of doing so?
C. Models of Default and Prepayment
    There are a number of approaches to relating the factors discussed 
above, such as LTV, mortgage type, mortgage age, and prepayments, to 
the performance of the Enterprises during the stress period. A 
comprehensive way to incorporate all of these factors into the stress 
test would be to estimate joint multivariate models of default and 
prepayment.\24\ A joint model of default and prepayment would ensure 
the consistency of these key variables and reflect an appropriate time 
pattern of defaults as well. Researchers have estimated a number of 
such models.\25\

    \24\Due to the unique difficulties of modeling multifamily 
default and prepayment, multifamily and single-family loans would 
probably need to be modeled separately. The modeling of loss 
severity is discussed in the next section.
    \25\Multinomial logit models for default have been estimated by 
Campbell and Dietrich (1983) supra; P. Zorn and M. Lea, ``Mortgage 
Borrower Repayment Behavior: A Microeconomic Analysis with Canadian 
Adjustable Rate Mortgage Data, AREUEA Journal, 17(1):188-136, 1989; 
and Cunningham and Capone (1990) supra. More recently, proportional 
hazards models have been used to analyze default and prepayment. 
See, for example, J. Quigley, ``Interest Rate Variations, Mortgage 
Prepayments and Household Mobility, Review of Economics and 
Statistics, 119(4):636-643, 1987; and J.M. Quigley and R. Van Order, 
``More on the Efficiency of the Market for Single Family Homes: 
Default,'' Center for Real Estate and Urban Economics, University of 
California, Berkeley, 1992. [[Page 7474]] 
---------------------------------------------------------------------------

    A joint approach to default and prepayment would generate default 
rates reasonably related to the stress benchmark, while simultaneously 
generating prepayment rates that are consistent with the interest rate 
characteristics of the ten-year stress period. To estimate a 
multivariate default/prepayment model, OFHEO could draw on all relevant 
historical data, not just data from the stress benchmark. The model 
might include explanatory variables such as LTVs at origination, 
current LTVs (determined through the application of an appropriate 
house price index), differences between actual mortgage coupons and 
current market rates, interest rate paths, mortgage age, dummy 
variables for time period and location of mortgaged property, and 
additional characteristics specific to different mortgage products. The 
estimation procedure could allow for changing coefficients over time to 
reflect structural changes in prepayment and default behavior. During 
the stress period, explanatory or dummy variables, reflecting the 
special circumstances of the stress benchmark, would be set at their 
benchmark levels.
    While multivariate models allow for the most realistic estimates of 
defaults and prepayments, OFHEO recognizes the difficulties of such an 
approach. Insufficient data may complicate model selection and the 
estimation of some individual parameters. One of the most simple 
approaches would be to measure cumulative defaults in the stress 
benchmark for the most common 30-year, fixed-rate, 80 percent LTV 
mortgages and then spread those defaults evenly or according to some 
predetermined pattern over the ten-year stress period, with no 
consideration of prepayments. Losses on other mortgage types and LTVs 
could be set at simple multiples of the ``standard'' loss rate based on 
average historical experience. All other possible variables might be 
ignored.
    Many approaches of intermediate complexity exist. For example, 
OFHEO could determine the stress benchmark default rates for standard 
30-year, fixed-rate, single family mortgages for several LTV categories 
and a few other types of mortgages. Relative defaults on additional 
mortgage types would be determined from more recent data using 
multivariate models, which would also provide adjustment factors for 
some mortgage features and other relevant variables. Prepayments could 
be modeled separately, affecting projected defaults by changing the 
volume of surviving loans (See ``Mortgage Prepayments--Interest Rate 
Risk'' below). The time patterns of defaults could also be modeled 
separately as a function of mortgage age.
    Question 15: What are the relative merits of using a joint model of 
default and prepayment in the stress test?
    Question 16: What is an appropriate statistical method for 
estimating a joint model of default and prepayment?
    Question 17: Should defaults be expressed in terms of conditional 
failure rates (hazards), cumulative default rates, or in some other 
manner?
    Question 18: What explanatory variables should be included in a 
statistical model for default and prepayment?
    Question 19: What is an appropriate level of statistical 
aggregation for the estimation of a joint model of default and 
prepayment?
    Question 20: How should the impact of house price trends, interest 
rates, and other economic factors be incorporated into a model of 
default and prepayment?
D. Models of Loss Severity
    Due to the varying quality of data on losses on defaulting loans, 
OFHEO may be unable to establish actual loss severities for the stress 
benchmark. Even if loss severities are incorporated in the stress 
benchmark, OFHEO may make adjustments to reflect changes in factors 
that affect loss severities. Consequently, OFHEO will conduct a 
separate analysis of loss severity based on all available data. This 
section examines some of the issues involved in modeling loss severity, 
including approaches for linking loss severity rates to the stress 
benchmark.
    Loss severity refers to the actual dollars lost on a defaulted loan 
and allows credit risk to be quantified in dollar terms. Severity is 
the extent to which the costs associated with default, foreclosure, and 
disposition exceed the revenues associated with these processes. The 
major costs are the loss of loan principal, transaction costs at both 
foreclosure and disposition, and carrying costs throughout the process. 
The major revenues are foreclosure sale price and mortgage insurance 
payments.
    Loss severity, like default, depends on numerous factors. Some 
factors--original LTV ratio, LTV ratio at time of default, original 
loan size, occupancy status, type of structure, and presence or absence 
of mortgage insurance--are the factors that also influence the 
likelihood of default. Other factors--methods of disposition, state 
foreclosure laws, and home price movements after default--influence 
severity without affecting the likelihood of default.\26\

    \26\See, for example, T. Clauretie and T.N. Herzog, ``How State 
Laws Affect Foreclosure Costs,'' Secondary Mortgage Markets, 
6(Spring):25-28, 1989; T. Clauretie and T.N. Herzog, ``The Effect of 
State Foreclosure Laws on Loan Losses: Evidence from the Mortgage 
Insurance Industry,'' Journal of Money, Credit, and Banking, 
22(2):221-233, 1990; E. Bruskin and M. Buono, ``A New Understanding 
of Loss Severity: Time is (of) the Essence,'' in Mortgage Securities 
Research, Goldman-Sachs, September 1994; and V. Lekkas, J. Quigley, 
and R. Van Order, ``Loan Loss Severity and Optimal Mortgage 
Default,'' AREUEA Journal, 21(4):353-371, 1993.
---------------------------------------------------------------------------

    OFHEO is considering using a multivariate statistical model to 
estimate the separate effects of these factors on severity. OFHEO may 
develop a separate model for each of the cost and revenue components of 
loss severity since each component is affected by different factors. In 
the event that data on the individual revenue and cost components of 
loss severity are unavailable, an alternative approach would be to 
model overall loss severity directly.
    Another less complex option is to estimate the individual 
components without multivariate statistical analysis. OFHEO could set 
fixed parameters for the components of severity--foreclosure costs 
might be x percent of unpaid principal balance (UPB), carrying costs 
equal to y percent of UPB and sales prices being z percent of UPB--
while allowing severity to vary based on, for example, the presence or 
absence of private mortgage insurance or state foreclosure laws. The 
simplest possible option would be to assume that all defaulted loans 
face the same level of severity as a percentage of UPB.
    There are a number of ways in which rates of loss severity may be 
related to the stress benchmark rates of default and the corresponding 
rates of default during the stress period. Given the impact of state 
foreclosure laws on loss severity, default rates and loss severity will 
be linked through the geographic location of the mortgages. For 
example, loss severities are likely to be lower in states where 
foreclosure laws are relatively more favorable to the lender.
    The assumptions about changes in house prices in the stress 
benchmark and during the stress period will affect the determination of 
foreclosure sales prices and loss severity. Defaults are more likely to 
have occurred when borrowers' properties have appreciated much less 
than the average for their region. This implies that house price 
indexes used to model loss severity would best be based on properties 
that have experienced lower than average appreciation. [[Page 7475]] 
    Question 21: What are the explanatory factors OFHEO should consider 
in modeling loss severity?
    Question 22: Should OFHEO model the individual cost and revenue 
components of severity or should OFHEO model only overall severity?
    Question 23: What is an appropriate house price index for real 
estate owned (REO) properties? In estimating foreclosure sales prices, 
should OFHEO use a house price index based on all properties or a house 
price index based only on REO properties?
E. House Price Indexes
    The Act requires that OFHEO use house price indexes to determine 
changes in the values of properties securing mortgages owned or 
guaranteed by the Enterprises and the corresponding changes in LTVs. 
Changes in property values are--

determined on an annual basis by region, in accordance with the 
Constant Quality Home Price Index published by the Secretary of 
Commerce (or any index of similar quality, authority, and public 
availability that is regularly used by the Federal Government).\27\

    \27\Section 1361(d)(1) (12 U.S.C. 4611(d)(1)).

    Since the second quarter of 1994, the Enterprises have published 
the quarterly Conforming Mortgage House Price Index (CMHPI) for the 
nine Census divisions. This represents a significant improvement over 
the annual four Census region Commerce Constant Quality Index (CCQI). 
The CMHPI is based on a weighted repeat sales (WRS) approach in which 
multiple transactions, i.e., mortgage originations, for individual 
properties are matched by street address to obtain changes in sales 
prices or appraisal values. Observed property values and transactions 
dates are then combined in a multivariate statistical model to estimate 
an index of housing values.\28\

    \28\See W. Stephens, Y. Li, V. Lekkas, J. Abraham, C. Calhoun, 
and T. Kimner, ``Agency Repeat Transactions,'' revised August 1994, 
forthcoming in Journal of Housing Research (available from OFHEO).
---------------------------------------------------------------------------

    OFHEO believes that a WRS index based on Enterprise data offers a 
number of advantages for estimating the changing LTVs of the 
Enterprises' mortgage assets. Perhaps foremost among these is the 
direct correspondence between index data and the housing segment 
serviced by the Enterprises. This factor, along with others, should 
make the index more accurate for establishing the current market values 
of properties securing mortgages held or guaranteed by the Enterprises. 
In addition, a WRS index based on Enterprise data will allow OFHEO to 
estimate changes in housing values at lower levels of geographic and 
temporal aggregation, and with greater statistical precision, than the 
CCQI allows. In order to meet the requirements of the Act regarding the 
use of an alternative house price index, OFHEO will produce and publish 
a similar house price index or indexes using data on the historical 
mortgage transactions of the Enterprises.
    Issues that have a bearing on the application of house price 
indexes to the risk-based capital test include the appropriate level of 
geographic aggregation, sample selection and appraisal bias, and the 
effect of index revisions as new data becomes available.\29\

    \29\Methodological issues related to the estimation of repeat 
transaction house price indexes are discussed in the following 
papers: M.J. Bailey, R.F. Muth, and H.O. Nourse, ``A Regression 
Method of Real Estate Price Index Construction,'' Journal of the 
American Statistical Association, 58:933-942, December 1963; K.E. 
Case and R.J. Shiller, ``Prices of Single-Family Homes since 1970: 
New Indexes for Four Cities,'' New England Economic Review, 45-56, 
September/October 1987; K.E. Case and R.J. Shiller, ``The Efficiency 
of the Market for Single Family Homes,'' American Economic Review, 
79:125-137, 1989; J.M. Abraham, J.M. and W.S. Schauman, ``New 
Evidence on Home Prices from Freddie Mac Repeat Sales,'' Journal of 
the American Real Estate and Urban Economics Association, 19:333-
352, 1991; C.A. Calhoun, ``Estimating Changes in Housing Values from 
Repeat Transactions,'' Federal National Association International 
meetings (available from OFHEO); and C.A. Calhoun, P. Chinloy, and 
I.F. Megbolugbe, ``Temporal Aggregation and House Price Index 
Construction,'' Federal National Mortgage Association, forthcoming 
in Journal of Housing Research (available from OFHEO); and B. Case, 
H.O. Pollakowski, and S.M. Wachter, ``On Choosing Among House Price 
Index Methodologies,'' Journal of the American Real Estate and Urban 
Economics Association, 19(3):286-307, 1991.
---------------------------------------------------------------------------

Geographical Aggregation

    Aggregation across housing markets with imperfectly correlated 
house price changes will result in biased estimates of the average 
levels of appreciation in individual markets. This bias can be 
characterized in terms of the smoothing of market-wide indexes, with a 
corresponding increase in the apparent volatility of individual house 
prices around the market index. Excessive disaggregation, however, may 
reduce the frequency at which indexes can be meaningfully computed and 
subject them to large revisions.
    Question 24: What principles should OFHEO use in selecting the 
optimal level of geographic aggregation for the stress test?

Bias

    As discussed below, potential sources of statistical bias include 
sample selection bias and appraisal bias.

Sample Selection Bias

    Even within the total database of Enterprise mortgages, non-random 
sampling of individual properties with repeat transactions could result 
in an index that is biased for the larger population of Enterprise 
properties. For example, the conforming loan limit and year-to-year 
changes in the limit could result in sample selection bias in the 
estimated parameters of a repeat transactions index. A closely related 
form of sample selection bias can occur when the waiting time between 
repeat transactions is correlated with the change in house prices. For 
example, if more rapidly appreciating properties turn over within 
shorter time intervals, they will appear in the repeat sample more 
quickly. In this case, appreciation rates for repeat transactions near 
the end of the sample period will not be representative. Thus, sample 
selection bias would be greater near the end of the index.

Appraisal Bias

    Approximately 85 percent of the repeat transactions used by the 
Enterprises to estimate WRS house price indexes involve a refinance 
transaction.\30\ Appraisals provide useful information on house values 
in the absence of sales transactions. However, the use of appraisals in 
real estate valuation is thought to impart bias by smoothing the 
fluctuations in housing values. Appraisals are derived through 
comparisons with properties that have either been sold or listed for 
sale within the past several months and may fail to indicate more 
recent changes in housing values.

    \30\See Stephens, et al., supra.
---------------------------------------------------------------------------

    Question 25: Should house price indexes estimated using Enterprise 
data include adjustments for identifiable sources of statistical bias?
    Question 26: What additional sources of statistical bias exist and 
what are possible corrective actions that may be taken to address them?
    Question 27: What methods of accounting and correcting for sample 
selection bias should be used?
    Question 28: Should a statistical adjustment to the WRS house price 
index be made to address the impact of appraisal bias?

Revision Volatility

    As data on new transactions are obtained each quarter, new repeat 
transactions can be combined with transactions that occurred in the 
past. Thus, the quarterly index estimation process involves the 
revision of the entire index in light of new information. 
[[Page 7476]] Depending on the level of geographic aggregation, this 
can result in substantial changes in historical values of the index and 
the implied changes in the LTVs of Enterprise mortgages.
    Question 29: Should changes in WRS indexes resulting from revision 
volatility be reflected in indexes used in a stress test? If so, what 
should be the frequency of such revisions?
F. Third Party Credit Issues
    The Enterprises have credit exposure to institutions that provide 
mortgage credit enhancements or that serve as counterparties to 
derivative transactions. This exposure arises because the adverse 
economic environment of the ten-year stress period may cause some 
fraction of these institutions to fail and be unable to meet their 
financial obligations to the Enterprises.

Credit Enhancements

    The Enterprises reduce their exposure to mortgage credit losses 
through a variety of credit enhancements that transfer some or all of 
the risk to other parties. These credit enhancements include lender 
recourse, mortgage insurance, and pool insurance.
    The use of mortgage insurance illustrates how credit enhancements 
work to mitigate credit losses and highlights some of the issues OFHEO 
must address. Generally, the Enterprises may not purchase a 
conventional mortgage whose LTV ratio exceeds 80 percent unless the 
seller retains a participation interest or enters into a repurchase 
agreement, or unless the mortgage is insured by a qualified 
insurer.\31\ If insured mortgages experience actual losses, the 
insurance fully or partially compensates the Enterprises for those 
losses.

    \31\Federal National Mortgage Association Charter Act, section 
302(b)(2) and (5)(C) (12 U.S.C. 1717(b)(2) and (5)(C)), and Federal 
Home Loan Mortgage Corporation Act, section 305(a)(2) and (4)(C) (12 
U.S.C. 1454(a)(2) and (4)(C)).
---------------------------------------------------------------------------

    Applying an approach used by credit rating agencies for private 
mortgage insurers, some insurers may be assumed to go out of business 
during the stress period.\32\ To reflect this possibility, OFHEO's 
stress test might assume the failure of some fraction of the private 
mortgage insurers who would then be unable to entirely fulfill their 
contractual obligations to the Enterprises.

    \32\``S&P's Structured Finance Criteria,'' Standard & Poor's 
(1988).
---------------------------------------------------------------------------

    Question 30: How should OFHEO calculate loss mitigation due to 
credit enhancements?
    Question 31: What should OFHEO assume about the scope of coverage 
provided by credit enhancements?
    Question 32: What assumptions should OFHEO make regarding the 
failure of credit enhancements over the stress period?

Derivatives Counterparties

    The Enterprises use non-mortgage derivatives--interest rate and 
foreign exchange rate contracts--to hedge interest rate and foreign 
exchange rate risk. Should a counterparty default on its obligation 
under a derivative contract, an Enterprise may have to pay a new 
counterparty to take on the remaining obligation.
    Derivatives counterparties present some of the same issues as 
credit enhancements. Generally, during an economic downturn, as one 
counterparty's credit deteriorates, the other party to the transaction 
may increase collateral requirements until eventually the value of 
pledged collateral more than covers risk exposure. Therefore, with 
prudent counterparty risk management, losses are most likely to occur 
due to unexpected counterparty bankruptcies. Such losses may be more 
directly related to potential financial market disturbances than to 
general economic conditions.
    Question 33: How, if at all, should OFHEO incorporate the effect of 
counterparty defaults in the risk-based capital test?
G. Non-Mortgage Investments
    The Enterprises maintain non-mortgage investment portfolios that 
include Treasury securities, federal funds, time deposits, obligations 
of states and municipalities, auction rate preferred stock, medium-term 
notes, asset-backed securities, repurchase agreements, and other 
instruments. At the end of the third quarter in 1994, these investments 
totaled $11.5 billion at Freddie Mac and $35.1 billion at Fannie Mae. 
On average in recent quarters, these investment portfolios have ranged 
from two to five percent of assets plus MBS.
    Many of these investments or their issuers are rated by the credit 
rating agencies. Even though these are very short-term and liquid 
investments, some of the issuers or the investments may be assumed to 
default during the stress period. To reflect this possibility, OFHEO's 
stress test might assume the failure of some fraction of the 
investments or issuers, based on their credit rating.
    Question 34: How should OFHEO simulate the default behavior of 
investments or issuers of short-term, liquid investments?
    Question 35: What assumptions should OFHEO make about the 
performance of rated investments or issuers over the stress period?
    Question 36: What assumptions should OFHEO make about gains and 
losses on the sale of collateral for repurchase agreements?

II. Interest Rate Risk

    Interest rate risk, associated primarily with the maintenance of a 
retained portfolio, caused the most serious losses ever experienced by 
the Enterprises. For a time during the early 1980's, Fannie Mae, which 
was then almost exclusively a portfolio institution, was insolvent on a 
mark-to-market basis.33 (Freddie Mac focused much more completely 
on mortgage pass-through securities during that time period.) As did 
much of the thrift industry at the time, Fannie Mae funded long-term, 
low-yield, fixed-rate, single family mortgages with short-term 
liabilities; rising interest rates drove up funding costs, causing 
Fannie Mae to incur significant losses.

    \33\The market value of Fannie Mae's liabilities (primarily 
market-rate, short-term securities) exceeded the market value of its 
assets (primarily below market-rate residential mortgages).
---------------------------------------------------------------------------

    Since then, Fannie Mae and Freddie Mac (the latter has built a 
substantial retained portfolio over the past decade) have developed 
funding strategies that reduce their exposure to interest rate risk. To 
protect against rising rates, liabilities have been lengthened to match 
more closely the maturity of mortgage assets. When falling interest 
rates result in accelerated mortgage prepayments, callable debt 
structures now allow the Enterprises to retire some debt early or issue 
new debt to maintain more closely their desired net interest margin. 
Adjusting hedging strategies for adjustable-rate mortgage investments 
presents a more difficult problem.
    The Enterprises have recently been building mortgage derivative 
portfolios that have an interest rate risk profile more complex than 
those of whole mortgages.
    Interest rate risk also affects income from the Enterprises' 
securitization businesses. Float income--the return on invested 
mortgage principal and interest payments prior to the corresponding 
payment to investors--varies with the level of interest rates at which 
the Enterprises reinvest such funds. Interest rates affect prepayment 
rates, and changing prepayments affect float income at each Enterprise.
    A number of issues related to the interest rate risk of the 
Enterprises are discussed below. [[Page 7477]] 
A. Yield Curve Construction
    The Act provides specific instructions concerning the ten-year CMT 
yield over the ten years of the stress test, but other points on the 
Treasury yield curve are important as well. The Treasury yield curve 
determines, directly or indirectly, the yields on adjustable-rate 
mortgages, the returns on non-mortgage investments and the costs of 
borrowing. The Act calls for Treasury yields of different maturities to 
be determined in a way that is ``reasonably related to historical 
experience and are judged reasonable by the Director.''34

    \34\Section 1361(a)(2)(D) (12 U.S.C. 4611(a)(2)(D)).
---------------------------------------------------------------------------

    Question 37: How should OFHEO determine the remainder of the 
Treasury curve and apply the curve through the ten-year stress period?
    Question 38: How should the other points on the yield curve change 
during the first year when the ten-year CMT yield is rising or falling?
    Question 39: How, if at all, should those yields vary after the 
one-year period when the ten-year CMT yield has reached its maximum or 
minimum level?
B. Mortgage Prepayments--Interest Rate Risk
    The financing of a mortgage portfolio presents one of the greatest 
challenges of asset/liability management. A portfolio manager can 
eliminate interest rate risk only if he or she issues liabilities with 
maturities, rate adjustments, and embedded options matching those of 
the mortgage assets. In a declining rate environment, should mortgages 
pay down more quickly than liabilities, new low-yield mortgages added 
to the portfolio will likely reduce the net interest margin; in a 
rising rate environment, if liabilities run off more quickly than the 
mortgage assets, the net interest margin will likely fall due to higher 
funding costs.
    Since the Enterprises absorb the credit risk of MBS, MBS dealers 
and investors principally concern themselves with interest rate risk. 
The tremendous volume of MBS outstanding, and the great sensitivity of 
MBS value to interest rate movements and resulting prepayment rates, 
have resulted in a significant research emphasis on prepayments by Wall 
Street analysts. Although most Wall Street MBS pricing models focus on 
prepayments, these models are estimated based on mortgage termination 
data that do not distinguish prepayments from defaults. For the purpose 
of modeling interest rate risk, the distinction is irrelevant.
    The section above titled ``Models of Default and Prepayment'' 
suggests an approach to the stress test that combines the simulation of 
defaults and prepayments in a joint multivariate model, making a 
termination model unnecessary. Use of a mortgage termination model for 
interest rate risk analysis runs the risk of generating implausible 
patterns of prepayments because, depending on the approach to default 
projections, defaults in some years of the stress period might approach 
or exceed total projected mortgage terminations.
    Question 40: What are the relative merits of the alternative 
approaches, e.g., a joint multivariate default/prepayment model versus 
a mortgage termination model, to modeling mortgage prepayments in the 
stress test?
C. Liabilities
    The Enterprises' liabilities may take the form of bonds and notes 
with simple structures; so-called ``structured notes,'' possibly 
combined with interest rate swap, cap or floor contracts; and foreign 
currency denominated debt coupled with foreign exchange swap contracts. 
Many bonds and contracts incorporate call or cancellation options, 
respectively. Enterprise funding costs are affected by management 
decisions to retire debt or cancel derivative contracts prior to stated 
maturities, as well as decisions about the characteristics of debt 
issued and derivatives activities initiated during the stress period.
    Even though the initial stress test involves a ``winddown'' of the 
Enterprises' businesses, decisions with respect to bond calls and 
derivatives contract cancellations must be simulated. The financing of 
mortgages purchased to fulfill contractual commitments may require the 
issuance of new liabilities and possibly the initiation of new 
derivatives contracts. The run-off of liabilities at a faster rate than 
assets may also require new issuances.
    Question 41: What should be the decision rules that OFHEO applies 
in the stress test related to the exercise of bond calls and 
derivatives contract cancellations?
    Question 42: What should be the characteristics of simulated 
liabilities issued by the Enterprises during the stress period, e.g., 
maturities, option structure, and coupon structure?
    Question 43: What are the implications for simulated liabilities of 
the pattern of interest rate movements modeled during the initial year 
of the stress period?
D. Yield Curve Volatility and Option Pricing
    The Act states that the ten-year CMT yield will be held at a 
constant level for the last nine years of the stress period,35 but 
remains silent on the volatility of the remainder of the Treasury yield 
curve. Theoretically, the historical volatility of the yield curve has 
some bearing on expectations of future volatility. Expectations of 
future volatility, in turn, are a determinant of the current value of a 
call option on debt.

    \35\Section 1361(a)(2) (B) and (C) (12 U.S.C. 4611(a)(2) (B) and 
(C)).
---------------------------------------------------------------------------

    Question 44: How does OFHEO implement the link between the 
volatility of the yield curve experienced during the stress test and 
the market's expectations of future volatility?
    Question 45: What assumptions should OFHEO make about the speed 
with which the Enterprises adjust to changes in volatility during the 
stress period?
    Question 46: If the actual volatility of yields experienced during 
the stress test reaches extraordinarily low levels, what assumptions 
should OFHEO make to ensure reasonable pricing and use of call options 
on new debt?
E. Enterprises' Costs of Borrowing
    As any organization depletes its capital reserves, the 
organization's cost of borrowing increases due to its higher perceived 
risk. Spreads over Treasury securities might also be affected by other 
aspects of the stress period, including the sharp interest rate changes 
early in the period and the prolonged general economic weakness.
    Question 47: What techniques should OFHEO use to project the 
Enterprises' borrowing costs? How should the stress test link capital 
levels and quality spreads (borrowing rates relative to Treasuries)?
    Question 48: Should yields relative to Treasuries widen during the 
stress period in response to general interest rate changes or credit 
problems? If so, by how much should they widen?
F. Hedging Activities
    Hedging activities associated with structured notes, which convert 
specific securities into a preferred debt structure, are addressed 
above under ``Liabilities.'' The Enterprises engage in other hedging 
activities to manage interest rate risk more generally. The Act 
provides that:

    Losses or gains on other activities, including interest rate and 
foreign exchange hedging activities, shall be determined by the 
Director, on the basis of available [[Page 7478]] information, to be 
consistent with the stress period.36

    \36\Section 1361(a)(4) (12 U.S.C. 4611(a)(4)).

    Question 49: How should OFHEO simulate gains and losses (other than 
those associated with counterparty failures) on derivative activities 
in the stress test?
G. Investment of Excess Cash
    Under certain circumstances, simulation of the stress scenarios may 
require decision rules concerning the investment of excess cash. For 
example, in the stress test scenario where the ten year CMT yield 
falls, mortgage prepayments will increase. The proceeds from 
prepayments of mortgages in the retained portfolio may exceed the cost 
of retiring associated debt. Likewise, in the rising rate stress test 
scenario, mortgages will prepay more slowly than in other scenarios. 
Slower prepayments may lead to the receipt of more guarantee fee income 
than initially anticipated on the Enterprises sold portfolio because 
the mortgages remain outstanding longer than originally anticipated.
    Since the Act does not permit the simulation of new business in the 
initial stress test model, any excess cash generated during the stress 
test period must be assumed to either be retained as cash or reinvested 
in an interest-bearing asset.
    Question 50: What decision rules should govern the investment of 
excess cash during the stress period?
    Question 51: What rate of interest should excess cash be assumed to 
earn?
    Question 52: Should excess cash be assumed to earn a single rate or 
a weighted average rate, representing a range of possible investment 
choices?
H. Other Indexes and Yields
    Values must be created for other indexes and yields, e.g., the 
Federal Home Loan Bank Eleventh District Cost of Funds Index and the 
London Interbank Offer Rate, over the stress period in order to 
reasonably project liability costs, as well as amortization, 
prepayment, and default rates on affected adjustable rate mortgages. 
One reasonable approach might be for OFHEO to create equations that 
project these indexes based on their relationship to points on the 
Treasury yield curve and assumed market conditions consistent with the 
circumstances of the stress test.
    Question 53: What techniques should be used to simulate the 
behavior of these indexes and yields?

III. New Business and Other Considerations

    OFHEO's risk-based capital test must incorporate a number of 
decision rules to reflect management actions that would significantly 
affect the financial performance of the Enterprises during the stress 
period. Initially, the Act requires that OFHEO's stress test 
incorporate no new business for the Enterprises during the stress 
period other than the fulfillment of contractual commitments to 
purchase mortgages or issue securities.37 The Act specifically 
states that:

    \37\Section 1361(a)(3) (12 U.S.C. 4611(a)(3)).

    The characteristics of resulting mortgage purchases [and] 
securities issued * * * will be consistent with the contractual 
terms of such commitments, recent experience, and the economic 
characteristics of the stress period.38

    \38\Id.

The Act also requires that characteristics of the stress period other 
than those discussed above in the ``Credit Risk'' and ``Interest Rate 
Risk'' sections (such as, for example, dividend policies and operating 
expenses) be determined by the Director, on the basis of available 
information, to be most consistent with the stress period.39

    \39\Section 1361(b)(2) (12 U.S.C. 4611(b)(2)).
---------------------------------------------------------------------------

A. Commitments
    At this time, the only ``new business'' OFHEO can assume during the 
stress period is the fulfillment of contractual commitments to purchase 
mortgages or issue new securities. As a regular business practice, the 
Enterprises enter into commitments to purchase mortgages for periods 
that may extend from a few weeks up to a year. The commitments specify 
underwriting and pricing criteria for the mortgages to be delivered. If 
the Enterprise intends to securitize the mortgages listed in the 
commitment, then the Enterprise will hedge the commitment at the time 
it is executed by selling the mortgages forward.
    Often the seller/servicer that has agreed to sell to an Enterprise 
under a commitment has not yet originated the mortgages at the time the 
commitment is executed. When the seller/servicer actually delivers 
mortgages, their characteristics may differ from those specified in the 
original commitment.
    Question 54: How should OFHEO define the term ``commitments''?
    Question 55: On what basis, if any, should OFHEO simulate the 
fulfillment of outstanding commitments?
    Question 56: What mix of product types and underwriting qualities 
should be assumed?
    Question 57: What delivery timing should be assumed?
    Question 58: What assumptions should be made with regard to 
securitization versus retention in portfolio?
B. Dividend Policies
    During the stress period, net income will fall, reducing cash 
available for distribution to shareholders. In such circumstances, 
Enterprise management might be expected to suspend dividends or reduce 
the dividend rate. However, Enterprise management may be reluctant to 
take such actions, because dividend reductions send a negative signal 
to investors and would be expected to depress the market price of 
Enterprise shares.
    Question 59: Should OFHEO assume continuation of the present 
dividend policies of each Enterprise for the entire stress period?
    Question 60: If OFHEO simulates a reduction in the dividend payout 
rate, at what point in the scenario should it take place?
    Question 61: By how much should dividends be reduced if they are 
reduced?
C. Operating Expenses
    The Act is silent on how operating expenses should be treated in 
the stress test, but OFHEO interprets the Act to require that OFHEO 
model operating expenses in a manner most consistent with the stress 
period. Operating expenses lower the Enterprises' earnings or increase 
their losses, and thereby reduce their capital. The major portion of 
operating expenses at each of the Enterprises consists of costs related 
to personnel, occupancy, and equipment. Each Enterprise is divided by 
business function, such as purchase of mortgages, credit analysis, and 
investment management. Each Enterprise has regional offices. The 
cessation of additional business at the commencement of the stress 
period (beyond the fulfillment of contractual obligations) creates 
conditions that would quickly eliminate some operations and gradually 
reduce others.
    Question 63: How should OFHEO appropriately model operating 
expenses in the stress test?
    Question 64: To what extent, if any, should operating expenses be 
disaggregated and treated in distinct categories?
    Question 65: How, if at all, should the stress test distinguish 
between the Enterprises in their management of operating expenses 
during the stress period? [[Page 7479]] 

Conclusion

    OFHEO has identified and highlighted many of the significant issues 
that must be addressed in connection with development of the stress 
test and the associated risk-based capital regulation. OFHEO seeks 
comment on these and any additional issues that may be identified.
    The development of the stress test and the risk-based capital 
regulation is one of the critical statutory responsibilities of OFHEO. 
In carrying out this responsibility, OFHEO is committed to a regulatory 
process that will provide the broadest possible range of opinions from 
the widest array of information sources for consideration during the 
regulatory process. The development of the stress test and the 
implementation of the risk-based capital regulation will provide 
regulatory and analytical standards and tools that will safeguard the 
financial safety and soundness of the Enterprises and in turn will 
ensure that the Enterprises continue to accomplish their public 
missions. Given the significance of this undertaking, OFHEO encourages 
all interested parties to analyze the issues raised in this ANPR and 
submit comments on the specific questions. OFHEO will thoroughly 
analyze and carefully consider all comments during the course of the 
development of the stress test and risk-based capital regulation.

    Dated: February 2, 1995.
Aida Alvarez,
Director, Office of Federal Housing, Enterprise, Oversight.
[FR Doc. 95-3076 Filed 2-7-95; 8:45 am]
BILLING CODE 4220-01-P