Monday, September 20, 2004

Professor Sandra Arlinghaus

UP 507 Advanced Geographic Information Systems (Winter, 2001)

Introduction

This course project is divided into two sub-projects, each of which constitutes research on the evaluation of empowerment zone (EZ) policy.   The research evaluates program impacts of six urban empowerment zones on employment and business establishments. Two methodological issues will be considered.  The first issue is to designate the control area for each EZ; once established, the control area will be compared with the EZ in terms of program impacts.  The control area should have socioeconomic characteristics similar to each EZ.   To ensure that it does have appropriate characteristics, the control area will be developed in geographical space using GIS and factor analysis.  The second project involves the acquisition of annual employment and business establishment data, for the control areas and for the EZs which do not have their own statistical data.  For simplicity, the Detroit Empowerment Zone (DEZ) is employed as a case study.

Project I : Designating control areas in geographical space

Research Question

What is a good procedure for measuring similarity and dissimilarity of socioeconomic characteristics between geographical units such as census tracts?
How can GIS contribute to managing the values representing socioeconomic characteristics in space as a tool to find control areas?

Research Scenario

Select variables which represent socioeconomic characteristics for each census tract in the Detroit Metropolitan Area.  I considered 25 variables.
Execute factor analysis to get factor scores for each census tract.  From the factor analysis, reduce the number of variables to several factors, each of which represents some socioeconomic characteristics such as income or poverty.  Compute factor scores for each census tract.  I extracted three types of factor scores for each census tract.

Draw a thematic map of the Detroit Metropolitan Area with factor scores by census tract. 

Designate a control area that falls into the same class as the DEZ. 

Pursue a statistical hypothesis test to verify whether or not the experimental area (DEZ) and the control area are the same in terms of the socioeconomic characteristics that are represented by factor scores.Iterate the process from the thematic map stage until getting a suitable control area. 

Method (using Detroit Empowerment Zone)

				Hypothesis Test
Step I	Step II	Step III	Step IV	T-test	Overview

References

Metadata for the maps

Control Areas of six urban EZs

Project II : Areal Interpolation

Research Question

Suppose two areal units with different boundaries in geographical space; one is the area that does not have necessary data, and the other is the area that has the data.  Let us call the former the target zone and the latter the source zone.
In the case that the source zone area contains the target zone area, can we get the data for the target zone from the source zone using GIS?

Research Scenario

We need annual employment and business establishment data by census tract of Detroit Empowerment Zone (DEZ).

DEZ is made up of 48 census tracts, but the tracts do not have annual employment and establishment data.

The zip code business patterns has annual employment and establishment data by zip code.

The areal interpolation method using GIS can provide us with these statistical data by overlaying the DEZ map on the zip code boundary map with the data.

Method

Location of DEZ (Map 1)
DEZ is made up of 48 census tracts, which do not have annual employment and establishment data (Map 2).  Let us call DEZ the target zone which means that it does not have needed data.

The zip code map containing DEZ with annual employment and establishment data by zip code is made up of 17 zip code areas (Map 3).  Let us call it the source zone which means that it has needed data.

Superimpose the target zone map (DEZ map) on the source zone map (Zip code map) (Map 3).
Split the source zone layer by the census tract lines of the target zone (Map 4).  110 polygons with the annual employment and establishment data were made (Map 5).  Note that there is no unique key field in the attribute table.  Without a unique key field, we cannot perform a spatial join to create a new attribute table by linking attribute tables of multiple layers.  For the spatial join, each attribute table of the layers should have a common unique key field.

Superimpose the target zone map (DEZ map) on the source zone map split by DEZ census tract lines with 110 polygons (Map 6).
Execute union of the polygons into DEZ census tract.  Then, we can get a map with 48 polygons which are exactly consistent with DEZ (Map 7).  Note that the attribute table has annual employment and establishment data, but still does not have a unique key field.

What we need is a data table with both annual employment/establishment and socioeconomic data by census tract.
First, link the table with socioeconomic data by census tract (Table 1) to the DEZ map by census tract (Map 2).  Then, we can get the DEZ map with socioeconomic data by census tract (Map 8)

Second, we want to perform a spatial join of Map 8 to Map 7 to create a new census tract layer with both socioeconomic data in Map 8 and employment data in Map 7.   However, the maps do not have a common key field to join with.  For this, create a key field named ID in the attribute table of  Map 7 (Map 9).  Seek POLID, which means Polygon ID, of each polygon in Map 8 to make a data table (Table 2) which matches ID's in Map 9 with POLID's in Map 8 (Map 10).  Join Table 2 to the attribute table of Map 7 with ID field as common key field to create the map which has the employment and establishment data table with POLID field as unique key field (Map 11).

Finally,  join Map 11 to Map 8 with POLID field.  Then, we can get the DEZ map (Map 12) with its attribute table containing both employment/establishment and socioeconomic data by census tract (Table 3).

Assumption

Areal interpolation is based on the assumption that employment and business establishment are evenly distributed throughout every zip code area.  This assumption, resulting in probable inaccuracy in the actual number of employment/establishments, functions as a limitation in collecting data using the areal interpolation method when we use the actual numbers for a study.
However, this assumption can be tolerable if the data are used for comparing changes over time in an area since we can assume that the annual % change of employment/establishment in a same area over time is not different.

References

Metadata for maps