U.S. patent application number 11/700832 was filed with the patent office on 2007-06-14 for automated loan evaluation system.
This patent application is currently assigned to Brandywine Building. Invention is credited to Brian L. Libman.
Application Number | 20070136187 11/700832 |
Document ID | / |
Family ID | 37768333 |
Filed Date | 2007-06-14 |
United States Patent
Application |
20070136187 |
Kind Code |
A1 |
Libman; Brian L. |
June 14, 2007 |
Automated loan evaluation system
Abstract
Disclosed is a system and method of creating a probability of
delinquency database using historical loan data and a plurality of
loan factors, for use in determining a loan rate, the method
comprising: identifying a plurality of multi-level loan factors;
creating a pool from the historic loan data, wherein the first pool
contains records relating to the plurality of multi-level loan
factors; separating the pool into a set of groups based on the
multi-level loan factors, calculating a probability of delinquency
for each group; and arranging the probability of delinquency for
each group into a database such that the probability of delinquency
for each group is accessible for any combination of multi-level
loan factors. Also disclosed is a system and method of determining
the loan rate for a loan for a borrower based on the probability of
delinquency database.
Inventors: |
Libman; Brian L.; (New
Canaan, CT) |
Correspondence
Address: |
MORGAN LEWIS & BOCKIUS LLP
1111 PENNSYLVANIA AVENUE NW
WASHINGTON
DC
20004
US
|
Assignee: |
Brandywine Building
Wilmington
DE
|
Family ID: |
37768333 |
Appl. No.: |
11/700832 |
Filed: |
January 30, 2007 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
09790374 |
Feb 21, 2001 |
|
|
|
11700832 |
Jan 30, 2007 |
|
|
|
60184150 |
Feb 22, 2000 |
|
|
|
Current U.S.
Class: |
705/38 |
Current CPC
Class: |
G06Q 40/02 20130101;
G06Q 40/025 20130101 |
Class at
Publication: |
705/038 |
International
Class: |
G06Q 40/00 20060101
G06Q040/00 |
Claims
1. A method of creating a probability of delinquency database using
historical loan data and a plurality of loan factors, for use in
determining a loan rate, the method comprising: (a) identifying a
plurality of multi-level loan factors; (b) creating a first pool
from the historic loan data, wherein the first pool contains
records relating to the plurality of multi-level loan factors; (c)
identifying a significant multi-level loan factor from the
plurality of multi-level loan factors; (d) separating the first
pool into a plurality of subpools based on levels of the
significant multi-level loan factor; (e) separating each subpool of
the plurality of subpools into a plurality of groups based on a
second plurality of multi-level loan factors, wherein the second
plurality of multi-level loan factors does not include the
significant multi-level loan factor; (f) calculating a probability
of delinquency for each group; and (g) arranging the probability of
delinquency for each group into a database such that the
probability of delinquency for each group is accessible for any
combination of multi-level loan factors.
2. The method of claim 1, wherein the multilevel loan factors are
selected from the group consisting of: Loan-to-Value ("LTV"),
mortgage pay history, and credit rating.
3. The method of claim 1, wherein the significant multilevel loan
factor is the mortgage pay history.
4. The method of claim 1, wherein the probability of delinquency
for each group is transformed into a base credit score.
5. The method of claim 4, wherein the database is represented as a
printed table.
6. The method of claim 1, further comprising: (a) identifying a
plurality of either-or loan factors; (b) creating a second pool
from the historic loan data, wherein the second pool contains
records relating to the plurality of either-or loan factors; and
(c) calculating an either-or probability of delinquencies for each
either-or loan factor in the plurality of either-or loan factors,
wherein each either-or probability of delinquency is combinable
with any probability of delinquency for each group.
7. The method of claim 6, wherein the either-or loan factors are
selected from the group consisting of: alternative documentation,
stated documentation, number of units, owner occupation, no
mortgage history, combined loan-to-value ratio, cash out,
debt-to-income ratio, and bankruptcy.
8. The method of claim 6, wherein the either-or probability of
delinquency for each group is transformed into an add-on credit
score.
9. The method of claim 8, wherein the plurality of either-or
probability of delinquency is represented as a printed table.
10-26. (canceled)
Description
CROSS-REFERENCE
[0001] This invention claims the benefit of U.S. Provisional Patent
Application 60/184,150 filed on Feb. 22, 2000.
BACKGROUND
[0002] This application relates generally to business methods for
evaluating loans, and more particularly, to a system and method for
providing a mortgage loan pricing model for various lending
scenarios.
[0003] In general, loans are often classified as either a prime
loan or a sub-prime loan. Sub-prime mortgage loans are loans which
do not meet the criteria of the Federal National Mortgage
Association and the Federal Home Mortgage Corporation
(collectively, the "Agencies") for purchase by the Agencies.
Typically, sub-prime loans have one or more credit issues related
to the borrower which the Agencies have determined would increase
the probability of the loan payments on such loans not being made
to the lender when due.
[0004] There is a market for sub-prime loans, however, through
whole loan purchasers and ultimately, through investors in
securities other than those issued by the Agencies. However,
pricing of such loans in the past has been driven primarily by
guesswork and competition.
[0005] It is desired to remove some of this guesswork and provide a
reliable, fair, and consistent evaluation for all loans, including
sub-prime loans.
SUMMARY
[0006] In response to the above described problems and
deficiencies, a methodology is provided for translating information
from historical data relating to similar loans. The historic
information can be used to evaluate loans, such as value-adjusted
and/or risk-adjusted mortgages for residential sub-prime lending.
The methodology may also convert the pricing model into a simple,
user-friendly system for grading and pricing such loans.
[0007] In operation, the first step is to determine a credit grade
for the subject borrower. This is determined by using only the
mortgage history of the borrower. A credit report is reviewed to
determine the borrower's 12 month mortgage history on the subject
property or similar type property (e.g., primary residence if the
new loan is for purchase of another primary residence). The
reviewer determines how many payments were over 30, 60, 90 and 120
days late, and this review results in the labeling of the loan as
A, A-, B, C or D credit grade.
[0008] In one embodiment, each credit grade has a separate Credit
Score/Loan-to-Value ("LTV") matrix. This matrix considers two
factors: the Credit Score, which is a Fair, Isaac credit ("FICO")
score obtained from a three-company consolidated credit report, and
the proposed loan-to-value for the subject loan. Based upon this
matrix, a "base score" is obtained. For the sake of reference,
these base scores are in 0.25 increments between 1.00 and 5.00.
[0009] Once the base score is obtained, various "Risk Add Ons" are
added to the base score, as applicable. The purpose of these is to
account for "add-on" risk factors considered by potential
purchasers other than "multi-level" loan risk factors, such as LTV,
Credit Score and mortgage payment history. Add-on risk factors may
include, for example, such characteristics as Alternative
Documentation, Stated Documentation (for self-employed borrowers),
3-4 unit property (rather than 1-2 unit), Non Owner Occupied, a
combined LTV of more than 95%, cash out to the borrower, lack of
mortgage history, excessive debt to income ratio, and prior
bankruptcy filings. Other risk factors may be added, based upon
conditions in the secondary market. The final Credit Score
determines whether the mortgage loan should be made and, if so, at
what rate.
[0010] An advantage of the present invention is that it avoids the
impracticality of individually pricing each loan based upon unique
characteristics. Instead, the system and method create groups of
loans which, although not identical to each other, might reasonably
be expected to perform in a similar manner and therefore, in a
risk-adjusted environment, and should be priced the same.
DRAWINGS
[0011] FIG. 1 is a flow diagram illustrating one embodiment for
developing probability of delinquency databases and procedures.
[0012] FIG. 2 is a flow diagram illustrating one embodiment of
using the database developed according to the diagram illustrated
in FIG. 1.
[0013] FIG. 3 is a flow diagram illustrating in detail one
embodiment for developing probability of delinquency databases and
procedures.
[0014] FIGS. 4a-4e are tables and associated graphs of an example
set of loan data.
[0015] FIGS. 5a-5e are example tables illustrating probability of
delinquency databases or tables.
[0016] FIG. 6 is an example probability of delinquency add-on
table.
[0017] FIG. 7 is an example set of rules for the probability of
delinquency add-on table of FIG. 6.
[0018] FIGS. 8a-8e are example tables illustrating credit score
databases or tables.
[0019] FIG. 9 is an example rate sheet.
DETAILED DESCRIPTION
[0020] The present invention provides a unique system and method
for evaluating loans. In one embodiment, the system and method
provide risk-based pricing, taking into account those factors which
make the loans "non-conforming" from the standpoint of the
regulating agencies. At the same time, it simplifies the
underwriting of these sub-prime loans. It is understood that the
following disclosure provides many different embodiments, or
examples, for implementing different features of this invention.
Techniques and requirements that are only specific to certain
embodiments should not be imported into other embodiments. Also,
specific examples of numbers, ranges, and formats are described
below to simplify the present disclosure. These are, of course,
merely examples and are not intended to limit the invention from
that described in the claims.
Glossary
[0021] The present disclosure uses terms that are well know in the
art of loan financing. For the sake of convenience, several of the
terms are defined below. [0022] Alterative documentation
("Alterative doc"): Verification of the borrower's income based on
a limited set of documentation. A lesser requirement for
documentation than "Full Documentation," usually consisting of
personal bank statements. [0023] Cash out: A loan where the
borrower receives more than $1,000 of the loan proceeds. [0024]
Combined loan-to-value ratio ("CLTV"): The ratio calculated by
dividing the sum of the principal balance of all loans constituting
a lien against the subject property by the appraised value of the
property or, if the loan is for purchase money, the lesser of
appraised value or the purchase price. [0025] FICO score: A credit
reporting score assigned to a borrower created by calculations
based upon the borrower's credit report; usually based upon a
merger of two or three different credit reports. A higher score
generally indicates a better credit history. [0026] Debt-to-income
ratio ("DTI"): The ratio calculated by dividing the borrower's
total installment debt to third parties (including mortgages,
automobile loans, and credit card loans) by the borrower's total
gross income from all sources. [0027] Full Documentation ("Full
Doc"): All standard documentation and verification of debts and
income provided [0028] Owner-occupied property: The borrower
maintains the subject property as his or her own residence: [0029]
Stated documentation ("Stated Doc"): Applicable to self-employed
borrowers only. Gross income of the borrower is assumed to be as
stated by the borrower; only required evidence of income is the
borrower's statement on an official Department of Housing and Urban
Development approved form used in loan originations (e.g., a 1003
application). System and Method Operation:
[0030] Referring now to FIG. 1a, a method 20 provides a broad
overview of one embodiment for determining a probability of
delinquency for a loan where several loan factors are known. The
method 20 begins at step 22 where historic loan data is analyzed in
light of various loan factors. The loan factors identified may
include loan-to-value ratio, FICO score (obtained from a tri-party
credit report), mortgage pay history, whether or not the loan was
qualified using alternative documentation, whether or not the loan
was qualified using stated doc (self-employed borrowers only),
number of units (measured as 1-4), whether the property is owner
occupied, the combined loan-to-value ratio (i.e., the ratio of all
liens against the property to the appraised value of the property),
whether the loan includes cash out to the borrower, debt-to-income
ratio, and whether the borrower is currently in bankruptcy. It is
important to note that other factors tracked by the origination
system on a loan-level basis could also be used.
[0031] As will be explained in more detail below, at step 24 each
loan factor is analyzed against historic loan delinquencies, so
that a historic probability of delinquency may be determined for
each loan factor. In step 26, a simplified procedure is developed
so that applicable probability of delinquency for each loan factor
may be combined into an overall probability of delinquency. Thus,
many factors may be used to determine an overall probability of
delinquency without the need for complex algebraic equations.
[0032] Once the overall probability of delinquency is determined,
the price of the loan can be determined. Referring to FIG. 1b, a
method 10 provides a broad overview of the pricing aspects of one
embodiment. The method 10 is for pricing a potential loan for a
client. The method 10 begins at step 12 where a credit grade is
determined. The credit grade may be based on various types of loan
factors. In one embodiment, the credit grade is based on historical
information regarding loan payments (for other mortgage loans) over
the last year. In one embodiment, a borrower's credit grade may be
classified as an "A", "A-", "B", "C", and "D". As will be explained
in greater detail below, this classification depends how many loan
payments were over 30, 60, 90, and 120 days late. Once a credit
grade is determined, a specific matrix or database can be chosen at
step 14 which is specific for each credit grade. In one embodiment,
each matrix includes a vertical axis of FICO scores from 500 to
680, and a horizontal axis of LTV ratios from 60% to 95%.
[0033] At step 15, a base probability of delinquency is determined
from the chosen matrix. For instance, given a FICO score and a LTV
for the loan in question, a process may easily determine the base
probability of delinquency. Given that the FICO scores are
represented by the vertical axis and the LTV ratio are represented
by the horizontal axis, the probability of delinquency will be
found in the cell which intersects the appropriate row and column.
Thus, at step 16, the process determines the probability of
delinquency from three different factors (payment history, FICO
score, and LTV) by the use of matrixes without having to use
complex equations.
[0034] At step 16, the probability of delinquencies representing
the appropriate add-on values may then be independently calculated.
At step 17, the base probability of delinquency may be combined
with the probability of delinquencies due to the add-on factors to
determine a total probability of delinquency.
[0035] At step 18, a predetermined rate sheet or database may be
accessed and the probability of delinquency may be cross-referenced
to a loan rate. Rate sheets are commonly used in the industry and
are frequently adjusted to interest rates and other costs unrelated
to the probability of delinquency. A price of the loan or loan rate
will be determined from the rate sheet, which take into effect
other pricing factors common in the industry, such as whether or
not a pre-payment penalty may be assessed.
[0036] Referring to FIG. 2, the method 10 may be performed manually
or on a computer 50. The computer 50 may be one or more mainframes,
servers, wireless telephones, personal digital assistants, and the
like. The computer 50 includes a processing unit 52, an
input/output ("I/O") 54, and a storage 56. The I/O 54 may include a
monitor and keyboard, an interface screen of a personal digital
assistant, a network interface, or other communication module. The
storage 56 may include a local memory including one or more local
storage devices, as well as a remote memory with one or more remote
storage devices. It is understood that each of the listed
components may actually represent several different components, and
some components may not be included in certain embodiments.
Risk of Delinquency and Matrix Creation:
[0037] The above described procedures and matrices are created by
analyzing historic loan data. In the present embodiment, it is
desired to determine the Probability of Delinquency of a loan with
a specific set of characteristics. It is also desired to associate
each Probability of Delinquency (or range thereof) to an interest
rate reflecting the price of the loan. Furthermore, it is desired
to create a format for presenting that information to employees
involved in underwriting, lending and processing loans, third-party
brokers and to borrowers in a simple and easily understood
form.
[0038] Referring now to FIG. 3, a method 100 can be used to
determine a probability of delinquency from a set of historical
loan data. Execution begins at step 102 where potential loan
factors that might affect risk of loan delinquency are
identified.
[0039] As previously discussed, potential loan factors identified
may include loan-to-value ratio, FICO score (obtained from
tri-party credit report), mortgage pay history, whether or not the
loan was qualified using alternative documentation, whether or not
the loan was qualified using stated doc (self-employed borrowers
only), number of units (measured as 1-4), whether the property is
owner occupied, the combined loan-to-value ratio (i.e., the ratio
of all liens against the property to the appraised value of the
property), whether the loan includes cash out to the borrower,
debt-to-income ratio, and whether the borrower is currently in
bankruptcy.
[0040] At step 104, each factor is categorized as "multi-level" or
"either-or." For simplicity, the number of multi-level factors may
be limited to a particular number (e.g., four.) A multi-level
factor is a factor with many gradations, e.g., LTV which may be
anywhere from 1% to 100% (or potentially higher). An either-or
factor has only two choices, e.g., owner-occupied or not. A
multi-level factor may be converted to an either-or factor by
locating a "break point" and changing the gradations to "above" and
"below"; e.g., LTV's may be described as "90% and over" and "less
than 90%."
[0041] Either-or factors may also be referred to as add-ons
factors. For instance, the unit will either be owner occupied or
not. If the unit is not owner occupied, the probability of
delinquency may be determined for this loan factor, and simply
"added on" to the overall probability or score.
[0042] In some embodiments, the mortgage pay history is the single
most significant indicator of performance. For purposes of this
application, the term "performance" describes the situation when a
borrower repays a loan according to its schedule. It can be thought
of as the opposite of delinquency. LTV and FICO scores are also
very significant, and their effect upon performance is gradual with
performance improving gradually with decreases in LTV or increases
in FICO scores, with other loan factors held constant.
[0043] At step 106, historic information for sub-prime loans is
obtained and collected into a set. A very large historic
information set from a variety of originators and servicers is
desirable, since a large number will dilute the effect of anomalies
in origination or servicing processes. The information set should
include data on each loan in most or all of the categories
identified in step 104 as loan factors which might affect the
probability of delinquency. Historic information may be purchased
directly from loan servicers.
[0044] Once the historic information set is obtained, the
multi-level loan factors can be analyzed. However, in step 108, the
data may have to be filtered to remove the effects of add-on
factors and other circumstances. For instance, the incidence of
delinquency is determined on a particular payment date. In one
embodiment, this date is the 12.sup.th payment date. Consequently,
loans which were liquidated or paid off prior to the particular
payment date should be excluded from the information set.
[0045] In addition, for purposes of determining the initial base
score, only loans with no "add-on" factors should be included in
the initial analysis. In other words, only full doc, 1-2 unit,
owner occupied properties when there is no cash out to the
borrower, the combined loan-to-value ratio is less than 95%, where
the borrower has a debt-to-income ratio of not more than 45%, and
where there is no existing bankruptcy. In one embodiment, loans
where no previous mortgage loan history are be excluded for this
analysis. Furthermore, one embodiment also excludes loans with LTVs
of 95% from this analysis. In step 108, for analysis purposes,
loans with these characteristics will be separated or filtered out
into a separate group or pool. For convenience, this filtered pool
will be referred to as "pool A."
[0046] After the information set has been filtered into pool "A,"
each multi-level loan factor may then be isolated. It is then
possible, in step 110, to determine the actual incidence of
delinquency at each level for a given loan factor. For multi-level
loan factors with a large number of levels (e.g., LTV may have 100
or more), appropriate groupings may be made. Prior grouping may
significantly reduce the complexity of matrixes in those
embodiments using lookup tables or databases. However, the use of
groupings should not adversely affect the accuracy of the tables.
Thus, it may be necessary to ensure that the groupings selected
will contain a statistically significant number of loans.
[0047] To illustrate the method 100 of FIG. 3, an example
information set will be discussed. Referring to FIG. 4a, an
information set of 15,000 loans has been filtered (i.e., no
add-ons, etc.) and divided into groups by initial LTV increments of
5 for the range between 65 and 94 (see column (a) of FIG. 4a).
Column (b) indicates the number of loans in each group for this
example information set. Column (c) represents the number of loans
in each LTV grouping that are 90+ days delinquent (on the 12.sup.th
payment date). Column (d) represents the probability of
delinquency, which can be calculated from dividing the number of
loans (column (b)) into the incidences of delinquency (column (c)).
FIG. 4b graphically illustrates the relationship between the LTV
groups and the probability of delinquency.
[0048] Referring to FIG. 4c, which continues analyzing the pool
"A," pool "A" can be separated into additional groups or categories
using the loan payment history. In one group, the borrower has 1 or
no mortgage payments no more than 30 days late in past 12 months
("1.times.30"). In another group, the borrower has 2 mortgage
payments no more than 30 days late in past 12 months
("2.times.30"). In another group, the borrower has 3 or 4 mortgage
payments no more than 30 days late in past 12 months
("4.times.30"). In another group, the borrower has 1 mortgage
payment 60 days late and 1 or 2 payments no more than 30 days late
in the past 12 months ("2.times.30 & 1.times.60"). Additional
similar groups can be created. Thus, column (a) of FIG. 4c lists
the pay history groups. Column (b) indicates the number of loans in
each group of the example information set. Column (c) represents
the number of loans in each LTV grouping that are 90+ days
delinquent on the 12.sup.th payment date of each loan. Column (d)
represents the probability of delinquency, which can be calculated
from dividing the number of loans into the incidences of
delinquency. FIG. 4d graphically illustrates the relationship of
each mortgage pay history group against the probability of
delinquency.
[0049] Referring to FIG. 4e, the probability of delinquency for the
pool "A" can also be compared to a credit history score, such as
the FICO scores. Thus, column (a) of FIG. 4e lists the FICO scores
in increments of 10 for the range between 520 and 680. Column (b)
indicates the number of loans in each group of this particular
information set. Column (c) represents the number of loans in each
LTV grouping that are 90+ days delinquent at the 12.sup.th payment
date. Column (d) represents the probability of delinquency, which
can be calculated by dividing the number of loans into the
incidences of delinquency.
[0050] Referring back to FIG. 3, it is desirable to calculate the
probability of delinquency for a given loan based on all of the
multi-level loan factors (e.g., LTV, mortgage pay history, credit
score). As will be explained in detail below, when matrixes or
tables are used to store historic data, the probability of
delinquency for each combination of factors may be determined by:
selecting the most significant loan factor (step 112); dividing the
most significant loan factor into groups (step 114); sorting the
groups (step 116); and creating a matrix for each group where the
matrix reflects the contributions of the other two loan factors
(step 118).
[0051] Continuing with the example of pool "A", FIGS. 4a, 4c, and
4e show the correlation between the probability of delinquency and
the respective loan factors (i.e., mortgage pay history, FICO
score, and LTV). The probability of delinquency of the top two rows
in FIG. 4c (the mortgage pay history) are significantly greater
than the probability of delinquency reflected in any row of FIGS.
4a and 4b (the LTV and FICO scores, respectively). Assuming that
actual information sets are similar to the example above, it can be
assumed that the mortgage loan history loan factor is the single
most significant indicator of delinquency. In this context, "most
significant" means the factor which appears to correlate most
directly with probability of delinquency.
[0052] Referring back to FIG. 4d, significant breaks and
congruencies in the probability of delinquency for mortgage loan
history are located and used to establish credit grades (initially
designated A, A-, B, C and D) for each level of mortgage pay
history. FIG. 4d shows that the 1.times.30 level clearly
outperforms other mortgage pay history levels. For simplicity, this
level may be designated as an "A" credit grade. The 2.times.30
level is 2.5 times more likely to default than 1.times.30 level,
but half as likely to default as 4.times.30 level. Thus, borrowers
with 2.times.30 mortgage loan history may be designated as an "A-"
credit grade. The 4.times.30 and 2.times.30 & 1.times.60 levels
have almost identical probabilities of delinquency, and thus can be
designated together as a "B" credit grade. The 6.times.30,
1.times.60 & 2.times.90, and 3.times.60 levels are 50% more
likely to default than either 4.times.30 or 2.times.30 &
1.times.60, and the probabilities of default in these three levels
are very close. These levels may also be designated together as a
"C" credit grade. The 1.times.120 level of loans clearly has a
higher probability of default than any other level. These loans are
designated as "D" credit grade. Borrowers whose mortgage pay
history includes a payment more than 120 days delinquent during the
past 12 months are considered extremely high risk and are excluded
from consideration altogether.
[0053] Once credit grades are established (step 114 of FIG. 3) and
sorted (step 116 of FIG. 3), matrixes can be developed for each
credit grade (step 118 of FIG. 3) indicating the relationship of
the other multilevel loan factors (e.g., FICO score and LTV) at the
respective credit grade. In other words, for each credit grade, the
relationship between the FICO score, the LTV, and the probability
of delinquency may be determined.
[0054] To illustrate, refer to Table 1, below, which uses the pool
"A" discussed above. Table 1 shows the specific number of loans in
each credit grade established at step 114. TABLE-US-00001 TABLE 1
Grade Mortgage Pay History Level Number of Loans A (1 .times. 30)
750 A- (2 .times. 30) 1000 B (4 .times. 30 and 2 .times. 30 + 1
.times. 60) 4000 C (6 .times. 30 and 1 .times. 50 + 1 .times. 90
and 3 .times. 60) 6750 D (1 .times. 120 max) 1500
[0055] To create a series of matrixes, the loans in each Credit
Grade can be separated into subpools, first by FICO score, then by
LTV. At step 118 of FIG. 3, a series of analyses can then be
performed, calculating the probability of delinquency in each
combination of FICO score and LTV. The calculation of probability
of delinquency would be made similar to the process discussed in
reference to FIGS. 4a-4e, above. Then, for each Credit Grade, a
matrix can be created showing the probability of delinquency for
each FICO score/LTV combination in each Credit Grade. FIGS. 5a-5e
illustrate example matrixes for each credit grade created from the
pool "A". For example, the matrix illustrated in FIG. 5a, contains
seventeen levels of credit reporting scores (along the y-axis) and
six levels of LTVs creating 102 cells (17.times.6=102). Therefore,
the value of each cell in the matrix contains the probability of
delinquency for the subpool of loans having a particular credit
grade, credit reporting score, and LTV.
Creating Add-Ons:
[0056] Turning back to the example information set, those loans
filtered out into other pools (i.e., those loans with add-on
factors) at step 108, may now be analyzed. These pools of loans may
be sorted for those loans which have one, but no more than one,
add-on factor. In step 120, these loans can then be separated into
Pool B (alternative docs), Pool C (stated doc), Pool D (3-4 unit
properties), Pool E (non-owner occupied properties), Pool F (those
loans where the CLTV at origination was >95%), Pool G (those
loans where there was cash out), Pool H (DTI over 45%), and Pool I
(existing bankruptcies). Alternatively or in addition, other
factors which have an apparent effect on incidence of delinquency
and for which there is adequate data may be added.
[0057] For each loan pool, in step 122, a risk of delinquency
associated with the respective loan factor may be determined. For
instance, from the example information set, if pool E consisted of
200 loans and had an incidence of delinquency of 10, the
probability of delinquency associated with pool E (and the
respective loan factor of non-owner occupation) would be 10/200 or
0.05. Similar risks of delinquencies may be created for each
"add-on" factor. In embodiments that use charts and tables, an
"add-on" table may be created. Rules for using the table may also
be published. An example add-on table is illustrated in FIG. 6. An
example set of rules for using the add-on table of FIG. 6 is
illustrated in FIG. 7. In embodiments using computer devices,
add-ons may simply be "flagged" during the initial data entry.
During processing, all of the flags are tested, if a flag for a
particular loan factor is set, the corresponding probability of
delinquency is added to the total probability of delinquency
variable.
Operation:
[0058] Once the matrixes have been created and the add-on factors
calculated, the overall risk of delinquency may be determined for
any given borrower, and a corresponding loan rate may then be
determined. Referring back to FIG. 1b, the overall risk of
delinquency associated with any given loan may be found by:
determining the credit grade (step 12), selecting the matrix or
database associated with the credit grade (step 14), determining a
base risk of delinquency from the applicable matrix (step 15),
determining the risks of delinquency for the applicable add-on
factors (step 16), and combining the risks to arrive at a total
risk of delinquency(step 17).
[0059] For instance, assume a borrower has had a mortgage history
of 2 late payments of less than thirty days, the applicable credit
score is 596, the LTV is 78, and the owner will not occupy the
property. From Table 1, it is determined that the borrower's credit
grade is A-, thus the matrix illustrated in FIG. 5b is selected.
Using a FICO score of 596 and an LTV of 78, the base risk of
delinquency can be determined from FIG. 5b to be 0.0624. Because
the owner does will not occupy the property, the add-on risk of
0.05 is also added to the base risk. The overall risk of
delinquency for the example loan, therefore, is 0.1124.
[0060] Once the total risk of delinquency is calculated, the loan
rate may now be obtained using the following formula: P = [ F
.function. ( A ) .times. C + L - C .function. ( L ) + G ] AC
##EQU1## Where: [0061] P=Loan Rate [0062] F=Funding Rate (i.e.,
cost of funds) [0063] A=Average Life of Loan (i.e., number of years
to average payoff) [0064] C=Percentage of loans that are not
delinquent (i.e., 90 days late at 12 months) [0065] L=Expected Loss
in points (this value will vary with the LTV, it also varies from
state to state) [0066] G=Profit Goal (hoped-for premium in the
market)
[0067] For illustrative purposes, certain assumptions will be made.
It is understood, however, that different assumptions can be made
for different scenarios, and the assumptions are not meant to limit
the invention. For purposes of example, expected Loss will be held
constant at 40 percent. Average Life will be assumed to be 2.0
years. The Funding Rate will be assumed to be 7 percent. The Profit
Goal is set for each individual product (e.g., fixed, ARM, 2/28 or
5/25). The state in which the property is located will also affect
the Loan Rate because the state will affect Average Life, due to
existence of prepayment penalties, and Expected Loss, due to time
necessary for foreclosure. The calculation of Funding Rate is
currently an interest rate on warehouse lines of credit, but may
eventually include the cost of funding through securitization,
which is based upon the weighted average coupon of the issued
securities and the required over collateralization level.
[0068] The variable "C" is the percentage of performing loans,
expressed as a percentage. This value is: C=(1-Total Probability of
Delinquency).times.100
[0069] To illustrate, assume the following: [0070] P=Loan Rate, the
variable to be solved [0071] F=Funding Rate (7%) [0072] A=Average
Life (2 years) [0073] C=Performing Loans (95%) [0074] L=Expected
Loss (40%) [0075] G=Profit Goal (5%)
[0076] Inserting the above variables into the following loan rate
equation yields: P = .times. [ 7 .times. ( 2 ) .times. 95 .times. %
+ 40 - 95 .times. % .times. .times. ( 40 ) + 5 2 .times. 95 .times.
% = .times. 13.3 + 40 - 38 + 5 180 .times. % = .times. 20.3 1.8 =
.times. 11.28 .times. % ##EQU2##
[0077] The loan rate (P), therefore, may be readily determined.
ALTERNATIVE EMBODIMENTS
[0078] As previously discussed, the above pricing process may be
implemented on a personal computer, an Internet webpage, or a
personal digital assistant (such as a Palm handheld device). It is
also possible to use charts or tables to implement the pricing
process. If charts are used, it may be desirable to use credit
scores rather than using risks or probabilities. Credit scores may
be easier for loan officers to manipulate. Furthermore, it may not
be desirable to have a borrower know that he or she is at a
particular delinquency risk.
[0079] For the sake of example, the base credit scores are
arbitrarily assigned a value from 1.00 to 5.00 (changing in 0.25
increments). Then, each probability matrix (FIG. 5a-5e) can be
translated to tables consisting of base credit scores. In this
example, 1.00 base credit score will represent the least
probability of delinquency; a 5.00 will represent the maximum
probability of delinquency which the company has determined to
assume for any individual loan. Table 2, below provides one way of
cross referencing the base credit scores to the probability of
delinquencies. TABLE-US-00002 TABLE 2 Risk of Credit Score
Delinquency 1.00 0 1.25 .0156 1.50 .0312 1.75 .0468 2.00 .0624 2.25
.0780 2.50 .0936 2.75 .1092 3.00 .1248 3.25 .1404 3.50 .1506 3.75
.1716 4.00 .1872 4.25 .2028 4.50 .2184 4.75 .2340 5.00 .2496
[0080] With the correlation data from Table 2, the matrixes of
FIGS. 5a-5e can be converted to the tables of 6a-6e, respectively.
Thus, given a credit grade, a credit reporting score, and an LTV,
it is possible to determine a base "credit score" from the tables
illustrated in FIGS. 8a-8e. For instance, using a known FICO score
and a LTV for the loan in question, a user may determine the base
credit score by reading the corresponding value from the
appropriate table. The user simply locates the FICO score on the
vertical axis to determine a row, locates the LTV ratio on the
horizontal axis a column, and the base credit score will be found
in the cell which intersects the respective row and column. Thus,
the user will be able to determine the base credit score from three
different factors (payment history, FICO score, and LTV) without
having to use complex equations.
[0081] Similarly, the risks of delinquency due to add-on factors
may also be converted into a credit score. With the information
derived above, an add-on table (such as in FIG. 6) can then be
created. Thus, once a user has determined the base score, the user
may simply add to the base score any applicable add-on factors to
arrive at a total credit score. In the present example, each risk
factor has an add-on ranging from 0.25 to 3.25. This number is
added to the base score to determine the final "Credit Score".
[0082] Rates sheets may then be developed based on the final credit
score. An example rate sheet is illustrated in FIG. 9. Rate sheets
are commonly used throughout the industry.
[0083] The process and information provided above may be
distributed in various manners. It may be part of a computer
program, such as can be performed by the computer 50 of FIG. 2. It
also can be physically published in tables and charts for use. The
information can include: 1) Credit Grade Determination; 2) Matrices
for each Credit Grade showing LTV and FICO score yielding Credit
Score; 3) a Table for add-ons; and 4) Pricing sheets showing the
appropriate rate for various Credit Grade and LTV combinations (if
applicable, separating out different loan products and states).
This information can be used in method 10 of FIG. 1, as discussed
above.
[0084] It is understood that modifications, changes and
substitutions are intended in the foregoing disclosure and in some
instances some features of the disclosure will be employed without
corresponding use of other features. Accordingly, it is appropriate
that the appended claims be construed broadly and in a manner
consistent with the scope of the disclosure.
* * * * *