Population
Growth and Affluence Distribution
Dallas-Fort
Worth: 1970-2000
David Rardon
GISC 6387
Summer 2005
Contents
Step 2: Identification of Concentration
Step 3: Evaluation of Consistency
Background
Over the 30 year period from 1970 to
2000 the
Population growth from 1970 to 2000 was particularly strong
in the
This study focuses on the
Project Approach
The study was conducted by first identifying the areas where population growth and affluence are concentrated within the region and then comparing the consistency between the two. By locating areas of concentration the comparison for consistency will be more clear and specific.
For the purposes of this project, affluence a general term intended to describe how “well-off” the population may be, or conversely, how much the population may be struggling economically. Three factors were considered to represent affluence:
Income
Home
Ownership
Education
Attainment
The comparison between population growth and affluence was conduced for each decade as well as for the three decades combined, producing the following time periods of study:
1970
- 1980
1980
- 1990
1990
– 2000
1970
- 2000
Expected Results
At the outset, the expectation was that the concentration of affluence factors would generally be consistent with the concentration of population growth for each time period studied.
Literature Review
Much has been studied and written about the social and environmental impacts of population growth and urban sprawl (e.g. economic growth follows population). A literature review produced some material on related topics. Some examples include:
Economic
Growth in Cross Section of Cities,
Edward L Glaescr, et al, Journal of Monetary Economics, Aug 95
Population
and Economic Growth, Gary Stanley Becker, American Economic Review,
May 1999
A
Theory of Urban Growth, Duncan Black and
On
the Distributional Aspects of Urban Growth, Christopher H. Wheeler, Journal
of Urban Economics, March 2004
The analysis for this project was executed through the following three steps:
Data
Preparation
Identification
of Population Growth and Affluence Concentration
Evaluation
of Consistency Between Population Growth and Affluence
This task involved the gathering and preparing all of the data necessary for analysis.
Data Required
Census
Tract boundary files for
Census
tracts were used because that was the smallest area for which 1970 data was
available
Attribute
data for each census year:
Population:
Total
Population by Tract - converted to Population Density to account for variation
in Census Tract sizes
Income:
Per
Capita Income
Home
Ownership:
Percent
of Housing Units Owner Occupied
Education
Attainment:
Percent
of Population with Bachelors Degree or Higher (>15 Years of Education)
Data Sources
Boundary
files were obtained using Geolytics CensusCD products available at The
University of Texas at
All
attribute data used was US Census data obtained directly from the Census
Bureau, the North Central Texas Council of Governments, and the CensusCD
products
Major Tasks and Challenges
One
of the greatest challenges of the data preparation task was obtaining the data
for 1970 and 1980. It was not as readily
available as initially expected
The
selection of variables available for 1970 was considerably less than for 2000,
limiting the choices for analysis
Boundary
file differences
The
boundary files obtained for 1970 and 1980 differed from those in 1990 and 2000
1970
and 1980 did not include land beneath area lakes as part of the census tract,
whereas in 1990 and 2000 that land was included
In
order to maintain consistency, the 1970 tract boundaries were edited to include
the land beneath lakes, using 2000 boundaries as a guide
Step 1
Output:
The results of this step are illustrated in the following maps:
Population
Income
Home Ownership
Education
Maps of data by census tract often appear to have a patchwork effect, making it difficult to determine where the overall concentration is specifically located. Therefore the following method was used to locate the concentration of each factor.
Method: Raster
Neighborhood Analysis
Raster
neighborhood analysis involved converting all data to consistent sized raster
cells, then calculating a value for each cell based on the values of all other
cells within a specified neighborhood of the cell (e.g. average population
density for all cells in a 10-cell radius)
Advantages
of this method include:
Neighborhood
analysis produces a smoothing effect that helps highlight the areas of
concentration
The
consistent geographic areas of raster cells reduces concerns about Census Tract
boundaries over time
Raster
layers can be easily compared using ArcMap and the Spatial Analyst extension in
ArcGIS
Major Tasks and Challenges
All
data was rasterized for all time periods
A
raster cell size had to be established
The
size had to be small enough so that the smallest census tracts would be
included in at least one cell, but large enough to prevent an unnecessarily
large volume of data and long processing times
Ultimately
a cell size of 1200 feet was selected
Population
density changes for each decade and the entire period were easily calculated
using the rasterized data
Parameters
for the Neighborhood Analysis calculation had to be determined
The
neighborhoods were defined to be a circle calculating statistic type: mean
Several
different alternatives were tested to determine the size (radius) of the
neighborhood that produces the most useful results
Smaller
neighborhoods produced results that were sill rather spotty and large
neighborhoods smoothed away much of the variation. For an illustration click here
Ultimately,
a neighborhood radius of 22 cells was selected due to the balance of the
results and the even 5-mile value that resulted when multiplied by the 1200
foot cell size
Neighborhood
Size Differences: 10 cells (left) vs. 25
cells (right)
Step 2 Output:
The results of this step are illustrated in the following maps:
Population Growth
1970-1980 Population Growth Concentration
1980-1990 Population Growth Concentration
1990-2000 Population Growth Concentration
1970-2000 Population Growth Concentration
Income
Home Ownership
1970 Home Ownership Concentration
1980 Home Ownership Concentration
1990 Home Ownership Concentration
2000 Home Ownership Concentration
Education
The results of step 2 produced a number of maps that individually produce some potentially interesting results; however the objective of the project was to compare the affluence to population growth. For example:
Are
the areas of highest population growth consistent with the areas of highest
income?
Are
the areas of lowest population growth consistent with the areas of lowest home
ownership?
Visual inspection can be used to identify some general similarities and differences, but it is extremely difficult to determine any specific conclusion, therefore a means to provide a quantified evaluation is necessary.
Method:
Compare Tiers for Consistency
The
approach used for this task involved classifying the values in the neighborhood
analysis output layers into similarly defined tiers
This
was accomplished by ranking from highest (tier value = 6) to lowest (tier value
= 1).
This
method allows for quantitative comparison between two layers by comparing each
cell’s tier value (rank) in one layer with the corresponding tier value in the
other layer to calculate the difference between the two.
Assignment of Tiers
The
data for each factor/year was assigned to tiers based on bands of standard
deviation originating from the mean. For
example in the illustration, those cells greater than 2 Standard Deviations
above the mean were assigned to Tier 6, while those with a value between the
mean and one Standard Deviation below were assigned to Tier 3, and so on
By
using standard deviation, the dispersion of the data is taken into account, applying
a consistent method for assigning tiers
Tiers
based on standard deviation also generally provide smaller sets of data at the
extremes (highest and lowest), further helping to pinpoint the areas of
concentration
Although
the example illustration is of a normal bell curve, all of the data did not
have an exact normal distribution.
Some
dataset were slightly skewed to the to the high end and some to the low end
However,
all cases fit within 5 or 6 bands of Standard Deviation originating from the
mean
Classification of Consistency Between Layers
Once
the data was classified into uniform tiers, the consistency between two layers
could be evaluated by subtracting one layer from the other.
The
result of this calculation indicates how well the tiers are aligned
A
difference of 0 indicates that the values in both layers were in the same tier,
whereas a difference of 2 indicates that the values are two tier levels apart
The
output from this calculation could be used to classify the consistency between
the layers, as illustrated in the following table:
Consistency By Tier
After
the consistency between layers was classified, the analysis was taken a step
further to evaluate exact consistency tier by tier. For example:
How
well do the areas of highest (Tier 6) income align with the areas of highest
population growth?
How
well do areas in Tier 3 home ownership align with areas of Tier 3 population
growth?
Major Tasks and Challenges
The
primary challenge for this step was determine the method to quantify the
results.
Several
methods were considered and tested before the method that was ultimately used
was defined
Step 3 Output:
Consistency Between Layers
Income
1970-1980 Income and Population Growth Consistency
1980-1990 Income and Population Growth Consistency
1990-2000 Income and Population Growth Consistency
1970-2000 Income and Population Growth Consistency
Home Ownership
1970-1980 Home Ownership and Population Growth Consistency
1980-1990 Home Ownership and Population Growth Consistency
1990-2000 Home Ownership and Population Growth Consistency
1970-2000 Home Ownership and Population Growth Consistency
Education
1970-1980 Education and Population Growth Consistency
1980-1990 Education and Population Growth Consistency
1990-2000 Education and Population Growth Consistency
1970-2000 Education and Population Growth Consistency
Consistency by Tier
Income
1970-1980 Income and Population Growth Consistency By Tier
1980-1990 Income and Population Growth Consistency By Tier
1990-2000 Income and Population Growth Consistency By Tier
1970-2000 Income and Population Growth Consistency By Tier
Home Ownership
1970-1980 Home Ownership and Population Growth Consistency By Tier
1980-1990 Home Ownership and Population Growth Consistency By Tier
1990-2000 Home Ownership and Population Growth Consistency By Tier
1970-2000 Home Ownership and Population Growth Consistency By Tier
Education
1970-1980 Education and Population Growth Consistency By Tier
1980-1990 Education and Population Growth Consistency By Tier
1990-2000 Education and Population Growth Consistency By Tier
1970-2000 Education and Population Growth Consistency By Tier
Consistency Between Population Growth and Affluence
The results of consistency evaluation are highlighted in the following table
Consistency Between Population Growth and Affluence By
Tier:
The results for consistency by tier are highlighted in the following table
Observations
With
the exception of 1970-1980, the Income-Population Growth comparison is highest
in the Consistent range followed by the Somewhat Consistent range. The non-consistent ranges generally make up
less than 15% of the total
The
1970-1980 comparison still falls in the highest two categories, but is not as
strong as the other time periods
The
Education-Population Growth comparison produced the highest level of
consistency
As
in the Income comparison, the 1970-1980 comparison was slightly weaker than the
other time periods
The
non-consistent ranges were only around 5% of the total
The
similarity of the results for Education
and Income are not surprising given that a there is likely a
relationship between the two
The
Home Ownership-Population Growth comparison yielded results opposite to the
other two factors
The
percentage of Consistent results was less than 15% and the non-consistent
categories were greater than 50%
One
consideration may be that high population growth, particularly population
density, is partially driven by high density multi-family units, resulting in
lower home ownership rates
It
is possible that home ownership rates are not an accurate gauge of affluence
and therefore were not a good variable for this study.
The
results of the Consistency by Tier analysis further support the noted
observations
With
the exception noted for the 1970-1980 time period, the results were similar for
all time periods, whether individual decade or the entire period
In
conclusion, when income and education are used to represent affluence, the
results of this project indicate a good level of consistency between the
concentration of affluence and population growth.
However,
when considering home ownership as a measure of affluence, the results indicate
that there is little consistency with population growth.
Implications
A
practical use for this data could be for anyone seeking affluent people. For example, a high end retailer may want to
seek areas of high population growth in order to be near their prospective
customers
Additional
topics of study could be derived from these results. For example:
If
affluence and population growth are consistently aligned, what will become of
mature areas that experience little population growth or decline?
Do
similar results hold true in other regions?
What
differences may be observed in fast growing regions vs. slow growing regions?
Lessons Learned
Several lessons were learned in the execution of this project
Allow
plenty of time for data preparation
Actual
data collection and preparation took more time that initially expected,
particularly for the older data from 1970 and 1980
Try
various alternatives when selecting parameters to determine what produces the
most useful results
Several
cell sizes were considered for the rasterization of data