Using Administrative Data to Create a Longitudinal, Spatial Data Archive to Support Policy Analysis,

1
Using Administrative Data to Create a
Longitudinal, Spatial Data Archive to Support
Policy Analysis, Planning and Research

• Dennis Culhane
• University of Pennsylvania

2
The Philadelphia - U of Penn Approach
Two primary types of data

• 1) Property and Housing Records
• 2) Persons' and Families' Health and Social
  Service Information

3
Two Related Types of Applications

• 1) Property-specific analysis (Neighborhood
  Information System Parcelbase)
• 2) Block/Neighborhood-specific analyses
  (Services Utilization Monitoring System SUMS
  Social Indicators Analyses)

4
The Philadelphia- U of Penn Partnership Model
City Responsibilities

• City agencies provide data and in-kind services
  of data processing staff
• City agencies provide internal political support
  for interagency data requests
• City agencies identify critical policy questions
  to address

5
The Philadelphia- U of Penn Partnership Model
Penn Responsibilities

• U Penn archives the data (data warehousing)
•  
• U Penn coordinates data exchange agreements
•  
• U Penn designs GIS applications for end-users
•  
• U Penn hosts and maintains websites for
  applications
• U Penn conducts basic research and policy
  analysis (supports nonproject researchers as well)

6
Data Security/Access Issues

• City official arbiter of data requests and
  authorizations
• In general, property-specific information is
  public domain (though access to application may
  not be)
•  
• In general, person-specific information is
  confidential (though aggregate applications may
  be public)
•  
• U Penn created the Aggregator to facilitate
  human services data transfer (block, block group
  or tract options)

7
Other Issues

• Scheduled, periodic updates of data essential
• Consistent data quality audits needed
• Data warehousing is to the mutual benefit of
  researchers and city government
• Trade-off City shares data, University designs
  software for City to access the data in
  customized manner

8
Research Advantages

• 1) Produces discrete, continuous-time data, ideal
  for time series and event history analyses
• 2) Produces user-defined small area geographies,
  ideal for studying the "natural" clustering of
  phenomena, increasing statistical power, and
  creating more sensitive space-dependent models
•  
• 3) Supports and enhances spatial analytic
  statistical approaches econometrics, social
  ecology, epidemiology, multi-level modeling

9
Research Advantages

• 4) Creates new variable opportunities
  clustering-contiguity measures, distance, travel
  time, transit access, social boundaries/buffers,
  displacement effects, controls for spatial
  autocorrelation
• 5) Supports traditional hypothesis testing models
  as well as emergent exploratory models, e.g. data
  mining techniques.