Colorado Foreclosures

1
Colorado Foreclosures

• A New Way Of Visualizing Risk

Jeffrey Ayres M.S. GIS Candidate Faculty Advisor
Professor James Murdoch April 30, 2007
2
Introduction
3
(No Transcript)
4
Loan Performance, Market Pulse, June 30, 2006.
5
RealtyTrac Rankings Top 25 As of December 31,
2006
6
Colorado reclaims top foreclosure rate Colorado
posted the nations highest state foreclosure
rate in December 2006 one new foreclosure
filing for every 376 households. The state
registered the highest monthly rate for the ninth
time in 2006, reclaiming the top spot back from
Nevada. http//www.realtytrac.com/ContentManagem
ent/pressrelease.aspx?ChannelID9ItemID1742accn
t64847
Top three metro foreclosure rates are located in
Colorado, Texas and Michigan For the fifth month
in a row, Greeley, Colo., posted the highest
foreclosure rate among the nations 200-plus
largest metropolitan areas. The Greeley metro
area (Weld County) documented 391 properties
entering some stage of foreclosure, a decrease of
nearly 9 percent from the previous month and a
foreclosure rate of one new foreclosure filing
for every 169 households more than six times
the national average. http//www.realtytrac.com/
ContentManagement/pressrelease.aspx?ChannelID9It
emID1742accnt64847
7
(No Transcript)
8
How Did We Get Here?

• The 2001-2003 surge in mortgage demand prompted
  lenders to expand their operations to boost
  capacity. These conditions also attracted new
  market participants, often lenders with little
  business experience or financial strength. When
  loan demand slowed in 2004, the market was left
  with overcapacity. To maintain production levels,
  and satisfy continued strong investor appetite,
  mortgage originators shifted to innovative
  products, often designed to help borrowers cope
  with rising home prices or continue to tap idle
  home equity. Some of these innovations
  included relaxed underwriting standards and
  temporary payment reductions that increased risk
  for both borrowers and lenders.
• Statement of E. Wayne Rushton Senior Deputy
  Comptroller, Office of the Comptroller of the
  Currency, Committee on House Financial Services
  Subcommittee on Financial Institutions and
  Consumer Credit, March 27, 2007.

9
Who Is Originating Subprime Mortgages? About
16.8 of Colorados foreclosed home loans in 2006
originated from bank lenders whereas 77.5
originated from non-bank lenders and 5.6
originated from bank affiliates.
Colorado Foreclosure Analysis, 2006, The Colorado
Bankers Association
10
Prepared by the Denver Office of Economic
Development Policy Group January 2007
11
Problem Statement
It is important to understand why rates of
foreclosures differ spatially.  Existing research
has shown that minorities, the economic
disadvantaged and people with low education
levels experience higher rates of foreclosure. 
But, there may be other factors. By controlling
for minority status, income, educational levels,
and age, I test if there is still a spatial
aspect in foreclosures.  HO The spatial
distribution of foreclosures is explained by
median family income, percent minority,
educational level, and age.  H1 factors used
to model foreclosures do not completely explain
the spatial pattern of foreclosures . 
12
My Contribution

• Address level foreclosure data is very new
  information and not available to the general
  public. My analysis takes advantage of the
  granularity of this data to identify
  neighborhoods that are most at risk for
  foreclosure and offers the possibility of
  conducting this type of analysis on a nationwide
  level.

13
Data Sources
14
(No Transcript)
15
Who is RealtyTrac
Ranked as the third largest real estate site by
MediaMetrix and No. 53 on Inc. magazines 2006
Inc. 500 list of the nations fastest-growing
private companies, RealtyTrac Inc. is the leading
online marketplace for foreclosure properties,
providing all the resources that home seekers,
investors and real estate agents need to locate,
evaluate and buy properties below market
value.1 Founded in 1996, RealtyTrac publishes
the largest and most comprehensive national
database of pre-foreclosure, foreclosure, For
Sale By Owner, resale and new construction
properties, with more than 1 million properties
across the country, property reports,
productivity tools and extensive professional
resources. RealtyTrac hosts nearly 3 million
unique visitors monthly and has been chosen to
supply foreclosure data to MSN Real Estate,
Yahoo! Real Estate and The Wall Street Journals
Real Estate Journal.2 1 http//www.realtytrac.co
m/ 2 IBID.
16

• RealtyTrac Methodology
• Obtains data directly from public records in
  over 2,500 counties, representing 74 percent of
  all U.S. counties and 94 percent of U.S.
  households.
• Data is reported nationally, on a
  state-by-state, county-by-county and zip code
  basis.
• Default foreclosure filings include new Notices
  of Trustees Sales and Judgments of Foreclosure
  Sales entered into the database each month.
• Auction foreclosures include new Notice of
  Trustees Sales and Judgments of Foreclosure
  Sales entered into the database each month.
• Bank-Owned foreclosure filings include new REO
  (Real Estate Owned) properties entered into the
  database each month.
• Foreclosure rate is total number of new default
  and auction foreclosure filings and new
  Bank-Owned foreclosures (REOs) divided by total
  number o f U.S. households (Census Bureaus
  Housing Units).

17
Variable Listing
Note ESRI Business Analyst (BA) includes over
1,500 demographic variables from the state to
census block level. The data is update yearly.
Estimates are validated against the Bureau of the
Census and supplemented with other consumer
growth and expenditure data.
18
Descriptive Statistics
19
Correlation
20
Literature Review
21
Spatial Analysis of Foreclosures

• There is a two-tiered approach for the
  investigation of geographical data. The first
  stage is exploratory spatial data analysis (ESDA)
  and the second involves confirmatory data
  analysis (CDA). The former is more descriptive
  than explanatory and deals with basic spatial
  autocorrelation relationships. The later deals
  with spatial dependence and omitted variables
  which can be discovered through the interpolation
  of regression residuals. This is an important
  aspect to my analysis as shown later in this
  study. Can 1998
• Neighborhood definition and identification is
  critical to ESDA and CDA. There are two methods
  for neighborhood identification (1) spatial
  contiguity (sharing a common boundary) and (2)
  distance between spatial entities. Can 1998
• Anselin draws comparisons between the academic
  functions ESDA and CDA and the commercial GIS
  world which examines spatial analysis in terms of
  selection and manipulation. Manipulation focuses
  on three groups attribute data, spatial data,
  and a combination of both called data
  integration. At the time of the report
  commercial software was not available to offer
  true spatial analysis for real estate policy and
  business applications Anselin 1998.

22
Statistical Analysis of Foreclosures

• Credit scores, recent minority buyers, and age
  offered the best predictive indicators of
  foreclosures. Of the three, credit scores were
  the strongest indicator. Grover, Smith, Todd
• Better data is needed at the neighborhood level
  to limit social losses associated with
  foreclosure. Grover, Smith, Todd

23
Public Policy

• The ramifications to foreclosures are
  significant for several reasons. One study
  concluded
• low-income households need additional support to
  remain homeowners
• benefits of homeownership are not as significant
  for low-income households compared to wealthier
  counterparts
• more research is needed to evaluate specifically
  at the experiences of low-income homeowners.
  Reid 2004
• It is estimated that each conventional
  foreclosure within an eighth of a mile of a
  single-family home results in a decline of 0.9
  in property value, averaging 159,000 per
  foreclosure for the City of Chicago. Immergluck
  and Smith 2005.
• Estimated losses on foreclosed properties range
  from 30 to 60 of the outstanding loan balances
  due to legal fees, foregone interest, and
  property expenses. Pence 2003.
• The direct costs to Chicago city government
  involve more than a dozen agencies and two dozen
  specific municipal activities, generating
  government costs that exceed 30,000 per property
  in some cases. Apgar and Duda 2005.

24
Analysis and Methodology
25
Methodology

• Geocode foreclosure locations as of December 31,
  2006 to address level 20k records for Colorado.
• Select key demographic information for each
  census tract 1075 tracts in Colorado.
• Remove any tracts without population 14 census
  tracts.
• Examine outliers (see next page).
• Spatially join all foreclosure points to
  appropriate census tract geographies.
• Calculate foreclosure rates based on owner
  occupied housing units for each census tract.
• Evaluate summary statistics and correlation for
  each variable.
• Calculate regression, residuals, and spatial
  autocorrelation as each new variable is added to
  the regression. I expect R2 to and T Stat to
  increase
• Calculate Morans I for each iteration residual
  values. I expect the Morans I to decrease as
  more variables are controlled thereby reducing
  the magnitude of the residuals.

26
Methodology Examine Outliers

• One census tract 080590098.31 was identified as
  an outlier that caused distortion to the
  analysis.
• This census tract reported 8 foreclosures
  compared with only 14 owner occupied homes. The
  majority of properties in this census tract were
  rental.

27
Morans I

• Where N is the number of casesXi is the variable
  value at a particular locationXj is the variable
  value at another locationX is the mean of the
  variableWij is a weight applied to the
  comparison between location i and location j

28
Results and Discussion

• Colorado Study

29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
Morans I for Foreclosure Locations Based on
Appraised Value
33
Foreclosures By Type
Greeley, Colorado Foreclosure Status
34
Foreclosures By Value
Greeley, Colorado Appraised Value
35
(No Transcript)
36
When Actual Foreclosures Are Much Higher Than
Predicted
Colorado Springs
Fort Carson
37
When Actual Foreclosures Are Much Lower Than
Predicted
Ski Resort Area
38
Residuals MFI
Residuals MFI/Min
Residuals MFI/Min/Ed
Residuals MFI - All
39
Conclusions
40
While Selected Independent Variables Improve The
Strength Of The Model, There Is Still Spatial
Clustering At A Statistically Significant Level.
41
Conclusions

• There is a statistically significant relationship
  between foreclosures and median family income,
  minority percentage, education, and age.
• Other measures such as credit scores could
  improve the predictive capability even more.
• The challenge is obtaining consistent and
  accurate detailed information, but for large
  geographic areas.
• There are likely other factors such as
  foreclosure laws and home price appreciation that
  will affect results.

42
Future Studies

• The FDIC is in the process of procuring credit
  score information at the census tract level.
  This information should be helpful in future
  analysis.
• Analyzing different aspects of the foreclosure
  process default, auction, and REO could provide
  early warning of future problems.
• RealtyTrac is trying to include identification of
  subprime mortgages based on the interest rate at
  the time of origination. Subprime mortgages are
  most at risk.
• Vintage analysis could differentiate between new
  mortgages and older (more seasoned) mortgages
  which do impact foreclosure rates. Grover,
  Smith, and Todd

43
Prepared by the Denver Office of Economic
Development Policy Group January 2007
44
References
45
References

• Apgar, WG, M Duda, RN Gorey (2005), The Municipal
  Cost of Foreclosures A Chicago Case Study,
  Preservation Foundation, Housing Finance Policy
  Research.Anselin, Luc (1998), GIS Research
  Infrastructure for Spatial Analysis of Real
  Estate Markets, Journal of Housing Research,
  Fannie Mae Foundation.
• Belsky, Eric, Can, Ayse, and Issac Megbolugbe
  (1998), A Primer on Geographic Information
  Systems in Mortgage Finance, Journal of Housing
  Research, Fannie Mae Foundation.
• Bureau of the Census. Current Population
  Survey/Housing Vacancy Survey, Series H-111
  Reports. http//www.census.gov/hhes/www/housing/hv
  s/historic/histt14.html
• Can, Ayse (1998), GIS and Spatial Analysis of
  Housing and Mortgage Markets, Journal of Housing
  Research.
• Colorado Foreclosures Analysis, 2006. Colorado
  Bankers Association and Development Research
  Partners, February 2007.
• Foreclosures in Denver Preliminary Findings
  (2007), Denver Office of Economic Development,
  Division of Housing and Neighborhood Development,
  February 2007.
• Grover, Michael and Richard M. Todd (2005), A
  Case for Post-Purchase Support Programs as Part
  of Minnesotas Emerging Markets Homeownership
  Initiative.
• Grover, Michael, Smith, Laura, and Richard M.
  Todd (2006), Targeting Foreclosure Interventions
  An Analysis of Neighborhood Characteristics
  Associated with High Foreclosure Rates in Two
  Minnesota Counties Federal Reserve Bank of
  Minneapolis and Manchester College, October 2006.
  
• Haurin, Donald and Stuart Rosenthal (2004), The
  Sustainability of Homeownership Factors
  Affecting the Duration of Homeownership and
  Rental Spells. U.S. Department of Housing and
  Urban Development, Office of Policy Development
  and Research (December).
• Immergluck, Dan and Geoff Smith (2005) The
  External Cost of Single-Family Mortgage
  Foreclosures, Housing Policy Debate, Volume 17,
  Issue 1, 2006.
• Pence, Karen (2003) Foreclosing on Opportunity
  State Laws and Mortgage Credit, Board of
  Governors Federal Reserve System.
• Reid, Carolina Katz (2004), Achieving the
  American Dream? A Longitudinal Analysis of the
  Homeownership Experiences of Low-Income
  Households. University of Washington, Center for
  Studies in Demography and Ecology, Working Paper
  04-04 (April).
• Sharick, Merle Omba, Erin Larson, Nick and
  James D. Croft (2006), Eighth Periodic Mortgage
  Fraud Case Report to Mortgage Bankers
  Association, Mortgage Asset Research Institute,
  Inc., April 2006.

46
Acknowledgements

• I would like to thank the following individuals
  for their support and encouragement
• Dr. James Murdoch
• Dr. Ronald Briggs
• Ph.D. student Stephen Kiser