U.S. patent application number 12/573834 was filed with the patent office on 2010-04-08 for method for mortgage fraud detection.
This patent application is currently assigned to FIRST AMERICAN CORELOGIC, INC.. Invention is credited to Christopher L. Cagan.
Application Number | 20100088242 12/573834 |
Document ID | / |
Family ID | 34573693 |
Filed Date | 2010-04-08 |
United States Patent
Application |
20100088242 |
Kind Code |
A1 |
Cagan; Christopher L. |
April 8, 2010 |
METHOD FOR MORTGAGE FRAUD DETECTION
Abstract
A method of detection of fraud in a mortgage application: in a
computer system, maintaining a database of sales prices of real
properties in a geographic area where the property is located;
obtaining a valuation history for the property; obtaining
historical sales data for similar properties in the geographic
area; computing price ratios using these valuation histories;
computing a distortion index based on the price ratios, the
distortion index indicating the likelihood of a fraudulent
valuation.
Inventors: |
Cagan; Christopher L.; (Los
Angeles, CA) |
Correspondence
Address: |
SNELL & WILMER L.L.P. (1st AMERICAN RE)
600 ANTON BOULEVARD, SUITE 1400
COSTA MESA
CA
92626
US
|
Assignee: |
FIRST AMERICAN CORELOGIC,
INC.
Santa Ana
CA
|
Family ID: |
34573693 |
Appl. No.: |
12/573834 |
Filed: |
October 5, 2009 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
10713348 |
Nov 14, 2003 |
7599882 |
|
|
12573834 |
|
|
|
|
Current U.S.
Class: |
705/306 ;
705/317; 707/724; 707/E17.11 |
Current CPC
Class: |
G06Q 30/0185 20130101;
G06Q 40/025 20130101; G06Q 40/02 20130101; G06Q 30/018 20130101;
G06Q 50/167 20130101; G06Q 30/0278 20130101 |
Class at
Publication: |
705/306 ;
705/317; 707/724; 707/E17.11 |
International
Class: |
G06Q 40/00 20060101
G06Q040/00; G06Q 10/00 20060101 G06Q010/00; G06Q 50/00 20060101
G06Q050/00; G06F 17/30 20060101 G06F017/30 |
Claims
1. A method of detecting fraud during a real estate transaction,
the method comprising: using a computer processor to: receive an
estimated value of a subject real property; access a database of
spatial data, the spatial data comprising real estate prices in a
geographic area in which the subject real property is located, the
geographic area comprising at least one of the same zip code, city,
or county as the subject real property; generate a first spatial
variance by computing ratios corresponding to the accessed real
estate prices; generate a second spatial variance by computing
ratios corresponding to the accessed real estate prices; and
compute a spatial distortion based on the first spatial variance
and the second spatial variance to indicate a likelihood of
fraud.
2. The method of claim 1, wherein the spatial data is generated by
using sales data for properties in the geographic area.
3. The method of claim 1, wherein the spatial data is generated by
using an automated valuation model for properties in the geographic
area.
4. The method of claim 1, wherein the spatial data is generated by
using a combination of sales data and an automated valuation model
for properties in the geographic area.
5. The method of claim 1, wherein the spatial data is generated
from real estate prices from previous years.
6. The method of claim 1, wherein the spatial data comprises
property characteristics that are shared between the subject real
property and properties in the geographic area.
7. The method of claim 1, wherein one of the spatial variances
comprise a ratio of the estimated value of the subject real
property and a median real estate price of real property in the
same zip code.
8. The method of claim 1, wherein the spatial distortion comprises
a difference between two years of spatial variances of the subject
real property.
9. The method of claim 8, wherein the two years comprise the
current year and a previous year having the largest spatial
variance.
10. A method of detecting fraud during a real estate transaction,
the method comprising: using a computer processor to: receive an
estimated value of a subject real property; access a database of
spatial data, the spatial data comprising real estate prices in a
geographic area in which the subject real property is located, the
geographic area comprising at least one of the same zip code, city,
or county as the subject real property; generate a set of spatial
variances by computing ratios between a plurality of real estate
prices of the subject real property and a plurality of real estate
prices of properties in the geographic area; compute a spatial
distortion based on the set of spatial variances; and produce a
distortion ratio score to indicate a likelihood of fraud based on
the spatial distortion.
11. The method of claim 10, further comprising accessing a set of
temporal data.
12. The method of claim 11, further comprising generating a set of
temporal variances, computing a temporal distortion based on the
set of temporal variances, and computing a total distortion by
adding the temporal distortion to the spatial distortion.
13. A system of detecting fraud during a real estate transaction,
the method comprising: a computer processor; and a memory storing
program instructions, said program instructions when executed by
the computer processor causes the computer processor to: receive an
estimated value of a subject real property; access a database of
spatial data, the spatial data comprising real estate prices in a
geographic area in which the subject real property is located, the
geographic area comprising at least one of the same zip code, city,
or county as the subject real property; generate a set of spatial
variances by computing ratios between a plurality of real estate
prices of the subject real property and a plurality of real estate
prices of properties in the geographic area; compute a spatial
distortion based on the set of spatial variances; and produce a
distortion ratio score to indicate a likelihood of fraud based on
the spatial distortion.
14. The system of claim 13, wherein the spatial data is generated
by using a combination of sales data and an automated valuation
model for properties in the geographic area.
15. The system of claim 13, wherein the spatial data is generated
from real estate prices from previous years.
16. The system of claim 13, wherein the spatial data comprises
property characteristics that are shared between the subject real
property and properties in the geographic area.
17. The system of claim 13, wherein one of the spatial variances
comprise a ratio of the estimated value of the subject real
property and a median real estate price of real property in the
same zip code.
18. The system of claim 13, wherein the spatial distortion
comprises a difference between two years of spatial variances of
the subject real property.
19. The system of claim 18, wherein the two years comprise the
current year and a previous year having the largest spatial
variance.
20. A method of detecting fraud during a real estate transaction,
the method comprising: using a computer data processor to: access a
database of real property prices in a geographic area in which a
subject real property is located; using data from the database or
data from a requestor to generate a temporal data set comprising a
current yearly real property price for the subject real property
and a set of past yearly real property prices for the subject real
property; generate from the database a spatial data set comprising
a current yearly real property price for real property with similar
characteristics as the subject real property and a set of past
yearly real property prices for real property with similar
characteristics as the subject real property; generate a set of
temporal variances; generate a set of spatial variances; compute a
spatial distortion based on the set of spatial variances; compute a
temporal distortion based on the set of temporal variances; and
compute a total distortion by adding the temporal distortion to the
spatial distortion.
Description
CROSS REFERENCE TO RELATED APPLICATION
[0001] This application is a continuation of and claims the benefit
and priority of application Ser. No. 10/713,348, entitled "Method
for Mortgage Fraud Detection," filed Nov. 14, 2003, now U.S. Pat.
No. 7,599,882, which is assigned to the assignee hereof and hereby
expressly incorporated by reference herein.
BACKGROUND
[0002] 1. Field
[0003] The present invention relates to a method of detecting fraud
in loan application. More particularly, the present invention
relates to a method of estimating the risk associated with a
contemplated loan, and especially to estimating the risk that a
lender may be induced to rely on an unrealistically high estimate
of the value of the real property that is to secure the loan.
[0004] 2. Description of the Related Art
[0005] For the purpose of controlling the risk associated with
lending money secured by real property, a loan originator attempts
to estimate the value of the property being used to secure the
note. Traditionally, the originator paid an appraiser, who was
supposed to be knowledgeable in the type of real property in
question and skilled in comparing such properties, and relied on
the appraiser's estimate of the market value of the real property
in order to limit the risk that value would be inadequate to secure
the note. The use of appraisals continues. In recent years, lenders
who wish to rely less on the appraisal have begun using "automated
valuation models" ("AVMs"), methods of estimating the market value
of a property based on various methodologies such as price indexing
methods, hedonic models, adjusted tax assessed value models, and
hybrid models.
[0006] A lender may also inquire whether the property has undergone
certain patterns of frequent sales, loans or refinancings which
have, in that lender's experience, come to be associated with
attempts to cause artificially high estimates of the value of the
property. Additionally, the lender may investigate the
creditworthiness of the person applying for the loan. Finally, the
lender might seek out information about the applicant's history,
taking a particular interest in whether the applicant has been
involved in a cluster of activities involving real property in the
neighborhood.
[0007] Thus, existing methods for fraud detection tend to emphasize
either or both of (1) the history of the subject property (the
property proposed for a sale or loan), with a special view to any
possible "flipping" (rapid series of sales, loans, or refinances),
and (2) the history and creditworthiness of the applicant, with a
special view to any other transactions in the neighborhood of the
subject property.
[0008] These methods are not infallible. An appraiser charges a
hefty fee, usually requires several days or more to deliver an
appraisal, and occasionally turns out to be incompetent, gullible,
or corrupt. Persons attempting to inflate the estimated value of a
property have been known to engage in patterns of sham sales of the
subject property or of nearby properties. They may also act in
concert with others to create in the mind of a purchaser or a
lender a false impression that properties in the area are
appreciating rapidly. Such tactics might also have the effect of
feeding artificially inflated values to the automated valuation
models that the lender is relying on.
[0009] An inexperienced or careless loan officer may be taken in by
such schemes. Unfortunately, even a more wary loan officer may
hesitate to deny the application or to demand additional
information. Lenders are under pressure to avoid the appearance
that they are engaging in unfair discrimination against classes of
applicants or against neighborhoods which are perceived to be
underserved by the banking industry. The lender may fear being sued
and being forced at great expense to prove an objective basis for
denying an application.
[0010] It is therefore necessary for lenders to have more
efficient, reliable, to objective means of controlling the risk of
being victimized by mortgage fraud.
SUMMARY
[0011] It is an object of the present invention to improve a
lender's ability to control the risks associated with mortgage
fraud while also controlling the costs of avoiding those risks.
[0012] In accordance with these objects and with others which will
be described and which will become apparent, an exemplary
embodiment of a method for mortgage fraud prevention in accordance
with the present invention comprises the steps of maintaining a
database of sales prices in the computer system of a plurality of
real properties in a geographic area in which the subject real
property is located; obtaining from the computer system, valuation
history data for the subject property; obtaining, using the
computer system, historical sales data for property in the
geographic area in which the subject real property is located;
computing price ratio data using the valuation history for the
subject property and the historical sales data for the subject
property in the geographic area in which the subject property is
located, and computing a distortion index based on the price ratio
data to detect fraud in the mortgage application.
[0013] The distortion index may include a temporal distortion
index, a spatial distortion index, or a combination of these.
[0014] The matrices of data that are assembled in the process of
preparing the distortion index may also be reported.
[0015] The method may be applied prospectively or retrospectively,
with single properties or with large numbers of properties, and
with the aid of varying automated valuation models.
[0016] The method may be applied by a person far removed in time
and place from the transaction in question and having no particular
connection to it. Because a person using the method does not need
to single out any neighborhood or other geographical area as a
special danger area for fraud, the method in accordance with the
present invention helps reduce the danger that a lender would be
accused of improper exclusion, such as redlining.
BRIEF DESCRIPTION OF THE DRAWINGS
[0017] For a further understanding of the objects and advantages of
the present invention, reference should be had to the following
detailed description, taken in conjunction with the accompanying
drawing, in which like parts are given like reference numbers and
wherein:
[0018] FIG. 1 is a block diagram of the overall process and
information flow for an exemplary method of detecting mortgage
fraud in accordance with the present invention;
[0019] FIG. 2 is a block diagram of input steps for an example of
the method of detecting mortgage fraud in accordance with the
present invention;
[0020] FIG. 3 is a block diagram of computation steps for an
example of the method of detecting mortgage fraud in accordance
with the present invention;
[0021] FIG. 4 is a block diagram of report steps for an example of
the method of detecting mortgage fraud in accordance with the
present invention; and
[0022] FIG. 5 is an exemplary set of matrices of data reported in
accordance with the present invention.
DETAILED DESCRIPTION
[0023] With reference to FIG. 1, an exemplary embodiment of the
method of detecting mortgage fraud in accordance with the present
invention utilizes a general purpose computer with access to a
database of sales price information pertaining to a geographical
area where a parcel of real property is situated. The method
obtains the identification of the subject property. The data
processor 21 obtains from the requesting party 23 the location and
characteristics of the property along with the proposed valuation
of the property. The property characteristics and location are used
to formulate a query to a database 27 of historical median price
data for the zip code, city, and county in which the property is
situated. The city may be the "postal city," which is the default
United States Postal Service city name for the subject property's
zip code. Alternatively, the city may be the situs city, which
includes the literal bounds of the city. The results of the query
are used to create a matrix of spatial price information. From the
requesting party 23 or from the database 27, the data processor 21
also obtains the time and price of a prior sale or valuation of the
subject property. The data processor 21 passes this information to
the automated valuation model 25, which returns a set of yearly
price values for the subject property. The data processor 21 then
computes a spatial distortion index, a temporal distortion index,
and a total distortion index. Finally, the data processor 21
reports at least the total distortion index to the requesting
party.
[0024] The exemplary embodiment of the method of detecting mortgage
fraud in accordance with the present invention utilizes a general
purpose computer with access to a database of sales price
information pertaining to a geographical area where a parcel of
real property is situated.
[0025] The method in accordance with the present invention places
the subject property and its requested or alleged value for the
purpose of sale, loan, or refinance, within both a spatial and a
temporal context.
[0026] The method constructs a temporal context or "valuation
history" for the subject property. This context is both real and
virtual. It is real in that it includes all known valuations
attributed to prior sales, appraisals, and refinances of the
property. It is virtual in that it includes valuations of the
property, done by an automated valuation model (AVM), valuing the
property at regular intervals, such as yearly, back into the
past.
[0027] The method also constructs a spatial context for the subject
property at several geographic levels, obtaining median sale prices
for the property's zip code, city, and county from a high-quality
database, the data for which is usually obtained from the county
recorder's office. The database is queried for data pertinent to
properties sharing the characteristics of the subject property. For
example, if the subject property is a single family residence, the
database is queried for single family residences. If the property
is a condominium or townhouse and appropriately specific data are
available, the query will reflect this information. These prices
are extracted for the present and the past, for the time periods
used in building the temporal context.
[0028] In accordance with the present invention, the temporal and
spatial contexts together build a price level matrix. One column of
the matrix constitutes the prices of the subject property's
temporal context. The other columns of the matrix are built from
the prices of the various levels of the spatial context.
[0029] From the price level matrix it is possible to construct a
price ratio matrix. The rows of this matrix are assigned to the
time periods of the temporal context; the columns represent the
concentric levels of the spatial context. The entries in the matrix
represent the ratios of price levels (for the different time
periods) of subject property to zip code prices, subject property
to city prices, subject property to county prices, zip code to
city, and city to county.
[0030] Ratios outside of reasonable contextual levels are said to
be distorted. In the price ratio matrix, reasonable variations are
possible in the absence of fraud. Individual properties can be more
or less valuable than the main body of properties in their zip
code, city, or county. Prices in a zip code or city can be higher
or lower than those in their corresponding city or county. The
ratios can also vary according to basic principles of the business
cycle. For instance, it is generally accepted that during boom
times prices in affluent areas rise in greater proportions than do
prices in middle-class or poor areas, whereas during declining
markets, prices in affluent areas are subject to disproportionate
declines. Furthermore, price trends in affluent areas tend to
"lead" the rest of the market in that they pull out of a recession,
or stop rising near the end of a boom, before the main body of the
market does so.
[0031] Abnormal or distorted ratios may suggest the possibility of
fraud in a loan application, and indicate the wisdom of further
investigation such as an outside appraisal. If the past history of
a subject property suggests that its valuation is between 100% and
120% of the median value in its zip code, but the last two loan's
on the property were based on valuations which were first 180% and
then 250% of the median value in the zip code, this spatial ratio
distortion may suggest the possibility of fraud. In the same way,
if prices in the property's zip code have been rising at 5% to 7%
per year for the past few years, but the alleged valuations of the
subject property represent an increase of 40% over the previous
year's valuation, this temporal ratio distortion may suggest the
possibility of fraud.
[0032] The individual temporal ratio distortions and spatial ratio
distortions can be used to construct a distortion ratio score which
may be expressed in numerical or letter-grade form to suggest the
presence of unusual numbers and the possibility of fraud.
[0033] With reference to FIGS. 2-5, an example of the method of
detecting mortgage fraud in accordance with the present invention
utilizes a general purpose computer with access to a database of
sales price information pertaining to a geographical area where a
parcel of real property is situated.
[0034] With particular reference to FIG. 2, the process begins when
information is obtained as shown at reference number 41 about the
subject property including its address, the identity of its owner,
and physical attributes such as the type of structure, square
footage, lot size, and the like. Also obtained (see at 43) is the
valuation that is being requested or proposed by the applicant. The
process will now be described by way of example as of a time-point
of Jul. 1, 2003 for two proposed market values: a realistic value
and an unrealistically high value. In this example, it will be
assumed that for the property involved, a value of $400,000 is
reasonable and $550,000 is unrealistically high.
[0035] At this point, the proposed value and the time of the
proposed value may be tabulated for the two alternative values as
follows:
TABLE-US-00001 TIME PRICE COMMENT Jul. 1, 2003 $400,000 realistic
requested Jul. 1, 2003 $550,000 unrealistic requested
[0036] With reference to FIG. 3, at 45, the process next obtains
data about previous sales, appraisals, or refinance valuations for
the subject property, along with date and source of those
valuations. In this example, the subject property was sold in 1994
for $180,000. This prior sale may be represented as follows:
TABLE-US-00002 TIME PRICE COMMENT 1994 $180,000 prior sale
[0037] The process next calls upon an automated valuation model to
produce a valuation of the subject property as of the date of the
request and at one-year intervals into the past to the extent
permitted by the available data. In this example, the automated
valuation model being used is ValuePoint.RTM.4 ("VP4"), a
well-known service provided by First American Real Estate
Solutions. It can be seen that the VP4 automated valuation model
has generated a series of estimated price values for the subject
property for Jul. 1, 1999 ($268,000) through Jul. 1, 2003
($395,000). These data, which we may refer to as temporal
information concerning the subject property, may be tabulated as
follows:
TABLE-US-00003 TIME PRICE COMMENT 1994 $180,000 prior sale 1995
1996 1997 1998 1999 $268,000 automated value - VP4 AVM for July 1
2000 $276,000 automated value - VP4 AVM for July 1 2001 $335,000
automated value - VP4 AVM for July 1 2002 $371,000 automated value
- VP4 AVM for July 1 2003 $395,000 automated value - VP4 AVM for
July 1 2003 $400,000 Requested July - realistic 2003 $550,000
Requested July - unrealistic
[0038] At 47, the process next builds one or more columns of
information about the economic performance of other properties
having various spatial relationships to the subject property. In
this example, the process builds columns of median price data for
properties having characteristics similar to those of the subject
property. For example, median price data are obtained for
properties of the same type as the subject property, such as
"single-family residence" or "condominium/townhouse." Further
refinements based on studying detailed property characteristics are
also possible. These data are obtained or computed from a database
of available reported sales, appraisals, and the like in the same
zip code, postal city, and county as the subject property. The
postal city is the default city name as used by the United States
Postal Service for the zip code in which the property is situated.
Alternatively, the situs city of the subject property could be
used: the city inside whose borders the subject property is legally
situated.
[0039] These median price statistics are to come from a standard
and reliable source. These numbers must apply to the same type of
residential property as the subject property (single-family
residence, or condo/townhouse, etc.). These numbers are for time
periods to correspond with the months and years that appear in the
subject property columns. For dates in the most recent year, the
process uses tri-monthly medians for the most recent completed
three months. For dates in previous years, the process uses yearly
medians.
[0040] The process unites the information into a spatial price
median matrix, which may be tabulated as follows:
TABLE-US-00004 PRICE OF PRICE OF PRICE OF PRICE OF PROPERTIES
PROPERTIES PROPERTIES SUBJECT IN SAME IN SAME IN SAME TIME PROPERTY
ZIP CODE POSTAL CITY COUNTY 1994 $180,000 $175,000 $150,000
$160,000 1995 $167,000 $139,000 $153,000 1996 $163,000 $134,500
$150,000 1997 $178,000 $136,000 $150,000 1998 $195,000 $138,000
$162,000 1999 $268,000 $244,000 $142,000 $175,000 2000 $276,000
$261,000 $149,000 $192,000 2001 $335,000 $310,000 $170,000 $229,000
2002 $371,000 $388,000 $220,000 $267,000 2003 $395,000 $410,000
$241,000 $297,000 2003 $400,000 (proposed) 2003 $550,000
(proposed)
[0041] At 49, for each time value (each year), the process computes
ratios of the subject property value (price) to the median value
reported in the various reference data sets (here, the reference
data sets are the median price data for the same zip code, same
postal city, and same county as the subject property). The result,
which may be referred to as a matrix of spatial variances, may be
tabulated as follows:
TABLE-US-00005 RATIO, %, RATIO, %, RATIO, %, RATIO, %, RATIO, %,
RATIO, %, SUBJECT SUBJECT SUBJECT ZIP TO ZIP TO CITY TO TIME TO ZIP
TO CITY TO COUNTY CITY COUNTY COUNTY 1994 103 120 113 117 109 94
1995 120 109 91 1996 121 109 90 1997 131 119 91 1998 141 120 85
1999 110 189 153 172 139 81 2000 106 185 144 175 136 78 2001 108
197 146 182 135 74 2002 96 169 139 176 145 82 2003 96 164 133 170
138 81 2003* 98 166 135 170 138 81 2003** 134 228 185 170 138 81
*proposed realistic value **proposed unrealistic value
[0042] From this table, it is evident that the historical value and
the automated valuation of the subject property are typical of the
values of similar properties in the zip code. It also appears that
such properties in the subject property's zip code are more
expensive than those in the city or county as a whole, and have
risen faster in recent years than those in the city or county as a
whole.
[0043] It also is evident that the ratios computed from the
unrealistic valuation (134%, 228%, 185%) stand out somewhat from
the rest of the matrix. The unrealistic valuation is apparent when
measured arithmetically by subtraction. For example, by way of
subtraction, 134% is 24% higher than 110%, the maximum ratio found
in the column above it. Similarly, 228% is 31% higher than 197%,
the maximum ratio in the column above it, and 185% is 32% higher
than 153%, the maximum ratio in the column above it.
[0044] Alternatively, the ratios may be measured arithmetically by
division. For example, a comparison by way of division of the
realistic and unrealistic value ratios of the subject property
divided by their respective maximum prior ratios are as
follows:
TABLE-US-00006 Realistic Unrealistic 98%/110% = 89% 134%/110% =
122% 166%/197% = 84% 228%/197% = 116% 135%/153% = 88% 185%/153% =
121%
[0045] For purposes of the present example, however, this
alternative measurement of ratios by division will not be used in
the subsequent steps of the process.
[0046] At 53, the process generates a matrix of temporal variances
by computing the following ratios for each time period: [0047]
Subject value to subject value one year prior; [0048] Zip code
median to zip code median one year prior; [0049] City median to
city median one year prior; and [0050] County median to county
median one year prior.
[0051] "One year prior" in this example means the complete year
previous to the year in question. The temporal variances are
tabulated as follows:
TABLE-US-00007 ratio, ratio, ratio, ratio, %, subject, %, zip, %,
city, %, county, over year over year over year over year time
previous previous previous previous 1994 1995 95 93 96 1996 98 97
98 1997 109 101 100 1998 110 101 108 1999 125 103 108 2000 103 107
105 110 2001 121 119 114 119 2002 111 125 129 117 2003 106 106 110
111 2003* 108 106 110 111 2003** 148 106 110 111 *proposed
realistic value **proposed unrealistic value
[0052] It is noticeable that the temporal variance associated with
the unrealistic valuation (148%) stands out as inconsistent with
the rest of the matrix, while the temporal variance associated with
the realistic valuation (108%) does not.
[0053] At 51, the process identifies distortions that are
associated with the subject property valuation. The spatial
distortion is the amount that the subject to zip ratio exceeds the
maximum of the entries in the column above it. For the realistic
valuation of $400,000, that distortion is zero, since 98% does not
exceed 110%. For the unrealistic valuation of $550,000, that
distortion is 24%, since 134%-110%=24%.
[0054] In some zip codes, there might not be enough data on
historical prices of comparable properties to produce a reliable
matrix. In such a situation, where, for example, the average number
of sales in a zip code per year is less than 100, the process may
use the subject to city ratio or the subject to county ratio in
place of the subject to zip ratio.
[0055] At 55, the process computes a temporal distortion, which is
defined as the percentage change in subject valuation from the
prior year, minus the percentage change in the zip code valuation
from the prior year. For the realistic valuation, the temporal
distortion is 108%-106%=2%. For the unrealistic valuation, the
temporal distortion is 148%-106%=42%. As before, if the average
number of sales in a zip code per year is less than 100, the
process may use the subject to city ratio or the subject to county
ratio in place of the subject to zip ratio.
[0056] It should be pointed out that an alternative index of
temporal distortion is computable as the ratio of the percentage
change in subject valuation from the prior year to the percentage
change in the zip code valuation from the prior year. For the
purposes of the present example, however, this alternative
distortion index will not be used in the subsequent steps of the
process.
[0057] At 57, the process computes the total distortion, which is
defined as the sum of the spatial and temporal distortions. For the
legitimate valuation, the total distortion is 0%+2%=2%. For the
fraudulent valuation, it is 24%+42%=66%.
[0058] With reference to FIG. 4, at 59, the process reports this
total distortion to the customer. Optionally, at 61, the reported
information also includes the spatial and temporal components
individually and, if it is deemed appropriate, some or all of the
matrix data that gave rise to the computation as shown in FIG.
5.
[0059] A total distortion above 50% usually suggests an unrealistic
valuation and hence a possible fraud. A total distortion above 100%
almost always merits further investigation.
[0060] Optionally, the process allows the customer to specify any
desired total distortion value which would flag an application as
likely to be fraudulent.
[0061] If the customer or user requests it, the price, ratio, and
distortion matrices can be offered as supporting documentation. If
no fraud is suspected, this step will usually not be necessary. If
a fraud is suspected, this supporting documentation can be very
useful to support and guide any lending or other decision. It is a
particular advantage of the present invention that it avoids the
inference of redlining or other geographic discrimination, because
the person using the method does not need to single out any
neighborhood or other geographical area as a special danger area
for fraud and because the data produced in accordance with the
present invention is "mechanistic" rather than subjective or
personal.
[0062] FIG. 5 exemplifies documentation the present invention makes
available where a valuation is suspected to be fraudulent. Such
information can be delivered efficiently to the customer along with
other information such as the proposed valuation and the name
address of the applicant and the address and relevant information
concerning the subject property.
[0063] Many lenders will be satisfied with such a score or grade,
especially if the findings suggest normal price levels and thus a
low likelihood of fraud. In some cases, lenders will request the
supporting documentation of the price level matrix and price ratio
matrix, especially in cases where a high distortion ratio score
suggests the possibility of fraud.
[0064] The requirements of the method in accordance with the
present invention are not the same as the requirements of existing
fraud detection methods. The requirements are twofold: (1) the
availability of a high quality automated valuation model (AVM)
which is able to value a subject property in current time and at
selected times in the past; and (2) a database of median price
statistics at several geographic levels such as the zip code, city,
and county where the subject property is situated. It is a
significant advance to recognize how these methods can be combined
and to devise an effective way of combining them to detect mortgage
fraud with improved accuracy and efficiency. The method in
accordance with the present invention astutely combines these two
components to provide a number of advantages over existing fraud
detection methods. For example, the method in accordance with the
present invention can provide documentation if a lender is called
upon to support a decision on a loan application. The efficiency of
automated data processing reduces the cost of producing such
documentation when it is called for.
[0065] The scores and matrices produced by the method in accordance
with the present invention can be requested and transmitted in an
automated fashion for one property at a time or for thousands or
millions of properties. The method can be used whenever needed, for
subject properties and applications throughout the country. Thus,
the method can achieve economies of scale.
[0066] An entire bundle of loans may be analyzed pursuant to one
request. The loans may be proposed loans, or they may be loans that
were made in the past. Thus, the fraud sought to be detected may be
prospective, or it may already have occurred. One may, therefore,
use the present invention to re-evaluate the risk entailed in
choices that were made in the past or that were made by others who,
at the time, had a very different appreciation of the sensitivity
and selectivity with which others later would evaluate the risk of
fraud.
[0067] The method in accordance with the present invention also
embodies a recognition of the tactics and limitations of the
perpetrators of mortgage fraud, with a view to making fraud much
more difficult to accomplish and rendering fraudulent alleged
values easier to detect. For example, perpetrators of fraud
("fraudsters") are frequently able to arrange a series of real or
alleged sales, appraisals, or loans for a subject property, often
at ascending valuation levels, to support their claims of value for
a sale or loan. Fraudsters also often arrange a series of real or
alleged sales, appraisals, or loans for a set of properties in a
single neighborhood, often at ascending valuation levels, to
support their claims of ascending market valuations or prices, and
to support their alleged valuation of a subject property for the
purposes of a sale or loan.
[0068] However, fraudsters will not have the resources to arrange
enough false valuations or false sale prices to distort the overall
price levels in a well-populated zip code, much less in an entire
city or county, because there are many legitimate sales in such
areas. Fraudsters also probably will not have the persistence to
arrange a history of false valuations or prices on a subject
property that extends several years into the past. In general,
fraudsters are unlikely to spend three, five, or more years to
build a trail of false valuations of a subject property. Combining
a spatial distortion index with a temporal distortion index places
the concealment of the attempted fraud beyond the resources and
beyond the patience of fraudsters.
[0069] Thus, an advantage of the method in accordance with the
present invention is that it uses a combination of distortion ratio
methods to differentiate realistic valuations, which are likely to
result from underlying market phenomena, from various types of
unrealistic valuations that are likely to result from the behaviors
that fit the methods and motivations of fraudsters. The various
advantages of the method combine to result in a substantial, novel,
non-obvious advance over existing methods of detecting fraud in a
mortgage application.
[0070] The method in accordance with the present invention is also
useful in identifying past transactions which gave rise to
anomalous valuations. For example, one may wish to investigate the
history of a particular property not only for indications of
mortgage fraud, but also for indications that any party to a
transaction recognized a value inconsistent with market
conditions.
[0071] For this purpose, a database of valuation-generating events
is maintained in a computer system for a plurality of real
properties in a geographic area in which the subject real property
is located. A valuation-generating event may be a sale, a valuation
pursuant to a mortgage application, or any other event indicating a
value placed on the property by any party. A subject property
record set including a price (or other valuation) of the property
and the date thereof for at least one valuation-generating event is
obtained from the computer. A valuation history data set for the
subject property, comprising prices and the dates thereof for
several timepoints, is obtained from the computer. This step may
involve obtaining values generated by automated valuation models as
well as values derived from other sources of data. Finally,
historical sales data for property in the geographic area in which
the subject real property is located is obtained from the computer.
This data is obtainable, for example, from county recorders. To
obtain suitable data, a database query may be formulated comprising
such information as the location, type, and other economically
important characteristics of the subject property, to the extent
appropriate in light of the available data.
[0072] In the manner described herein, price ratio data is computed
using the valuation history for the subject property and the
historical sales data for the subject property in the geographic
area in which the subject property is located. Also in the manner
described herein, a distortion index is computed based on the price
ratio data to detect an anomalous valuation in the at least one
valuation generating event.
[0073] However, instead of computing and reporting a distortion
index for a single property for the purpose of detecting fraud in a
pending mortgage application, a distortion index is computed and
displayed for each of a plurality of conditions so that patterns
and relationships may be revealed. For example, for a single
identified property the distortion index may be reported as a
function of time. Alternatively, a particular timeframe may be
selected and the distortion index may be reported as a function of
other variables or combinations of variables. The variables may
include, for example, the identity of the transferor or the
transferee, the identity of the lender or loan officer, the
geographical location, the municipal jurisdiction, or any other
descriptive information associated with the property or with the
transaction. It is possible, therefore, to associate certain
individuals, brokers, lenders, environments or timeframes with
certain valuation patterns that signal an anomalous valuation.
[0074] Thus, the method in accordance with the present invention
becomes a potent tool for risk management in secondary sales of
financial instruments, as well as for historical and economic
research and other investigative purposes.
[0075] It will be appreciated that many variations are possible for
practicing the invention without departing from the spirit of the
present invention, whose scope is to be limited only by the claims
appended to this specification.
* * * * *