Final Project: Demographic Analysis of Property Foreclosures in Dallas County Russell Frith GISC 638

1
Final Project Demographic Analysis of Property
Foreclosures in Dallas CountyRussell FrithGISC
6384
2
Problem Statement

• For this report, property foreclosures are
  defined as a private residence where the owner
  has been legally served with an eviction for
  delinquency on mortgage payments.
• The purpose of this study is to analyze potential
  demographic trends in regions of Dallas County
  where there exist high numbers of residential
  property foreclosures.
• A companion web site for this report may be found
  at http//www.utd.edu/rfrith/fp/fp.htm

3
Topics

• This paper is comprised of six main topics.
• In the first topic, an exploratory statistical
  data analysis (ESDA) is performed on the
  foreclosure data in order to determine if the
  geographic distribution of the foreclosure point
  data are clustered or randomly distributed among
  the census tracts in Dallas County.
• In the second topic of this report, an analysis
  is conducted of the data. The principal topics
  presented in this section concern data sources
  and data definitions used as database schema for
  near-relational census tables.
• The third section of this project treats the
  architecture and implementation of an analysis
  engine used to draw conclusions about foreclosure
  distributions.
• The fourth section of this paper presents a
  series of observations obtained from overlaying
  foreclosure point data onto various thematic maps
  of the author's choosing.
• The fifth section presents results from zonal
  statistics tables.
• The sixth and final section presents a summary
  for this paper and suggests several extensions
  for this study.

4
Justification for Posing the Question

• As stated in the introduction, a justification
  must be made to conduct the study of demographic
  trends in foreclosure analysis. Three main data
  explorations were done and all essentially showed
  that foreclosures are clustered into distinct
  regions. The three methods of data exploration
  were
• Getis Statistic for zonal indifference,
• Spatial Autocorrelation for zonal indifference
  (Moran's I), and
• Point Density analysis.
• The results are shown in the following slide
  gallery. In addition, the point count data per
  census tract appear to be normally distributed,
  as opposed to randomly distributed.

5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
Data and Its Sources

• Foreclosure Data
• UNIX awk Script Used to Strip HTML Tags from
  Records
• Foreclosure Data Source
• Address Table
• Geocoded Data (Street Map w/ Points)
• Census Tract Data with Foreclosure Counts Per
  Tract (PntsInPoly field).
• Demographic Data
• NCTCOG
• Housing Code Table
• Economic Code Table
• Education Code Table
• Race Code Table
• Age Code Table
• Projection Data
• Map Projections

10
Analysis

• Mask Construction for Areal Interpolation
• Geocode - approximately 85 of the foreclosed
  addresses were matched.
• Point Density Map Generated
• Regionalize Point Density
• Reclass Region Map
• Conditional Processing of Region Map
• Region Mask Generated

11
(No Transcript)
12
(No Transcript)
13
Equations

• These equations were derived from the attribute
  definitions from the census tables.
• Blacks P007003Race Code Table
• Minority P007002 P007003 P007004 P007005
  P007006 P007007 P007008Race Code Table
• Whites P007001Race Code Table
• Married w/ Children P018008House Code Table
• Single Female w/ Children P018015House Code
  Table
• Over 65 P020017House Code Table
• Unemployed P038012 P038017 P038021Economic
  Code Table
• Educational Attainment
• High School Only Household Head P037011
  P037028Education Code Table
• Bachelor's Only Household Head P037032
  P037017Education Code Table
• Greater than 1 hr commute time P031013
  P031014Economic Code Table

14
Automated Model

• Model Diagram
• Model Paramters
• Input Census TableThis is an arbitrary census
  table downloaded from NCTCOG.
• ExpressionThis is the equation used to calculate
  attribute density.table_name.calculation_field
  (var1 var2 ... varn)/cell_count
• Census Table Join FieldThis is typically done by
  tract number. An example would be
  table_name.TRACT2000
• Calculation FieldThis is the expressiontable_na
  me.calculation_fieldIt holds the results of the
  calculation.
• Rasterization FieldThis will be
  table_name.calculation_field
• Model Source

15
(No Transcript)
16
Model Parameters
17
Observations

• The following slides are available
• A theme showing an extrusion of census tracts
  based on the numbers of foreclosures with the
  tract.The theme looks from the Southwest to the
  Northeast.
• A theme map showing the distribution of
  foreclosures by city.
• A theme map showing the distribution of
  foreclosures by census tracts.
• A theme map showing the distribution of
  foreclosures by school districts.
• The following "thematic implications" can be
  inferred
• Observe how the DART rail connects foreclosure
  hot spot regions.
• Observe the proximities of foreclosure regions to
  major freeways.

18
(No Transcript)
19
Foreclosures by City
20
Foreclosures by Census Tracts
21
Foreclosures by ISD
22
Results
23
Questions Answered

• What percentage of minorities reside within
  foreclosed regions? 20
• What percentage of the population residing within
  foreclosed regions are minority? 34
• What percentage of blacks reside within
  foreclosed regions? 26
• What percentage of whites reside withing
  foreclosed regions? 16
• What percentage of married couples with children
  reside within foreclosed regions? 14.8
• What percentage of single female household heads
  with children reside within foreclosed regions?
  27
• What percentage of household heads over 65 reside
  in foreclosed regions? 12.8
• What percentage of unemployed reside in
  foreclosed regions? 19.6
• What percentage of household heads with only a
  high school education reside in foreclosed
  regions? 17.9
• What percentage of household heads with only a
  B.S. degree reside in foreclosed regions? 13.9
• What percentage of household heads who require
  more than 1 hour of commute time to work reside
  within the regions? 19.6

24
Summary Extensions

• Problem with Model Builder The author encountered
  disappointment with Model Builder. When a raster
  request was programmed, Model Builder would
  inexplicably "change" the raster function's
  parameters. The dire consquence was that the
  zonal statistics would be wrong.
• Problem with Equations Admittedly, the author did
  not put too much research into formulating some
  of the demographic variables.
• Problem with Regional Classification A rigorous
  statistical "proof" for this mask was not
  conducted, although its construction is not
  completely arbitrary. More in-depth analysis
  needs to be done on the mask's limitations and
  biasness in order to make more rigorous
  assertions.
• Problem with Foreclosure Table The table is not
  complete. Competing foreclosure providers post
  conflicting data sets. Furthermore, not all of
  the data were sufficiently goecoded.
• What does it say about "fairness" of
  foreclosures?
• Are there discriminatory lending practices?
• Are lenders overlooking a borrower's earning
  potential?
• Are single women with children more susceptible
  to being foreclosed on?
• Are foreclosure regions contained within areas of
  high unemployment?
• Are school and city taxes contributing to
  foreclosures?
• Are foreclosures being induced by land grab
  conspiracies?
• Extensions
• Relate to rail corridors.Perform a GWR between
  foreclosure regions and the DART rail corrider.
• Relate to toxic sites. Perform a GWR between
  foreclosure regions and toxic waste sites.
• Relate to freeways. Perform a GWR between
  foreclosure regions and major thoroughfares.
• Land Use and Land Use ChangeAnalyze how land use
  change detection impacts foreclosure counts.