Spatial Clustering of Illegal Drug Dealers:

1

• Spatial Clustering of Illegal Drug Dealers
• Swarming for Safety or Agglomeration for Profit
• Dr. George F. Rengert
• Department of Criminal Justice
• Temple University
• Philadelphia, PA. 19122
• grengert_at_temple.edu

2

• Where can illegal drug markets locate?
• Folk wisdom any where they want to.
• Scientific knowledge any where they want to as
  long as
• Safe from neighbors and detection by police.
• Profits can be made.
• Safest areas to sell drugs generally thought to
  be in socially disorganized areas.
• If not socially disorganized, may experience
  resistance from neighbors.
• Example from North Philadelphia

3
(No Transcript)
4

• But most socially disorganized areas may be least
  profitable areas.
• Lack local demandabandoned houses.
• Drug dealing can lead to abandoned houses as more
  people sell than buy houses in this community.
• Would you buy one of these houses located in a
  drug sales area of North Philadelphia?

5
(No Transcript)
6
(No Transcript)
7

• The following is an example of housing
  abandonment measured by tax delinquency around a
  drug sales area.

8
(No Transcript)
9
(No Transcript)
10

• Which is it, profit or social disorganization?
• Critical issues concerning profits in retail
  operations location, location, location.
• Good locations allow
• ready access
• attract large numbers of customers
• increase the potential sales of retail outlets.

11

• Retail market analysts commonly use demographic
  variables to predict market share Demographic
  Profile.
• Who is likely to purchase illegal drugs?
• Young adults aged 15 to 29.
• High school drop-outs.
• Unemployed.
• Marketing geographers have developed several
  strategies for determining optimal locations of
  retail firms.
• Location-allocation model most often used.
• Includes the objective function, demand points,
  feasible sites, a distance matrix, and an
  allocation rule.

12

• We use the objective function of maximizing sales
  volume by minimizing distance to potential
  customers identified by the Demographic Profile.
• Data from Wilmington, Delaware.
• Demand points centroids of census tracts.
• Distance matrix distance between centroids of
  
  . census tracts.
  
• Allocation rule customers assigned to the
  census tract that minimizes total distance
  traveled by potential customers for
  illegal drugs.
• Assumption all users in city purchase drugs at
• this census tract
  

13

• Would have to travel the fewest person-miles if
  illegal drug market was located in census tract
  1600.
• This tract was fourth from the highest in
  reality.
• Limitations of the simple form of
  location-allocation model
• Planar model any location in city is a
  potential site.
• Residents of expensive housing areas not likely.
• Masked out areas where medium housing values
  above average.
• Local addicts will travel any distance for drugs.
• Assigned zero to distance if beyond a mile, 1 if
  less.

14

• Distance matrix is replaced with zeros and ones.
• Two clusters of census tracts identified
• 2200, 2300, and 1400.
• 602 and 601.
• Surprise not in the center of the city.

15

• The analysis yielded two clusters of census
  tracts.
• The census tracts that ranked first, second and
  third formed the first cluster.
• The census tracts that ranked fifth and sixth the
  second.
• The preceding map portrayed this analysis.
• It is not census tract 600 or 1600 that are in
  the center of the city.
• Rather it is a group of census tracts that are in
  the center of a population of potential drug
  users.
• Rather large areas.
• We need specific sites for our drug market.
• Requires more refined analysis possible with GIS.

16

• What can Geographic Information Systems do for
  us?
• Compare traditional analysis with what possible
  with GIS
• Traditional analysis assigns features to census
  space.
• Census tracts.
• Block groups.
• Block faces.
• Census boundaries are set and determine the
  spatial nature of the analysis.

17
(No Transcript)
18

• Refined method
• Create GIS buffers about features and allocate
  proportion of area of tract that is within
  buffer.
• Advantages
• Feature does not have to be in tract to impact
  it.
• Impact is proportional to size of tract.
• Disadvantages
• Assumes impact uniformly distributed across
  entire tract.
• Proportion not site specific.

19
(No Transcript)
20

• GIS method
• Create new geographies with buffers around
  features.
• Create interaction effects with overlays of
  buffers.
• Advantages
• Does not assume effect is uniform over census
  tract.
• Buffers can be sized to reflect spatial reach of
  a feature.
• Disadvantages
• New geographies vary markedly in size.
• Small slivers created that lack geographic
  meaning.
• Zero counts overrepresented.

21
(No Transcript)
22

• Drug Market Analysis of Wilmington, Delaware,
  GIS.
• Initially start with census data at block group
  level.
• Local Demand
• 1. Percent of population age 14 to 29.
• 2. Unemployed males.
• 3. Percent of population over age 18 with
• less than a high school education.
• 4. Median Income.
• 5. Number of children under age 5 living
• in poverty.
• R2 .467

23

• Identify features that attract potential drug
  users.
• Routine activities create crime generators.
• Schools, taverns, homeless shelters, etc.
• Create buffers around these features to determine
  their areal reach if any.
• Use location quotients to determine if feature
  associated with spatial aggregation of drug
  dealers.

24

• LQ CR / CN
• CR Number of drug arrests per square mile
  in
• GIS identified area.
• CN Number of drug arrests per square mile
  in
• entire city.

25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
Local Accessibility

• Routine activity nodes.
• Anchor points of daily activities.

34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
(No Transcript)
43
(No Transcript)
44
Homeless Shelters Social Service Centers Check
Cashing Stores Taverns Liquor Stores
Wilmington, DE
45
Homeless Shelters 800 feet Social Service
Centers 800 feet Check Cashing Stores 400
feet Taverns 400
feet Liquor Stores 400 feet
Wilmington, DE
46
The Analysis

• Create buffers around point and line features.
• Assign the buffer areas to census block groups.
• Statistically analyze the importance of each
  variable.
• Begin with drug sales figures and the plotting of
  each feature on a map of Wilmington, Delaware.

47
(No Transcript)
48
Zero Inflated Poisson Model

• Two phase analysis.
• Like analysis of number of children a couple
  chooses to have
• Choice to have children
• Choice of how many children to have once decide
  to have them.

49

• Factors Associated with the Establishment of a
  Drug Market
• Positively Associated
• Percentage of nonwhite residents.
• As the percentage of nonwhite residents
  increases, the chance that the area will never
  have a drug-market arrest decreases.
• The spatial lag term.
• As the number of arrests in the surrounding area
  increases, the chance of the parcel never having
  a drug-sale arrest diminishes.
• Negatively Associated
• I-95 exits.
• Being located near to an access ramp for I-95
  increases the chance that an area will not have
  drug-market arrests.
• Rest not statistically significant

50

• Factors associated with the size of drug markets
  given that a drug market exists
• Positively associated
• I-95 exits.
• Female headed households with children.
• Vacant homes.
• Non-white residents.
• Check-cashing stores.
• Liquor stores.
• Homeless shelters.
• Spatial lag term.
• Negatively associated
• Renter occupied units.
• Social service programs.
• Taverns.
• Rest not statistically significant

51

• Implications of the Study.
• Significant difference between taverns and Liquor
  stores.
• Place managers of tavern owners?
• Negative association between rental housing and
  drug sales arrests.
• Interaction between population density and
  neighborhood control?
• Significance of spatial lag term.
• Is it Agglomeration economies?
• Is it social networks?
• Is it a result of spatial diffusion?

52

• Association between Black population and drug
  sales arrests.
• Is it environmental racism?
• Noxious facilities are put in vulnerable
  neighborhoods
• Is it a lack of social efficacy.
• Do not use all the tools available including the
  police.
• Not police crackdowns.
• Rather, prioritize calls for servicecreate
  social efficacy.

53

• Clearly what is needed at this point is
  contextual analysis to determine interaction
  effects.
• We especially see this in the I-95 access.
• Not all are bad.
• But if is bad, is very bad as size of market
  illustrates.
• We also see this in the difference between
  taverns and liquor stores.
• Notice that the difference between location
  quotients is not great for taverns.
• Indicates they might locate in bad areas rather
  than attracting drug sales.
• Liquor stores have greater difference in LQ.
• In order to obtain contextual variables, can use
  GIS to visualize

54
Census Block Groups
Wilmington, DE
55
Homeless Shelters 800 feet Social Service
Centers 800 feet Check Cashing Stores 400
feet Taverns 400
feet Liquor Stores 400 feet
Wilmington, DE
56
Single Coverage combined polygons
Wilmington, DE
57
Single Coverage combined polygons
Wilmington, DE
58
Census Block Groups and Built Environment
Wilmington, DE
59
Areas within Buffers of Liquor Store or
Tavern and I95 Exits and Major Roads
Wilmington, DE
60
Census Block Groups and Built Environment
Single Coverage
Wilmington, DE
61

• How do you know which interaction effects are
  significant?
• Which should you choose?
• Answer Tree analysis in SPSS-- interaction trees.
• Can force first split.
• Drug Sales
  Arrests
• Low income High income