University of Southern California Dissertations and Theses

Population disaggregation for trade area delineation in retail real estate site analysis

(USC Thesis Other)

Population disaggregation for trade area delineation in retail real estate site analysis

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content POPULATION DISAGGREGATION FOR TRADE AREA DELINEATION IN
RETAIL REAL ESTATE SITE ANALYSIS

by

Alfredo David Cisneros

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2015

Copyright 2015 Alfredo David Cisneros

DEDICATION
I dedicate this document to my parents for their constant support, my family and in laws, and
most importantly to my wife who spurred me to get it done and agreed to make me the happiest
man in the world.

ii

ACKNOWLEDGMENTS
I will be forever grateful to all my professors in the GIST program, and most of all Professor
Karen Kemp. Thank you to my committee members Victor Bennett and Katsuhiko Oda. Thank
you also to my family and friends, all of whom displayed patience, stuck with me, and all
inspired me throughout this process and helped me complete this degree program.

iii

TABLE OF CONTENTS
DEDICATION ................................................................................................................................ ii
ACKNOWLEDGMENTS ............................................................................................................. iii
LIST OF TABLES ......................................................................................................................... vi
LIST OF FIGURES ...................................................................................................................... vii
LIST OF ABBREVIATIONS ...................................................................................................... viii
ABSTRACT ................................................................................................................................... ix
CHAPTER 1: INTRODUCTION ................................................................................................... 1
1.1 Summary of Methodology .................................................................................................. 2
1.2. Research Goals................................................................................................................... 3
1.3. Structure of Thesis ............................................................................................................. 4
CHAPTER 2: BACKGROUND AND LITERATURE REVIEW ................................................. 5
2.1. Methods for predicting sales potential ............................................................................... 7
2.1.1. Analogue Method 7
2.1.2. Regression Method 8
2.1.3. Trade Area Method 8
2.1.4. Esri’s Trade Area Definition Tools 10
2.2 Determining catchment population characteristics ........................................................... 12
2.2.1 Census geography 12
2.2.2. Issues related to the use of aggregated data 13
2.2.3. Collapsing aggregated data to polygon centroids for spatial overlay 13
2.2.4. Methods of areal interpolation 15
2.2.5 Land use as auxiliary data to disaggregate population within census zones 15
2.3 Background Summary ...................................................................................................... 17
CHAPTER 3: DATA AND METHODS ...................................................................................... 18
3.1 Geographic Data Sources .................................................................................................. 19
3.1.1 Assessor Parcel Data 19
3.1.2. Retail Analysis Data Demands 20
3.1.3. Land Use Data 21
3.1.4 Population and other Demographic Data 21
3.1.5 Developable Land Data 22
3.2 Case Study Store Sites ...................................................................................................... 22
3.3 Determining Catchment Population .................................................................................. 24
3.3.1 Disaggregation of Census Population Aggregates 25
iv

3.4 Calculation of Detailed Drive Time Trade Areas ............................................................. 29
3.5 Calculation of Radial and Network Distance Trade Areas ............................................... 29
3.6 Calculation of Trade Area Population and Characteristics ............................................... 30
CHAPTER 4: RESULTS .............................................................................................................. 31
4.1 Results of Drive Time Trade Areas .................................................................................. 33
4.2 Comparison of Trade Areas Created by Distance Measures to Drive Time Trade Area . 46
4.3 Summary of Results .......................................................................................................... 52
CHAPTER 5: DISCUSSION AND CONCLUSIONS ................................................................. 54
5.1 Directions for further research .......................................................................................... 57
5.1.1 Weighting Population 57
5.1.2 Different Trade Area Techniques 58
5.1.3 Other possible improvements for the future 59
5.2 Conclusions ....................................................................................................................... 60
BIBLIOGRAPHY ......................................................................................................................... 61

v

LIST OF TABLES
Table 1 Esri's Trade Area Tools and Descriptions ....................................................................... 11
Table 2 Population Household and associated demographic information for detailed drive time
trade areas of store sites at drive times of three five and seven minutes. ............................... 44
Table 3 Population Household and associated demographic information for detailed drive time
trade areas of store sites at drive times of five, ten, and fifteen minutes ................................ 46
Table 4 Population Household and associated demographic information Alpine, San Diego,
Poway, and Ramona sites for all trade areas of distance 3 ..................................................... 48
Table 5 Population Household and associated demographic information Alpine, San Diego,
Poway, and Ramona sites for all trade areas of distance 5 ..................................................... 50
Table 6 Population Household and associated demographic information Alpine, San Diego,
Poway, and Ramona sites for all trade areas of distance 7 ..................................................... 52

vi

LIST OF FIGURES
Figure 1. Project Workflow .......................................................................................................... 19
Figure 2: Store locations in suburban and rural areas of San Diego County ................................ 23
Figure 3: “Developable Land” (outlined in red) overlain on “Residential parcels”. . .................. 28
Figure 4 : “Developed residential parcels” in San Diego County. ............................................... 31
Figure 5 : Polygons with existing residential land use development aka “Populated parcels”
overlaid on yellowish background areas representing Census Blocks with population. ........ 32
Figure 6 : “Populated parcels” showing residences. A detail of “Populated parcels” transparency
overlaid on residential land use polygons is also shown. ....................................................... 33
Figure 7 : Trade Areas using detailed drive times for three, five, and seven minute intervals. San
Diego store .............................................................................................................................. 34
Figure 8: Trade Areas using detailed drive times for three, five, and seven minute intervals.
Poway store; ............................................................................................................................ 35
Figure 9 Trade Areas using detailed drive times for three, five, and seven minute intervals.
Ramona store .......................................................................................................................... 36
Figure 10: Trade Areas using detailed drive times for three, five, and seven minute intervals.
Alpine store ............................................................................................................................. 37
Figure 11 “Populated parcels” intersecting detailed drive time areas of 5, 10, and 15 minutes for
the San Diego, Ramona, and Alpine store sites. ..................................................................... 38
Figure 12 : All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Alpine location. ................................................................................... 40
Figure 13: All “Population parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Ramona location. ................................................................................. 41
Figure 14 : All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Alpine location. ................................................................................... 42
Figure 15: All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Poway location .................................................................................... 43
Figure 16: “Populated parcels” polygons for detailed drive time areas of 3,5, and 7 minutes ..... 45
Figure 17: 7 minute Drive Time Trade Areas and 7 mile network & radial distance trade areas 47

vii

LIST OF ABBREVIATIONS
AAG Association of American Geographers
AVGHHINC Average Household Income
AVGNW Average Net Worth
GIST Geographic Information Science and Technology
PCI Per Capita Income
POP Population
PP Populated Parcels
SSI Spatial Sciences Institute
USC University of Southern California

viii

ABSTRACT
An appropriately sited retail location can turn a business into a veritable cash machine for the
owner. Siting a store location has financial implications for store owners, banks, real estate
professionals, store employees and company shareholders, all of whom are impacted by the
success or failure of a store. Determining catchment population -- the population within a store
site’s actual or potential trade area -- is essential for good retail site suitability analysis. An
accurate calculation of a store’s catchment population depends on the method of defining a
store’s trade area and the accuracy and the precision of population data.
This study explored how concentrating aggregated census population into existing
developed residential areas affects the results of trade area analyses likely to be used in retail real
estate marketing and decision making. Different methods of defining trade areas were also used
to explore how the differing trade area outcomes affected results of analyses used for retail real
estate decision making. It also seeks to show how different store sites with different population
densities ranging from very dense areas in suburban areas to areas bordering rural areas affect
population aggregations.
Results of these analyses showed only small changes in catchment population and
demographics when concentrated population areas were used in calculations as opposed to
census aggregates. Conversely using different distance measures for trade area creation resulted
in large differences in catchment population which should be taken into consideration for
analysis and marketing moving forward.

ix

CHAPTER 1: INTRODUCTION
An appropriately sited retail location can turn a business into a veritable cash machine for the
owner. Conversely, a poorly chosen site location may result in owners losing all investment in
their retail location. For first time business owners this could bankrupt their business and cost
them most if not all of their wealth. Siting a store location has financial implications for store
owners, banks, real estate professionals, store employees and company shareholders, all of whom
are impacted by the success or failure of a store.
When searching for an ideal retail site for a client, most retail real estate brokers can
readily list a number of desirable site characteristics that can be used to assess a site’s suitability.
These might include requirements such as:
• Busy retail area
• Signalized intersection
• Property located on a hard corner
• Average household income within three miles above a given threshold
• Population within three miles above a given threshold
• Daily traffic count on the main road above a specific number of vehicles
All of these characteristics are considered favorable in insuring there is sufficient exposure and
market potential for the retail location.
Determining catchment population -- the population within a store site’s actual or
potential trade area -- is essential for good retail site suitability analysis. Population and
associated demographic data are collected by the Census Bureau and to protect individual
privacy are reported at various levels of aggregation (i.e. block, block group and census tract).
Due to the fact that trade areas and aggregated census units are rarely identical and population is
1

often distributed unevenly within census aggregates, using census data to accurately determine
population within a retailer’s trade area is problematic.
This research addresses these shortcomings by using land classification and
supplementary data from San Diego County to disaggregate biennial census and American
Community Survey data in new ways. These disaggregated data were then used to conduct trade
area analyses. Finding better methods to disaggregate data may result in more accurate estimates
of population distribution and allow for better calculation of catchment population and associated
population characteristic statistics.
1.1 Summary of Methodology
This process used three sets of polygon layers in the study area along with census block
attributes to disaggregate the census block data. First the areas that were strictly residential were
extracted from the land use classification layer. Overlaying this with the parcel layer allowed for
the extraction of a “Residential parcels” layer separating residential land use areas into smaller
parcel divisions. A third polygon layer of developable land used was used to remove
undevelopable areas as designated by the county in order to concentrate the population into
“Developed residential parcels”. Finally, Census blocks without population were removed to
further concentrate population leaving only populated developed residential parcels, the
population data source used in the analysis referred to as “Populated parcels”.
These new data were used to determine a site’s catchment population for detailed drive
time trade areas to see if there were differences in the calculated population and related
population characteristics than when calculated by the more commonly used manner which
employs the census data polygons. Four store locations for suburban and rural cities of San
Diego County were chosen to see the effects of the resulting variations in population density on
2

these analyses. Drive Time trade areas for each site for a store of non-specialty goods were
calculated at three, five, and seven minute thresholds. Five, ten, and fifteen minute trade area
thresholds were calculated for specialty goods store sites. A manual calculation of the same
spatial analysis process was conducted to confirm the calculations of resulting catchment
population and characteristics were accurate.
Distance measures in miles are more commonly used to create trade areas in real estate
marketing and decision making due to fewer data requirements needed to create these trade
areas. Trade areas were also created using 3, 5, and 7 mile road network and radial distance
measures. Catchment population and demographics were calculated for each of these trade areas
using census aggregate and “Populated parcels.” These results were compared with trade areas
calculated with drive time minutes created previously to see how these differences could impact
marketing and decision making.
1.2. Research Goals
This case study explores how concentrating aggregated census population into existing
“Populated Parcels” affects the results of trade area analyses likely to be used in retail real estate
decision making. Different methods of defining trade areas were also used to explore how the
differing trade area outcomes affected these analyses as well as similar analyses used in retail
real estate decision making. It also shows how different store sites with different population
densities ranging from very dense areas in suburban areas to areas bordering rural areas affect
population aggregations.
3

1.3. Structure of Thesis
The next chapter reviews related research that has been undertaken previously on determining
site suitability for retail sites, methods for disaggregating census population data to determine
sales potential, the definition of trade areas and other fundamental themes in this thesis. Details
of the methodology outlined above are discussed in Chapter 3. Results of the analyses are
discussed in Chapter 4. Conclusions from this study and recommendations for future
improvements and studies are discussed in Chapter 5.
4

CHAPTER 2: BACKGROUND AND LITERATURE REVIEW
Retailers provide products to the public for consumption. In Intelligent GIS, Birkin et al. (1996)
review how the success of a retailer is dependent on how well they execute on the retailer
marketing mix. The retailer’s marketing mix is made up of Product, Price, Promotion, and Place
commonly referred to as the 4 Ps of marketing. Product refers to the good or service available for
consumers. Demand for products is determined by consumer’s level of need and product
desirability. If there is demand for a product then the consumer considers the second P price, at
which the product is available. Price has a negative impact on demand meaning the higher the
cost of a product the less likely a consumer is to purchase the product at all or from that
particular retailer.
Birkin et al. (1996) explain that if a desirable product is available at a price consumers
are willing to pay the retailer relies on the third P, promotion, to inform consumers the product is
available and appeal to consumers to purchase from the retailer’s location. Promotion can cause
a consumer to visit a particular retailer when products or pricing are similar. Promotion is
especially important in today’s market with many online retailers offering similar or identical
products at lower prices. Competing with other retailers and online retailers is difficult but the
immediacy of a purchase from a brick and mortar retailer can make a large difference in a
consumer’s decision. If a retailer is located near consumers and can be reached with minimal
difficulty, then consumers are more likely to patronize this location.
Place, the fourth P of the marketing mix, refers to the retailer’s location. Distance to a
retailer’s location has a negative relationship with sales. Hence, the greater the distance the less
likely a person will be to purchase a product from that location. Other factors can also influence
a consumer’s decision to purchase from a store location such as nearby attractions (including
5

stores) as well as other products offered by a retailer at their store. Location also plays a large
role in product pricing as real estate costs for all stores are factored into the price of products
offered at a location (Birkin et al. 1996).
As location is integral in determining site suitability and pricing, it is crucial to find the
right site at the right price for a retail location. Real estate professionals should understand the
importance of finding the best site to the business owner’s ultimate success. From my experience
working with McDonald’s corporation to locate sites for their new locations, larger corporations
understand the importance of location to their businesses’ success and have developed their own
departments for finding ideal new sites.
The use of site suitability analysis in retail market analysis has become much more
sophisticated and even the smaller companies now use GIS with demographic or traffic count
data to make models of varied sophistication to estimate sales potential (Birkin et al. 1996). Sales
potential for a retail location is the estimate of retail sales for a specific time frame, usually a one
year period for existing stores or multiple years for new stores to allow them to reach
profitability. Accurate estimation of sales potential is crucial to the success of a retailer and
better techniques for calculating a site’s catchment population are crucial to improving these
estimates.
A store is dependent on the local population likely to be consumers at that store site.
Catchment population is the population within a store’s trade area which is more likely to
patronize the store location. To determine a store site’s suitability for future success, knowing the
catchment population’s counts as well as demographic and consumer spending information is
very important (Birkin et al. 1996).
6

The accuracy of a calculation of a store’s catchment population is dependent on the
method of defining a store’s trade area and the accuracy and precision of population data. As
such, this chapter reviews past and current methods of predicting sales potential for retail sites
and the trade area method used in this case study. In particular Esri’s Business Analyst trade area
tools are discussed including the various trade area creation methods used in this study. This is
followed by a review of Census data and issues arising from its use in such analysis. Lastly, a
review of areal interpolation methods utilized in the past, and specifically dasymetric mapping
techniques similar to those used in this case study are discussed.
2.1. Methods for predicting sales potential
Fenker and Zoota (2000) and Birkin et al. (1996) review intuitive models of predicting sales
which can be applied to modeling and GIS approaches for predicting sales potential. Sales
potential estimation in the past has followed approaches referred to as analogue, regression, and
trade area methods (Fenker and Zoota 2001; Birkin et al. 1996). A summary of these approaches
are reviewed in subsequent sections.
2.1.1. Analogue Method
Use of the analogue method involves seasoned real estate professionals and business owners
making decisions using the confluence of their experience. They apply known past store location
successes and failures to prospective sites determining suitability and potential future success
from past sites with analogous characteristics and their “gut feelings”. Sites deemed similar to
past successful sites are treated as favorable whereas sites similar to past failures are treated
unfavorably. This method is subjective because it relies on the judgment of people and a “gut
feeling” based on their experience from site visits. Given their subjective nature, these judgments
7

may be clouded by emotion surrounding a past experience and are less objective than using a
method involving an unbiased model or method (Fenker and Zoota 2001).
2.1.2. Regression Method
Regression models can be used to assess a site’s suitability for a retail location and have been
used often by real estate professionals (Fenker and Zoota 2001; Birkin and others 1996). The
regression method seeks to calculate a score for overall site attractiveness, the dependent
variable, from independent variables which are believed to be positively and negatively
correlated to sales potential for retail sites (Mitchell 2009). Independent variables used often
include whether a property is on a corner, road visibility, the attractiveness to consumers of
neighboring stores, competitor site proximity, road access, whether the retail location has a left
turn signal, traffic counts, demographic profiles and more. Each independent variable in the
regression equation is assigned weights of positive or negative coefficients reflecting the impact
of each parameter on the site’s likelihood for success or failure (i.e. sales potential). After
applying the regression analysis, retail sites with higher scores are deemed to have the greatest
likelihood of success.
The equations can be determined either mathematically using various regression
techniques or intuitively by assigning variable weights subjectively (Fenker and Zoota 2001).
Such models are applied with varied sophistication based on the mathematical and technological
abilities of the particular buyer or real estate professional (Fenker and Zoota 2001).
2.1.3. Trade Area Method
Another method used to assess sales potential for retail sites analyzes the characteristics of the
population within site’s trade area. A trade area is the area surrounding a store’s location from
8

which patrons are likely to travel to the store; in other words, trade areas enclose the catchment
population. Trade areas have been determined using many different methods.
Many of the trade area techniques use distance measures to characterize costs of travel to
a store. As distance increases, so do the financial and intrinsic costs involved in patronizing a
store location. Difficulty in getting to a store was historically approximated by a distance
measure using a straight line radial distance drawn as a circular buffer (Birkin et al. 1996). Use
of straight line radii for distance measures generally produced poor estimates of a trade area
because this method fails to address the travel realities of road paths and natural obstacles, such
as mountains or lakes, which require travel around or over these objects adding time and distance
cost (Miller 2010).
As digital road data became more available, road network distance measures became
preferred to straight line radial distances. Network distance measures more accurately reflect
distance traveled to reach a location. While road networks better estimate distance they do not
provide the best approximation of difficulty getting to a location (Birkin et al. 1996). Traveling a
few miles may sound difficult to most people but anyone stuck in LA traffic who was running
late for work with a few miles to go understands how frustrating and stressful traveling short
distances might be.
Today, calculation of drive times using speed limit data in conjunction with road network
data provide better approximations of the cost of reaching a retail site location. Time intervals
are used to characterize the relative travel difficulty endured in order to reach a store. Typical
examples of time intervals used in analysis are 3, 5, and 7 minutes for highly substitutable goods
within urban and suburban areas. Willingness to travel is greater for specialty goods which are
not readily available in most general or grocery stores so larger time intervals of 5, 10 and 15
9

minutes are typically used for analysis of these items. Trade areas created with drive time
distance measures produce trade areas with greater distances from the site along freeways and
major roads and shorter distances from a site along side streets and streets with slow speed limits
(Birkin et al. 1996; Miller 2010).
Drawbacks of this method are that speed limits are not always abided by in light traffic
and may not be achievable in heavy traffic. Additionally, traffic fluctuates greatly depending on
the day of the week, time of day and with local school schedules. Further investigations into
these variances are warranted but are outside the scope of this case study (Miller 2010).
For the purposes of this investigation, detailed drive times were found using Esri’s
Business Analyst to delineate trade areas for each store site. The next section provides an
overview of the tools available in Business Analyst.
2.1.4. Esri’s Trade Area Definition Tools
Esri’s Business Analyst has many trade area definition tools which can be used to delineate a
site’s trade area. These seventeen trade area tools are summarized in Table 1. These tools allow
users to define trade areas in a variety of ways, using methods ranging from simple to complex.
Additionally, to help a user characterize trade areas, Esri’s Business Analyst contains a
substantial data resource about businesses, business performance, population, demographics, and
consumer spending information, as well as Tapestry Segmentation data which provide consumer
profile information. This software also allows for custom data layers to be imported. A custom
population data layer was created and imported for this case study.

10

Table 1 Esri's Trade Area Tools and Descriptions

(Miller 2010)
Esri’s trade area toolset utilizes geographic, distance and other data driven variables to
delineate trade areas. Geographic features utilized to define trade areas include census tracts and
Tools Description
Create Trade Area From Geography Levels
Generates trade areas based on standard
geographic units.
Create Trade Area From Sub-geography Layer
Generates trade areas from the features of an input
polygon layer that intersects a defined boundary
layer.
Customer Derived Trade Areas
Creates trade areas around stores based on the
number of customers or volume attribute of each
customer.
Data Driven Rings
Creates a new feature class of ring trade area
features. The radii are determined by a field in the
ring center (store) layer.
Dissolve by Attribute Range
Aggregates and dissolves features based on
specified attributes.
Drive Time
Creates a new feature class of trade areas, based
on drive time or driving distance, around store point
features.
Grids
Generates an equidistant vector based grid network
for a specified area.
Huff's Equal Probability Trade Areas
Generates areas of competitive advantage
boundaries between stores weighted on one or more
variables. These weights can be calculated based on
the results of a Huff Model.
Market Penetration
Calculates the market penetration based on
customer data within an area.
Measure Cannibalization
Calculates the amount of overlap between two or
more trade areas.
Monitor Trade Area Change
Creates a new feature class and report that analyze
how trade areas have changed over time
Remove Trade Area Overlap
Removes overlap (cannibalization) between trade
areas
Static Rings
Creates a new feature class of ring trade area
features using a set of radii
Thiessen Polygons
Generates competitive advantage trade areas for
each store by creating boundary lines equidistant
from each of the store locations.
Threshold Data Driven Ring
Creates rings around stores. The radii of the rings
are determined by expanding from the store location
until they meet the criteria included in the store layer.
Threshold Trade Areas
Creates rings around your stores. The radii of the
rings are determined by expanding from the store
location until they meet your criteria.
11

administrative units. Distance measures are used to define Thiessen polygons, static rings, grids,
and drive time trade areas. The rest of Esri’s trade area tools incorporate distance with other
supplementary variables such as store and competitor information, customer data, demographics
and consumer spending information. Business Analyst is a robust tool that provides a user with
all that is needed for most trade area definition tasks (Miller 2010).
2.2 Determining catchment population characteristics
Catchment population characteristics are calculated using existing census and other demographic
and economic data which are widely available. The key challenge here is how to divide and
redistribute the population counts and characteristics from the standard census and other
polygons into the trade areas determined for a specific store location site analysis. In this section,
the geography of census data is briefly summarized in order to explain the nature of aggregated
census data. This is followed by a discussion of issues that arise from the use of aggregated data
and some techniques that have been used to disaggregate it, including the use of land use data.
Finally the concept of areal interpolation and specifically dasymetric mapping techniques similar
to those used in this case study are introduced.
2.2.1 Census geography
Privacy concerns prevent census data from being released at the household level. Thus,
household level information is aggregated to blocks which are geographic areas delineated in
such a way that the total population is between 600 and 3,000 people and the demographic
characteristics within the block are somewhat homogeneous. Blocks are themselves aggregated
into larger block groups and block groups into census tracts. Importantly, for ease in some kinds
of trade area analyses, blocks are also often reduced to centroid point features called block points
12

(Peters and MacDonald 2004). Block level data tables contain only population counts. Full
census data are provided at the block group and larger aggregates.
2.2.2. Issues related to the use of aggregated data
Household count data can be aggregated into an infinite number of different size and shape
polygons, each of which may be just as valid due to spatial autocorrelation. Spatial
autocorrelation is the idea that sampled geographic data will likely be more similar to that from
nearby locations than from more distant ones. As stated by Tobler in 1970, “everything is related
to everything else but near things are more related than distant things” (Tobler 1970). Thus when
dealing with Census data, population characteristics are likely to be similar to others nearby due
to a human tendency to group near like individuals.
However, geographic boundaries can be manipulated to produce population characteristic
distributions that are favorable or unfavorable to a specific end. Aggregation of geographic data
results in the Modifiable Areal Unit Problem (MAUP) (Wong 2009). Congressional District
gerrymandering and research conducted un-objectively to support desired outcomes are
examples of such intentional manipulations related to the MAUP (Kelly 2012). Many dissimilar
instances grouped together in arbitrary or manipulated ways also lead to research outcomes and
real world outcomes that are biased and unrepresentative of realities.
2.2.3. Collapsing aggregated data to polygon centroids for spatial overlay
When using census aggregates above blocks (e.g. block groups or tracts) as the source layer for a
spatial overlay, Business Analyst uses a method called weighted block centroid retrieval. Here
each block centroid is assigned a proportion of its higher level aggregate’s data values based on
each block’s population as a percentage of the enclosing census aggregate’s population. Once
population values are assigned to block points, during the overlay process, values from the
13

source layer are included in the target layer results based on inclusion of block points within the
target polygons. According to Business Analyst help documents on Spatial Overlay this will be
more accurate than simple centroid inclusion retrieval method using the larger census aggregate
centroids.
Using aggregated data collapsed to the centroid block points as done by Esri’s weighted
block centroid retrieval can still be problematic. In an overlay analysis of spatially incongruent
layers, “centroid containment” rules for inclusion of source layers features in an overlay result is
the most basic form of dealing with features having spatially mismatched boundaries (Miller
2010). For example, if a block centroid falls outside of a trade area, the entire block population
will be treated as not intersecting the trade area even though in reality some of the block
population may fall within the trade area. Conversely, a block polygon that intersects only a
small sliver of the trade area but whose centroid is within the trade area’s extent produces
outputs reflecting complete inclusion of all block population from the source layer in the result.
When applied to population polygons, entire populations are accounted for in the features
intersecting the centroid and no population is reflected for features where the intersecting layer
does not intersect the centroid despite potentially intersecting a majority of the population
aggregate polygon (Miller 2010).
Using centroid containment for inclusion and exclusion in analysis is undesirable unless
the target areas of a spatial overlay are large relative to the census aggregates. Large errors will
occur when source and target features are similar in size (Ignizio and Zandbergen 2010).
Methods which are more advanced than rules for inclusion and exclusion in addressing spatially
different polygons in analysis are commonly referred to as areal interpolation according to
Goodchild and Lam as cited in (Zandbergen 2011).
14

2.2.4. Methods of areal interpolation
The most basic form of areal interpolation is areal weighting which weights population included
and excluded in a source zone by the ratio of its intersection to the source layer feature area.
Another form of aerial interpolation is “surface fitting” where a surface is fitted to the data in
source areas and typically inferential statistics are used to interpolate values (Zandbergen 2011).
Another method of areal interpolation is dasymetric mapping in which ancillary data are
used to distribute population unequally within the source layer features. Dasymetic mapping is
the process of disaggregating data into finer units of analysis to help refine locations of
population or other phenomena (Mennis 2003). The results preserve known population within
each source area in the target area results. This is referred to as the pycnophylactic property (Qiu
and Cromley 2013).
2.2.5 Land use as auxiliary data to disaggregate population within census zones
Land cover is the ancillary data most often used to refine population distribution using
dasymetric mapping. Dasymetric mapping using land cover data usually employs an overlay of
population polygons with land cover classified data. Population is apportioned to varying land
cover areas by assigning weights to their land cover classifications. Importantly, some land cover
classifications such as water or natural areas are weighted zero and not apportioned any
population. The remaining populated areas are apportioned by aerial weighting. These aerial
weighted polygons assume population distribution is uniform in target zones but since these
areas are usually much smaller than the aggregates, the results are a more accurate estimate (Qiu
and Cromley 2013; Zandbergen 2011).
Early on, spectral signatures in Landsat images were used to spatially delineate various
kinds of land cover (Amaral et al. 2012). Training areas with known land cover were used to
15

determine the recorded spectral signatures for these known land cover types. These spectral
signatures were then used to classify land cover throughout the Landsat images.
Another approach is to use remotely sensed thermal images to classify land use and land
cover by comparing heat emissions from morning and night images (Wen and Yang Xiaofang
2011). Water bodies produce almost no heat emission so they are easiest to identify. Increasing
levels of daytime heat emission are recorded from undeveloped land, residential and commercial
areas. Industrial areas and high rises produce the largest heat emissions during the day.
Residential areas tend to have a much greater contrast of heat emission levels between day to
night as people are home and use energy more frequently at night. Commercial and industrial
areas display the opposite effect as these areas are not typically operational at night.
Zanbergan and Ignazio (2010) used census block group population with large scale land
cover data as the ancillary data to estimate population. Actual census block group population
counts were then compared against calculated block group population to see the error produced.
The authors used areal weighting, land cover, total imperviousness, imperviousness above 75
and 60 percent, “cleaned imperviousness” total roads, local roads and nighttime lights datasets to
conduct dasymetric mapping. Imperviousness refers to the imperviousness of surfaces to show
where population is likely with population more likely on the least impervious surfaces due to
paving and structures likely in these areas. Similarly road density is another surrogate for the
presence of population.
Land cover and imperviousness performed the best among the datasets used with the
lowest errors produced. Zandbergen and Ignazio found errors ranging from 11.9 to 14.5 using
landcover data. Later in 2011, Zanbergen augmented his previous study and added address point
16

and residential address point data to the same analysis. Again he found land cover had a similar
error, this time at 11.6 percent, while imperviousness performed better at 10.8 percent.
Far outperforming the other datasets, address points had only 4.9 percent error and
residential address points produced only 4.2 percent. For this reason, a dataset of only populated
developed residential parcels called “Populated parcels” in short was created and used in this
study. To produce this dataset census aggregated populations were areally weighted to residential
areas with existing structures.
2.3 Background Summary
Methods for estimating a store’s sales potential by analogue, regression and trade area methods
were illustrated. Esri’s trade area creation tools were described. Census geographic data was
overviewed and issues arising from using aggregated data were discussed. Methods of areal
interpolation used in past research to disaggregate aggregated data including dasymetric mapping
were reviewed.
Calculation of trade area catchment population and demographics used to estimate store
sales potential depends on trade area definition and population most often provided as census
aggregates. A methodology of dasymetric mapping to disaggregate population as well as
methods of trade area creation used by this study to find differences in calculated catchment
population and demographics are reviewed in the following Chapter.

17

CHAPTER 3: DATA AND METHODS
Analyzing a retail store’s trade area population is crucial to determining the potential of a retail
site with the results affecting decision making for constructing, locating or relocating, or
maintaining a retail site. A case study was conducted using different aggregation levels of
population data and various methods to create store site trade areas to show the effects on
calculated trade area population counts and related population attributes. The accuracy of results
will have an effect real estate decision making based on the calculated characteristics of
catchment population.
Potential retail sites located in various parts of San Diego County with different
population densities were selected for this case study to see if results varied by population
density. This case study explores how trade area population characteristics found by using
different population estimates from aggregated Census population data and concentrated
population areas impact the calculation of catchment population and related characteristics. A
generalized workflow for this case study is shown in Figure 1.

18

Figure 1. Project Workflow
3.1 Geographic Data Sources
Data for these analyses were collected from Esri Business’ Analyst, the US Census, and
SanGIS/SANDAG regional data warehouse. These data sources provide data in the form of
tables, shapefiles and geodatabase for boundary and attribute data. When data layers were used
in this spatial analysis, Esri’s Business Analyst automatically converted data layers from
differing datum and projects them to be compatible with the base layer datum.
3.1.1 Assessor Parcel Data
The Assessor Parcel Data produced by SanGIS contains boundary files for each tax parcel in San
Diego County. Each parcel has a unique assessor parcel number (APN) and the corresponding
parcel’s legal boundary and ownership information. Each APN number is unique to ownership
entity or group, however boundaries of APNs in the case of condominiums, which all occupy the
same parcel boundary, are overlaid resulting in multiple ownership entities or groups for a single
19

geographic boundary. Alternatively multifamily housing (Apartments and Senior Living
Facilities) while having many residents generally have one owner for one geographic boundary.
In the case of both condominiums and apartments, one parcel boundary can have many
households living in a single parcel boundary.
Single family residences have a one to one parcel boundary to household ratio. This
results in different parcel areas having one household and one owner, many households and one
owner or many households and many owners. “Populated parcels” were areally weighted by the
percentage of total area for “Populated parcels” within each block. The corresponding population
for each block was multiplied by these weights to arrive at a new estimate of population for each
polygon with source layer population preserved in target features referred to as the
pycnophylactic property of dasymetric mapping. This new layer assumes that population
distribution is uniform across these smaller polygons. However, this resulted in large single
family parcels of wealthy households being allocated the same population of a multifamily
parcel the same size, especially where estates are large and a large area of land surrounds the
residence. This means that a large estate may have received a population estimate similar to an
entire apartment complex which has a much greater population despite a similarly sized area thus
skewing population estimates.
3.1.2. Retail Analysis Data Demands
Retail site data utilized by most retail developers and analysts needs to be obtained easily and at
a low cost. Esri’s Business Analyst provides generalized albeit rich data about consumers.
Generally, survey data provide more detailed information about a specific retail site, such data
are more time consuming to obtain and beyond the needs of most retail developers. Based on my
experience in my office with current level of data used in real estate, data provided by Esri’s
20

Business Analyst are sufficient for most analysis needs and for this reason utilized in this study.
Retail site data for this case study are provided by Esri’s Business Analyst.
3.1.3. Land Use Data
Land Use data were provided by SanGIS’s regional data warehouse. These data show the land
use classification for each parcel in San Diego although contiguous APNs with the same land use
classification are aggregated into larger land use polygons. There are several different
designations within each classification of residential, commercial, retail, industrial, open space,
conservation and more. For this analysis only the residential areas of single family, multifamily,
condominium, student housing, were used to better classify population distribution. Other
residential distinctions such as prisons, hotels, and others were omitted.
3.1.4 Population and other Demographic Data
Population and corresponding demographic data for this case study were taken from both Esri’s
Business Analyst and the US Census. Population data obtained from Esri’s Business Analyst are
from the biennial census as well as the American Community Survey. Non-census year data are
derived from Esri’s own models which project population totals and demographic characteristics.
Esri’s Business Analyst has data from the Census for Tract and Block Group Census boundaries.
Block level data from Esri’s Business Analyst are available but block boundaries are not given
and blocks are represented by block centroids called block points. In order to account for the
boundaries of Census Blocks, data obtained from the Census website directly provided the Block
Boundaries and population counts within those blocks.

21

3.1.5 Developable Land Data
The county of San Diego develops many layers of ancillary data for use by its employees and the
public. One of the layers produced by San Diego County is the “Developable Land” layer. This
layer is comprised of areas that the county deems developable based on favorable topography not
prohibitive to development costs and that is currently vacant. Land which may be undeveloped
but having topography that is not relatively flat or with zoning restrictions or conservation
protections are not considered favorable and are excluded from this layer.
3.2 Case Study Store Sites
The four store sites in San Diego County chosen for this study are pictured in Figure 2. Two of
these store sites are located in suburban city areas (San Diego and Poway) and the other two are
in cities in more rural areas of the county (Alpine and Ramona). As described below, each of
these cities is unique in population, demographics and bordering communities. They were chosen
to provide a wide range of conditions over which to test this methodology.
22

Figure 2: Store locations in suburban and rural areas of San Diego County. Suburban store
locations are in the city of San Diego and Poway, rural locations are in the cities of Ramona and
Alpine. Study area for data extent is outlined in blue.

The San Diego store site is located in the affluent community of Carmel Valley which is
mostly comprised of single family residences and condominiums. The even more affluent cities
of Del Mar and Rancho Santa Fe with multimillion dollar beach homes and equestrian ranches
lie to the west and north respectively. To the east is the city of San Diego proper with many new
single family home developments which were constructed prior to the housing crash with graded
lots and open space awaiting future development. To the immediate south is a large commercial
area home to Qualcomm’s ever growing campus and other commercial buildings with few
apartment communities throughout the area. Even further to the south are the city of La Jolla and
the Miramar marine base. The San Diego store location is near the coast to the west and has
23

convenient access to Interstate 5 which connects to Interstate 805 to the south as well as Hwy 56
to the east.
The second suburban site located in the city of Poway is similar to the San Diego site in
density but is slightly less affluent due to being more inland where residential real estate costs
are slightly more affordable. The city of Poway and the Poway store are located next to Interstate
15 and Hwy 56 allowing for quick travel to the north, south, and west and Poway Road a major
arterial road allows for quick travel east.
The third store located in the city of Alpine is considered a more rural location. Alpine is
located close to Interstate 8 allowing for quick travel to the east and west, however Alpine is the
easternmost city in San Diego County and surrounded by Native American reservations and hilly
topography with sparse population to the north east, and south. Past the Native American
reservations, a few miles to the west are the outskirts of the San Diego County suburban area
which are more affluent and densely populated relative to Alpine.
Lastly, the fourth store is located in the rural city of Ramona. Ramona is located in the
hilly region in the north central part of San Diego County just west of the city of Escondido,
which is the north easternmost city on the edge of San Diego County’s urban sprawl. Ramona
has one access road leading into Escondido. However that road is a rural road and is time
consuming to travel. As a result, Ramona is a city largely isolated from the rest of the county and
getting in and out of Ramona is difficult. Ramona is surrounded by sparsely populated hilly areas
in all other directions further isolating the city.
3.3 Determining Catchment Population
Catchment populations, the population within a store’s trade area, are often modeled by looking
at where people have their residences, but it might also be derived based on where people work
24

or by taking into account travelers and commuters. While different ways of defining catchment
population distribution are important and deserve further exploration, for the purposes of this
case study the catchment population is determined using three different datasets. The first two
are the traditional approach using the Census level aggregations, in this case census block groups
and tracts. The third approach uses parcel and land use data to concentrate the census population
aggregates into truer representations of where population most likely resides by removing those
areas within census aggregates where population does not live. The next section describes the
disaggregation process used here.
3.3.1 Disaggregation of Census Population Aggregates
Disaggregation of census population aggregates for the third approach to determining catchment
population was achieved through the use of Land Use and other ancillary data from the county.
This required a multistep extraction process.
First, using the county’s “Land Use” layer, polygons with residential land use
designations of multifamily, condominium, single family residences, dormitories and military
housing polygons were isolated. Some other residential land use classifications such as hotels,
hospitals were not included as residential land use, despite the county classification of residential
land use, as they are not counted toward population by the Census. Prisons were also omitted
from inclusion despite their residential classification by the county due to prison population
lacking the mobility necessary to patronize a retail store. While prison areas were removed, their
population is included in Census data. These populations were not removed from the analysis
here, but since they are small numbers relative to the general population, inclusion of these
numbers in the total figures used is not considered an issue of concern. Also, since most San
Diego prisons are in the southern portion of the County, south of the study area, there should be
25

minimal impact on results. The isolated residential areas were extracted into a new feature layer
called “Residential land use”.
As noted earlier, land use polygons are aggregations of contiguous assessor parcels of the
same land use. Assessor parcel boundaries are completely within or identical to land use
boundaries. As a result, it is possible to isolate each assessor parcel within each land use
polygon. Thus, the next step involved overlaying “Residential land use” polygons on the assessor
parcels. The intersecting features were extracted and exported to a new feature layer called
“Residential parcels.”
Residential land use means that residential uses are allowed but this does not indicate
whether an area was developed or undeveloped. Undeveloped areas do not have residences
despite the residential land use designation. Thus, to ensure that the parcels to be included in this
analysis have structures on them and were not merely designated residential, undeveloped
parcels were identified. The “Developable Land” layer created by the county containing parcels
with potential for development, but not yet developed, was utilized. “Developable Land” was
overlaid on “Residential parcels” and the intersecting areas were removed as these areas were
currently undeveloped. The remaining parcels was exported to create “Developed residential
parcels” layer.
“Developed residential parcels” does not yet completely indicate the existence of a
residential structure on a parcel as some vacant areas within residential land use polygons are not
considered developable by the county due to issues of fitness for development. Land fit for
development is relatively flat, currently undeveloped, with permissible zoning and not subject to
conservation protections. Parcels which do not meet these requirements despite falling within
26

residential land use zoned areas cannot be extracted using the “Developable Land” data. Thus,
Census block data were used to further remove such unpopulated areas.
Census data collected about individual households are aggregated initially to census
blocks and combined to form higher levels of aggregation. Census blocks are contiguous for all
United States geographies including uninhabitable areas such as water bodies, conservation
areas, mountainous or hilly areas or valleys where residences are nonexistent or very sparse. As a
result, it is possible to use unpopulated census blocks to further identify unpopulated parcels.
This was accomplished by isolating census block polygons which have no population.
Census polygons do not share exact boundaries with the assessor parcels. In order to create
shared boundaries for both layers “Developed residential parcels” were split along all Census
block boundaries. New polygon boundaries were formed when “Developed residential parcels”
were located in more than one Census block. The resulting dataset assigned each new
“Developed residential parcels” area with its corresponding block attribute. All “Developed
residential parcels” area which intersected census blocks having no existing population were
removed. The remaining polygons, referred to as “Populated parcels”, have both population and
existing residences. Figure 3 shows an example of the result of this process.
27

Figure 3: “Developable Land” (outlined in red) overlain on “Residential parcels”. Here the final
“Populated parcels,” shown as blue areas, overlay the more extensive yellow “Residential
parcels” which shows through the clear developable land polygons. The Developable Land
polygons in this image are examples of polygons removed from the “Residential parcels” layer.

These remaining polygons provide a more accurate consolidated representation of where
census populations reside than the census aggregates which contain many areas where people do
not live. It was then necessary to apportion population to the “Populated parcels” from the block
aggregates. In order to do this, the percent of each block polygon area occupied by each of the
“Populated parcels” within each census block was calculated. Population for each resulting
parcel polygon was then apportioned by multiplying the block population and household count
by the calculated area percentage. All area percentages for each “Populated parcels” within a
block when added together by block summed to one. The result is population for each block
28

aggregate allocated by area to the existing “Populated parcels” within each block, providing a
more concentrated representation of each census block’s population distribution.
Weighting population and households by area of these polygons is only one method of
disaggregating population and has many shortcomings. For example, some densely populated
areas like apartment complexes or high rise areas which have much more population are
allocated here the same population as rural areas that contain only one house for an area similar
in size to a nearby apartment community. Accounting for these differences is worth further
investigation but not in the scope of this study.
3.4 Calculation of Detailed Drive Time Trade Areas
Drive Time Trade Areas were constructed in Business Analyst using the Business Analyst Trade
Area tools. For the purposes of this study, Drive Time Trade Areas were estimated around store
locations with varying densities of population selected in San Diego County. Drive Time Trade
Areas were created using Drive Time Trade Areas of 3, 5, and 7 minutes per site. The Detailed
Drive Times option was selected so that only the road network able to be traversed within the
selected time frames was included, excluding areas not reachable. This results in a more precise
output. If detailed drive time areas are not selected, the areas covered by road networks are
joined by distant end points to form larger polygons which may include areas which may not be
reachable within the allotted time.
3.5 Calculation of Radial and Network Distance Trade Areas
Trade areas were constructed in Business Analyst using the Business Analyst Trade Area tools
for Simple Rings and Drive Time distance. For the purposes of this study, distance Trade Areas
29

were estimated around all stores described in Section 3.2. Trade Areas were created using radial
and network distances of 3, 5, and 7 miles for each site.
3.6 Calculation of Trade Area Population and Characteristics
Esri’s Business Analyst Drive time tool allows for the creation of reports when a drive time area
is calculated summarizing the drive time area with population and consumer spending
characteristics. In lieu of this method, for this project spatial overlay was used to find the
catchment population and corresponding characteristics for each drive time trade area. Spatial
overlays were also used to calculate the population and characteristics of the radial and network
distance Trade Areas defined in the previous subsection.
When conducting these overlay analyses, Esri’s Business Analyst automatically used two
different methods based on the population dataset used as the source layer. When using
‘Populated parcels’, the overlay process apportioned the percentage of population equal to the
area of the polygon within the trade area. When using the census aggregates as the source layer,
Business Analyst used the weighted block centroid retrieval method described earlier. The
results of these spatial overlay functions using the three different population data sources for all
trade areas are discussed in the following chapter.

30

CHAPTER 4: RESULTS
As described in the previous chapter, for the study area used in this case study, shown in Figure 2
in the previous chapter, population data were concentrated to residential areas using a
combination of land use parcels, supplemental county data, and census block data. The results of
these operations are illustrated in Figure 4 which shows in blue the areas where “Developed
residential parcels” were found.

Figure 4 : “Developed residential parcels” in San Diego County. Supplemental data was used to
identify developable land for all land uses. Developable land was extracted and removed from
the Residential land use polygons. Remaining “Developed residential parcels” have residential
land use designation and existing development.

Lastly Census blocks without population were also removed. The resulting polygons,
illustrated in transparent blue in Figure 5, are referred to as “Populated parcels” for the remainder
of this document. The blow-up in Figure 5 with “Populated parcels” showing in a transparent hue
31

allows visual verification of the correct classification of these areas when overlain on satellite
imagery. The areas in Figure 5 in yellow are uninhabited areas within census blocks with
population. Figure 6 shows just the “Populated parcels”, in blue overlaid on an aerial image to
illustrate more clearly where these are located.

Figure 5 : Polygons with existing residential land use development aka “Populated parcels”
overlaid on yellowish background areas representing all Census Blocks. Census Block Group
boundaries are shown in Black and these are overlain by Census tract boundaries in red. The
extent of the yellow area shows that a much larger area is included in the Census zones whether
or not there is any population contained.

32

Figure 6 : “Populated parcels” showing imagery of underlying residences.
4.1 Results of Drive Time Trade Areas
For the first portion of this case study trade areas were formed around store sites using detailed
drive times for 2 different product type scenarios. The first scenario was for a store with non-
specialty goods and services meaning consumers would be willing to drive smaller distances to
obtain these items than more specialized items. For this scenario drive time areas were created
for each store at three, five, and seven minute intervals. Figures 7, 8, 9 and 10 show the
corresponding drive time trade areas for the four store sites in San Diego County.
33

Figure 7 : Trade Areas using detailed drive times for three, five, and seven minute intervals.
San Diego (Del Mar Heights Rd.) store
34

Figure 8: Trade Areas using detailed drive times for three, five, and seven minute intervals. Poway store;
35

Figure 9 Trade Areas using detailed drive times for three, five, and seven minute intervals. Ramona store
36

Figure 10: Trade Areas using detailed drive times for three, five, and seven minute intervals. Alpine store
37

To account for stores that sell more specialty goods which customers would be more
inclined to endure greater travel difficulty to procure, additional drive time areas at time intervals
of five, ten, and fifteen minutes were created. These larger trade areas, shown in Figure 11 were
created for the westernmost location of San Diego County and two easternmost locations
eliminating the Poway location. If the Poway location were included drive time areas for the
suburban locations which overlapped would cause errors in the spatial overlay results for this
study. For this reason the Poway location was omitted from analysis of these larger drive time
distance thresholds.

Figure 11 “Populated parcels” intersecting detailed drive time areas of 5, 10, and 15 minutes for
the San Diego, Ramona, and Alpine store sites. The suburban San Diego location has noticeably
more residential parcels than the more rural Ramona and Alpine.

38

This case study explored the effect on results of calculations of catchment population and
demographics using concentrated population data as opposed to census aggregates. The effects
on the results could impact retail real estate decision making. Spatial overlay was used to find the
catchment population characteristics for the Drive Time Trade Areas using the “Populated
parcels”, Census Block Groups, and Tracts. Differences in the calculated catchment populations
using the 3 base population layers were analyzed. In order to conduct the Spatial Overlay a
custom .bds (Business Analyst Dataset) layer was created using the “Populated parcels”
described in Chapter 3 along with the layer’s derived census attributes for each population
polygon. The “Populated parcels” were imported and the population and households for each
polygon were apportioned by area. All other attributes were joined to the associated block group
level and weighted by either population or household depending on the normalizing metric. The
normalizing metric is a count such as population or households for which measures such as per
capita income and average household income are derived for Census data (Peters and
MacDonald 2004). Block Group and Tract data from Esri’s default pre-packaged .bds layers
were used for Spatial Overlays at the corresponding aggregate population level.
All detailed drive time trade areas for each store were calculated individually for each
store and time increment in order to ensure that Spatial Overlays would be calculated as an
aggregate from store location to the outer extent of the detailed drive time trade area. This means
that despite overlapping of trade areas of different distances all population within the trade area
boundaries was used for the calculations for each trade area. The results of the Spatial Overlays
are shown in Table 2 for the grouping of three, five, and seven minute trade areas. The
“Populated parcels” overlaid in the Spatial Overlay process are pictured in Figure 6 through
Figure 9.
39

Figures 12 through 15 show all “Populated parcels” intersecting detailed drive time trade
areas of three, five and seven minutes for the store locations. These land use polygons are
overlaid on a less vibrant drive time trade areas of three, five and seven minutes of the same
color. Vibrantly colored areas indicate areas within the store’s trade area where population lives.
“Populated parcels” not within the trade area for each site are shown with a gray outline.

Figure 12 : All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Alpine location. These “Populated parcels” are overlaid on a less vibrant
detailed drive time trade areas of three, five and seven minutes of the same color. Vibrantly
colored areas indicate residences. “Populated parcels” outside the trade area for the site are
shown with a gray outline.
40

Figure 13: All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Ramona location. These “Populated parcels” are overlaid on a less vibrant
detailed drive time trade areas of three, five and seven minutes of the same color. Vibrantly
colored areas indicate residences. “Populated parcels” outside the trade area for the site are
shown with a gray outline.
41

Figure 14 : All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Alpine location. These “Populated parcels” are overlaid on a less vibrant
detailed drive time trade areas of three, five and seven minutes of the same color. Vibrantly
colored areas indicate where people live. “Populated parcels” outside the trade area for the site
are shown with a gray outline.
42

Figure 15: All “Populated parcels” intersecting detailed drive time trade areas of three, five and
seven minutes for the Poway location. These “Populated parcels” are overlaid on a less vibrant
detailed drive time trade areas of three, five and seven minutes of the same color. Vibrantly
colored areas indicate where people live. “Populated parcels” outside the trade area for the site
are shown with a gray outline.

The results show that disaggregated polygons (i.e. the Populated parcels) yielded
different results for Spatial Overlay of the trade areas than Block Groups and Tracts yielded.
Block Groups and Tracts yielded identical Spatial Overlay results. This was due to the weighted
block centroid method for retrieval and inclusion in a Spatial Overlay described earlier. Spatial
overlays for block groups and tracts would have the same results for the same trade areas given
inclusion of their populations based on weighted block centroids.
43

Table 2 Population Household and associated demographic information for detailed drive time
trade areas of store sites at drive times of three, five, and seven minutes. PP is an abbreviation for
Populated parcels and BG and T is an abbreviation for Block Groups and Tracts.

“Populated parcels” give a more accurate picture of where population is located. Analysis
of the San Diego and Poway site Spatial Overlay results show that population counts in suburban
areas experience a significant spike when using “Populated parcels.” Calculation of
demographics using “Populated parcels” for small trade areas yield results similar to block group
and tract spatial overlay most notably at higher densities. This is due to the fact that smaller trade
areas especially in dense suburban areas are comprised of smaller block groups and tracts which
when overlaid yield similar results.
Another trend of note seems to show deviation of results increasing with greater numeric
values. Larger percent changes in results seem spurred by larger counts or dollar amounts (i.e.
per capita income results show smaller changes than average income, and average income results
show smaller changes than average net worth which are numerically larger).

Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 98319 84026 14293 14.5% 99128 82728 16400 16.5% 100920 88908 12012 11.9%
San Diego 142762 144123 -1361 -1.0% 160212 144664 15548 9.7% 166672 151087 15585 9.4%
Poway 114510 93647 20863 18.2% 117737 103389 14348 12.2% 126961 119573 7388 5.8%
Ramona 69902 65531 4371 6.3% 78491 70939 7552 9.6% 83550 79822 3728 4.5%
Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 31961 31490 471 1.5% 33432 31406 2026 6.1% 34575 33614 961 2.8%
San Diego 53630 53871 -241 -0.4% 57529 57041 488 0.8% 59022 59161 -139 -0.2%
Poway 32001 30876 1125 3.5% 36223 35004 1219 3.4% 41654 41556 98 0.2%
Ramona 20947 20827 120 0.6% 23925 22717 1208 5.0% 25990 25993 -3 0.0%
Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 759178 520392 238786 31.5% 785006 463783 321223 40.9% 823286 582145 241141 29.3%
San Diego 995967 1022112 -26145 -2.6% 1170041 995131 174910 14.9% 1186639 1041376 145263 12.2%
Poway 953819 654552 299267 31.4% 964756 772205 192551 20.0% 1028999 946598 82401 8.0%
Ramona 374462 325255 49207 13.1% 501238 404958 96280 19.2% 578765 536022 42743 7.4%
Drive Time 5 Minutes
Average Household Income
Drive Time 5 Minutes Drive Time 3 minutes Drive Time 7 minutes
Average Net Worth
Per Capita Income
Drive Time 3 minutes
Drive Time 3 minutes Drive Time 7 minutes
Drive Time 7 minutes Drive Time 5 Minutes
44

Figure 16 – “Populated parcels” polygons for detailed drive time areas of 3, 5, and 7 minutes.
“Populated Parcels” within 3, 5, and 7 minutes are shown in yellow, red, and blue respectively
without overlap. Block Groups are outlined in black and Tracts are outlined in green. This figure
demonstrates how population distribution is not uniform across census blocks or tracts. The
northernmost tract shows population is concentrated mostly in the southern area of the tract.

The results of five, ten, and fifteen minute Drive Time Trade Areas are shown in Table 3.
The results at these larger trade areas were also identical for block groups and tracts which is
further evidence of the weighted block centroid retrieval. Similar effects for those seen in the
results of smaller trade areas were seen in these larger trade areas. Differences in population
calculated were found but were not as large relative to smaller trade areas. This makes sense as
45

larger areas intersect an increased number of the census blocks due to their larger size resulting
in an averaging effect as more source layer features are include in larger trade areas.
Table 3 Population Household and associated demographic information for detailed drive time
trade areas of store sites at drive times of five, ten, and fifteen minutes. PP is an abbreviation for
Populated parcels and BG and T is an abbreviation for Block Groups and Tracts.

4.2 Comparison of Trade Areas Created by Distance Measures to Drive Time Trade Area
The standard method of calculating population and demographic information used in real estate
marketing and decision making as described in Chapter 1 is to utilize radial distance rings in
such analysis. This case study also sought to determine how calculated catchment population
changed using radial distance and two other trade area creation methods using network distance
measures as well as drive time distance measures in miles and minutes respectively. Calculations
for each of the four store sites in San Diego were conducted at 3, 5, and 7 mile radial distance,
and network (road) distance as well as 3, 5, and 7 minute drive times. A spatial overlay for all of
these differently defined trade areas with both the “Populated parcels” and population aggregates
Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 7631 6056 1575 20.6% 15182 14036 1146 7.5484% 55310 50849 4461 8.0654%
San Diego 30611 28211 2400 7.8% 80954 74182 6772 8.3652% 286593 279359 7234 2.5241%
Ramona 10754 9199 1555 14.5% 17271 15677 1594 9.2293% 25098 22560 2538 10.1124%
Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 99128 82728 16400 16.5% 105988 95580 10408 9.8% 95246 85572 9674 10.2%
San Diego 160212 144664 15548 9.7% 166152 152141 14011 8.4% 144149 119247 24902 17.3%
Ramona 78491 70939 7552 9.6% 90025 86050 3975 4.4% 97357 94417 2940 3.0%
Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 33432 31406 2026 6.1% 36356 35828 528 1.5% 31027 30711 316 1.0%
San Diego 57529 57041 488 0.8% 57592 58867 -1275 -2.2% 46735 46211 524 1.1%
Ramona 23925 22717 1208 5.0% 28723 28001 722 2.5% 31972 31100 872 2.7%
Site
PP BG and T Change % Change PP BG and T Change % Change PP BG and T Change % Change
Alpine 785006 463783 321223 40.9% 934549 750152 184397 19.7% 840523 682198 158325 18.8%
San Diego 1170041 995131 174910 14.9% 1188120 1059119 129001 10.9% 1029372 786510 242862 23.6%
Ramona 501238 404958 96280 19.2% 668838 612131 56707 8.5% 792018 762617 29401 3.7%
Drive Time 10 Minutes Drive Time 5 minutes
Drive Time 3 minutes
Drive Time 15 minutes
Average Net Worth
Per Capita Income
Drive Time 5 minutes
Drive Time 5 minutes Drive Time 15 minutes
Drive Time 15 minutes Drive Time 10 Minutes
Drive Time 10 Minutes
Drive Time 15 minutes Drive Time 5 Minutes
Population
Average Household Income
46

was conducted to find the effect on the results. Figure 18 below pictures these differently defined
trade areas for each store at 7 minute and mile distance measures as an example.

Figure 17: 7 minute Drive Time Trade Areas and 7 mile network and radial distance trade areas.

Table 4 shows the results calculated for “Populated parcels” and census aggregates for 3
minute drive time trade areas, beneath these are the results for “Populated parcels” and census
aggregates for 3 mile network distance trade areas, with the results for “Populated parcels” and
census aggregates for 3 mile radial trade areas at the bottom of the table. Table 5 and Table 6
show similar results but at distance measures of 5 minutes and miles and 7 minutes and miles
respectively. The results of these analyses are discussed below.
47

Table 4 Population Household and associated demographic information Alpine, San Diego,
Poway, and Ramona sites for all trade areas of distance 3. PP is an abbreviation for Populated
parcels and CENSUS is short for census aggregates. POP is an abbreviation for Population.
AVGHHINC is an abbreviation for average household income. PCI is an abbreviation for per
capita income. AVGNW is an abbreviation for average net worth.

The results in Table 4 show that for 3 minute and mile distance measures calculations of
population and demographics using the same trade area were roughly the same when using either
the “Populated parcels” or census aggregates. The differences that do exist show that for each
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 3694 85096 31961 530545
San Diego 15023 141760 53630 972439
Poway 9753 96211 32001 669394
Ramona 7418 65560 20947 316226
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 3642 84026 31490 520392
San Diego 12372 144123 53871 1022112
Poway 7330 93647 30876 654552
Ramona 7148 65531 20827 325255
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 9648 94560 35259 690459
San Diego 39910 156907 59431 1109670
Poway 43062 116033 38761 927667
Ramona 12810 76545 24668 471347
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 7481 90697 34065 615426
San Diego 37205 149484 58776 1027426
Poway 38810 113081 38412 901039
Ramona 11897 76539 24654 477058
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 13511 99248 36689 789520
San Diego 66172 159157 59366 1096045
Poway 85962 120734 41597 948290
Ramona 18653 83108 27196 565243
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 13401 98498 36708 773433
San Diego 66464 155702 59518 1059901
Poway 85422 117977 41433 906868
Ramona 18349 83934 27174 578919
Network Distance Trade Area
Miles
Simple Ring Trade Area Miles
3
Drive Time Trade Area Minutes
48

population and demographic characteristic calculations using the “Populated parcels” are higher
for each result as opposed to those calculations using census aggregates. These differences in
results are more pronounced in urban store trade areas than the rural store trade areas.
While comparison of the results found using the same trade areas had little change, large
changes can be seen in the results found using the different trade areas. As can be expected using
radial distance provides the highest estimation of population with figures almost double in some
cases the population of the network distance trade areas and roughly four times the population of
drive time trade areas. This same trend is however not seen in the results of the demographics as
demographics will likely be similar at shorter distances due to autocorrelation.
Table 5 shows results for trade areas defined using 5 minute and mile distance measures.
The results for trade areas at these distances are similar to those found at 3 minute and mile
distance measures. Generally calculations of demographics were similar but tend to be
underestimated using aggregated population when compared to those found using “Populated
parcels.” Population calculations for trade areas found at distance measures of 5 follow the same
trends as those found at trade areas found at distance measures of 3. The one exception would be
Store 2 which is coastal and therefore trade areas are constrained by the coast and limited in
expansion to one side. As a result the differences in population are not to the same magnitude for
the coastal store site as the other store sites.
49

Table 5 Population Household and associated demographic information Alpine, San Diego,
Poway, and Ramona sites for all trade areas of distance 5. PP is an abbreviation for Populated
parcels and CENSUS is short for census aggregates. POP is an abbreviation for Population.
AVGHHINC is an abbreviation for average household income. PCI is an abbreviation for per
capita income. AVGNW is an abbreviation for average net worth.

Table 6 shows results for trade areas defined using 7 minute and mile distance measures
to construct trade areas. The results are similar to those results found using trade areas delineated
with 3 and 5 minute and mile distance measures. As with the previously defined smaller trade
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 7631 88916 33432 584622
San Diego 30611 151948 57529 1072966
Poway 25212 109135 36223 839904
Ramona 10754 74152 23925 440390
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 6056 82728 31406 463783
San Diego 28211 144664 57041 995131
Poway 20870 103389 35004 772205
Ramona 9199 70939 22717 404958
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 13353 98960 36423 781709
San Diego 69787 156715 59065 1095429
Poway 113522 121273 42503 962195
Ramona 16494 83792 27647 584441
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 12052 96254 35900 732027
San Diego 65994 151621 59265 1052956
Poway 108880 117463 42151 916911
Ramona 14962 82791 26915 573707
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 18156 102501 37433 847659
San Diego 105899 154838 56715 1116241
Poway 189917 123234 42483 991728
Ramona 29511 95773 32216 772277
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 18001 101710 37474 832697
San Diego 106595 150448 56524 1059742
Poway 187782 119883 42401 954646
Ramona 29604 96957 32246 794279
Drive Time Trade Area Minutes
Network Distance Trade Area
Miles
Simple Ring Trade Area Miles
5
50

areas the 7 minute and mile trade areas reflect minimal change in population calculations within
the same trade area but large population changes found using different distance measures for
trade area delineation. This difference is however not as drastic in the larger distance trade area
population calculations. However the smaller distance trade areas were more uniform in their
results for demographics. At larger distance measure more room for deviation is more likely
which is reflected in the results having less consistent trending than the smaller distance trade
areas.

51

Table 6 Population Household and associated demographic information Alpine, San Diego,
Poway, and Ramona sites for all trade areas of distance 7. PP is an abbreviation for Populated
parcels and CENSUS is short for census aggregates. POP is an abbreviation for Population.
AVGHHINC is an abbreviation for average household income. PCI is an abbreviation for per
capita income. AVGNW is an abbreviation for average net worth

4.3 Summary of Results
Census population aggregates produced calculations of Drive Time Trade Area catchment
population which were less than calculations produced by “Populated Parcels”. Population
differences were more significant at smaller distances but insignificant overall. Significant
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 9305 92624 34575 657480
San Diego 45494 156598 59022 1096948
Poway 61192 121309 41654 964233
Ramona 13955 79535 25990 523983
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 7595 88908 33614 582145
San Diego 40229 151087 59161 1041376
Poway 55983 119573 41556 946598
Ramona 13105 79822 25993 536022
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 16445 100424 36843 818781
San Diego 87348 154452 58193 1078976
Poway 222146 121886 42246 966998
Ramona 23348 94043 31246 745614
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 14731 97190 36273 767085
San Diego 81820 152207 59113 1056063
Poway 216912 118648 42187 936064
Ramona 21594 94123 30805 750308
SITE POP PP AVGHHINC PP PCI PP AVGNW PP
Alpine 26544 103864 38000 889477
San Diego 273615 125537 45920 882579
Poway 291443 121797 41687 958207
Ramona 35503 101206 34140 851466
SITE POP CENSUS AVGHHINC CENSUS PCI CENSUS AVGNW CENSUS
Alpine 25939 102508 38019 871014
San Diego 284790 123246 45456 839547
Poway 294964 119761 41589 937457
Ramona 35697 102002 34160 870044
Drive Time Trade Area Minutes
Network Distance Trade Area
Miles
Simple Ring Trade Area Miles
7
52

differences in catchment population were found when different distance measures were used for
creation of trade areas. Simple Ring trade area estimates were roughly double and quadruple
population counts than those found using network distance and drive time distances.
Simple ring is most common trade area creation method used in retail real estate
marketing and subsequent decision making. The large differences in population estimates show
that radial distance measures overestimate catchment population and how crucial trade area
definition is to catchment population calculations. The differences found in this case study show
further investigations into trade area delineation and disaggregating population are warranted.
Recommendations for further studies are given in the next Chapter.
53

CHAPTER 5: DISCUSSION AND CONCLUSIONS
The results of this study expose possible shortcomings of methods of analyses currently used in
retail real estate marketing and decision making. This case study explored how calculation of
catchment population and related demographics using “Populated parcels” contrasted with
analysis using census aggregates commonly used in real estate. It also examined how more
precise methods for calculating a store’s trade area affected calculated catchment population and
related demographics compared with catchment population calculated using simple ring and road
network distance trade areas, typically used in real estate.
Concentrating population data to areas where population lives provides improved
calculations of catchment population compared to those calculated using census aggregates.
When used to calculate catchment population and corresponding characteristics for Drive Time
trade areas, census aggregates only include population where trade areas pass through the block
centroids within the census aggregates. As such, all population for those census blocks whose
centroids are not within the trade area are fully omitted from catchment population calculations
despite having some population within a trade area. Simultaneously, census blocks whose
centroids are within a trade area have the entire population of the census blocks included in
catchment population calculations for trade areas.
Concentrating population to only residential areas and weighting those areas by
percentage of their area as a ratio of total residential areas within their census block provides a
more precise estimate of population. This overcomes the shortcoming of total inclusion or total
exclusion of population using weighted block centroid inclusion for calculation. Inclusion of
previously omitted population and exclusion of some portions of previously over-included
population aggregates results in more accurate catchment population calculations which were
54

more inclusive of actual population within a trade area. Results found using census block groups
and tracts produced identical results which reinforces this idea. Block groups and tracts should
ideally have results that vary as they are not identical. Results for both census units should
intersect each aggregate level differently, pointing to the problem with using the aggregate
centroids as the qualification for inclusion in calculated characteristics is a flaw in this approach.
This study also showed how using “Populated parcels” produces different calculations of
catchment population than census aggregates. Calculated catchment population differences were
minimal, however, the calculated related demographics experienced much larger percent
changes. Even if catchment population estimates are similar, large differences in demographics
show the potential for problems in subsequent real estate modeling or decision making using
these demographics for analysis. These differences in population and demographic calculations
found with concentrated population vary depending upon population density of the trade area
and other study area community characteristics.
While this study showed how using concentrated population data may provide greater
insight, obtaining population characteristics collected via surveys at the block level is also
necessary as much demographic information is not distributed below Block Group aggregates.
This would be very time consuming and add cost. However, this study’s results display
ecological fallacy, where characteristics of larger aggregate data were imposed on disaggregated
data and may not correlate with the new lower level geography. Using parcel level survey data
would allow for characteristics to be verified against new lower level polygon aggregates and
disaggregates.
The effects of using concentrated population and aggregate population data were then
extended to using different distance trade area creation methods as well. Distance measures used
55

were road network minutes and miles as well as simple ring radial miles which is most
commonly used in real estate to construct trade areas. Distances of 3, 5, and 7 minutes and miles
were used to construct store trade areas for each site and catchment population was calculated
using each set of population data.
The results of this investigation showed similar results to the previous results for all trade
areas. However, what was most notable were the drastic increases in trade area size and
calculated catchment population from trade areas constructed using network minutes to miles
and again from network miles to simple ring trade areas. Population calculations for network
minutes to network miles roughly doubled. The same effects can be seen from network miles to
simple ring miles, resulting in roughly four times the calculated population from network
minutes to radial miles.
The shortcomings of using simple rings or radial distance measures were discussed in
Chapter two. Network distance is a much better measure of distance traveled and Drive Time
distance measured in minutes is an even better indicator of actual costs of reaching a store site
which impacts a consumer’s willingness to travel. Despite this, simple ring distances are
commonly used in real estate marketing and decision making. The large differences in
calculations of catchment population using these different trade area creation approaches show
what may be a major shortcoming of using simple rings for real estate marketing and analysis as
simple rings largely overestimate catchment population.
The investigations undertaken in this case study show that using concentrated data may
result in large differences in demographic calculations as opposed to using census aggregates.
Additionally it was shown that traditional methods of trade area calculation may drastically
overestimate population in a trade area, meaning that the actual population likely to patronize a
56

store may be less by roughly double or four times that calculated traditionally. These
shortcomings can be detrimental to real estate decision making and ultimately could cost a great
deal of money to investors, retail business owners, and banks which finance them a great deal of
money. Further investigation into these outcomes is warranted and future studies could be
expanded in a variety of ways.
5.1 Directions for further research
This study shows that further investigation of how dasymetric mapping along with different trade
area creation methods impact a store site’s calculated catchment population and demographics is
warranted. Possible directions for future research should include augmenting this study with new
methods of weighting population in dasymetric mapping or using other trade area creation
techniques to define trade areas to see these effects. These and other possibilities for future
research are discussed in the following sections.
5.1.1 Weighting Population
Population was areally weighted by percentage of area for all residential parcels within a block.
Other methods for weighting population could also be used in future analysis and again verified
with survey data to see how these methods affect outcomes of disaggregating populations and
weighting the corresponding characteristics. Three different weighting methods include
weighting residential land areas by number of units from assessor data, weighting population by
different land use classifications, and weighting populations for all land uses by land use as well
as area.
A simple improvement to the current study’s weighting would be to weight the value by
number of units in the case of multifamily and student housing as well as by number of APN’s in
57

the case of condominiums. This would ensure that each area would be weighted by possible
households. Average Household Size would be multiplied against this weighting to get
population estimates. Further investigations could also be done which take in to account the
vacancy factor for the area homes which could be found from real estate literature derived from
surveys. Without taking into account a vacancy rate, an area could be overestimated, though with
a current vacancy factor of two percent in San Diego County overestimation may not be
significant in this study.
Weighting land uses by land use classification means weighting all existing built
structures by land use classification. This would be done to simulate where people are during the
day. For example, residential and commercial properties could have a higher weight for their
classifications than say industrial as industrial properties tend to have few workers. Retail parcels
would be weighted similar to industrial for number of employees but have an increased
weighting due to customer draw, which would increase likelihood that people would be coming
from these places.
An extension of weighting land use by land use classification would be to weight land
uses by land use and area as well. Weighting by area would give increased weight to larger areas.
Weighting by a land use classification alone would mean that larger residential areas would be
weighted the same as small residential areas. By weighting populated areas by both land use
classification and area as weights areas more likely to have population as determined by land use
would also be weighted proportionally to their size.
5.1.2 Different Trade Area Techniques
Another variation on the current study would be to use the disaggregated data to see how
different trade area creation methods affect the resulting population of trade areas for store site’s
58

trade area population, demographic and economic characteristics. This method would use
different trade area creation methods in Business Analyst not utilized in this study. For example,
Thiessen polygons could be used to construct trade areas defined by creating trade areas where a
store’s trade area is defined by all points nearest to the store than any of its competitors.
Another trade area method where store competitors are considered in defining a store’s
trade area is the Huff Gravity Model. In this model, store sites are assigned trade areas by many
factors similar to the regression method discussed in Chapter two. Other factors such as store
square footage, which is often used as a proxy for store sales, might also be used in the model.
5.1.3 Other possible improvements for the future
Additional improvements to this study would involve doing more store trade areas for more sites
and in different counties with varied population densities. Studies similar to this investigation
could be conducted for other suburban counties in southern California and other urban counties
in areas like Chicago and very rural counties in states like Arizona or Wyoming on the opposite
end of the spectrum. Having a larger sample size of sites of varying densities would allow for
greater identification of trends resulting from different site population densities.
Such additional methods would be important to investigate because retail sites could fail
or succeed depending on estimates of catchment population and demographics. Better
information would increase level of success of site suitability analysis and allow financers to be
more accurate. Real estate transactions and sale lease backs are based on these metrics and priced
accordingly. Improper or less accurate data can lead to poor investing and cost real people real
money.
59

5.2 Conclusions
The methodology used in this case study produced results which were more representative of the
population within each trade area than the calculations found with the census aggregates.
Removing portions of census aggregates and concentrating population to where people actually
live is an improvement over the centroid inclusion method which does not take into account
actual population distribution. This case study shows that further investigation into this type of
population concentration is warranted. Future directions for improvement on this research
described earlier in this chapter and analysis of these effects will lead to a far better calculation
of catchment population and population characteristics and could be immensely beneficial to
calculations of store site suitability analysis. Employing these techniques and improvements
could also be a differentiating factor for a researcher or real estate agent in marketing oneself
versus their competitors.
Regardless of the methods used, calculation of underlying population and population
characteristics should be as accurate as possible. Businesses that have better information are
better equipped to pick a location for a store that will help to ensure the success of the store and
of the investment in the store. Additionally in real estate sales, retail property owners, banks and
investors make determinations about store profitability and worth based on many of the
population characteristics determined by researchers for each property. Information of better
quality and precision can lead to better decision making and increase the probability of a site’s
success and of investor and bank returns.

60

BIBLIOGRAPHY
Amaral, Silviana, Andre Augusto Gavlak, Maria Isabel Sobral Escada, Montiero, and Anotnio
Miguel Vieira. 2012. "Using Remoe Sensing and Tract Data to Improve Representation of
Population Spatial Distribution: Case Studies in the Brazilian Amazon." Population and
Environment 34: 142-170.
Birkin, Mark, Graham Clarke, Martin Clarke, and Alan Wilson. 1996. Intelligent GIS : Location
Decisions and Strategic Planning. 1st ed. Vol. 1. New York, NY: Pearson Professional
Limited.
Fenker, Richard and Juli Zoota. 2001. "Intuitive Retail Modelling: Does Science have Anything
to Offer?" Journal of Corp Real Estate 3 (3): 248-259.
Ignizio, Drew A. and Paul A. Zandbergen. 2010. "Comparison of Dasymetric Mapping
Techniques for Small-Area Population Estimates." Cartography and Geographic
Information Science 37: 199-214.
Kelly, Jason P. 2012. "The Strategic use of Prisons in Partisan Gerrymandering." Legislative
Studies Quarterly 37 (1): 117-134.
Mennis, Jeremy. 2003. "Generating Surface Models of Population using Dasymetric Mapping."
Professional Geographer 55 (1): 31-42.
Miller, Fred L. 2010. Getting to Know ESRI Business Analyst. 380 New York Street, Redlands,
CA: Esri Press.
Mitchell, Andy. 2009. The Esri Guide to GIS Analysis Volume 2 :Spatial Measurement and
Statistics. First ed. Redlands, CA: Esri Press.
Peters, Alan and Heather MacDonald. 2004. Unlocking the Census with GIS. First ed. Redlands,
CA: ESRI Press.
Qiu, Fang and Robert Cromley. 2013. "Areal Interpolation and Dasymetric Modeling."
Geographical Analysis 45 (3): 213-215.
Tobler, W. R. 1970. "A Computer Movie Simulating Urban Growth in the Detroit Region."
Economic Geography 46 (Supplement: Proceedings. International Geographical Union.
Commission on Quantitative Methods): 234-240.
Wen, Xingping and Hu Guangdu Yang Xiaofang. 2011.Journal of Indian Science of Remote
Sensing June 2011 (39): 193-201.
61

Wong, David. 2009 "The Modifiable Areal Unit Problem (MAUP)." In The SAGE Handbook of
Spatial Analysis, edited by A. Stewart Fotheringham and Peter A. Rogerson, 105-125.
London: SAGE Publications Ltd.
Zandbergen, Paul A. 2011. "Dasymetric Mapping using High Resolution Address Point
Datasets." Transactions in GIS 15: 5-27.

62

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Selection of bridge location over the Merrimack River in southern New Hampshire: a comparison of site suitability assessments

PDF

Spatial delineation of market areas: a proposed approach

PDF

Estimating populations at risk in data-poor environments: a geographically disaggregated analysis of Boko Haram terrorism 2009-2014

PDF

Providing a new low-cost primary care facility for under-served communities: a site suitability analysis for Service Planning Area 6 in Los Angeles County, California

PDF

A site suitability analysis for an inland port to service the ports of Los Angeles and Long Beach

PDF

Community gardens for social capital: a site suitability analysis in Akron, Ohio

PDF

Installing public electric vehicle charging stations: a site suitability analysis in Los Angeles County, California

PDF

Finding food deserts: a study of food access measures in the Phoenix-Mesa urban area

PDF

Relocation bay: identifying a suitable site for the Tampa Bay Rays

PDF

Soil lead contamination from the Exide battery smelter: the role of spatial scale in cleanup efforts

PDF

A model for emergency logistical resource requirements: supporting socially vulnerable populations affected by the (M) 7.8 San Andreas earthquake scenario in Los Angeles County, California

PDF

Preparing for earthquakes in Dallas-Fort Worth: applying HAZUS and network analysis to assess shelter accessibility

PDF

Defining neighborhood for health research in Arizona

PDF

Evaluating predator prey dynamics and site utilization patterns of golden eagles using resource selection modeling and spatiotemporal pattern mining

PDF

An analysis of racial disparity in the distribution of alcohol licenses and retailers in Orange County, California

PDF

Site suitability analysis for implementing tidal energy technology in southern California

PDF

Analysis of park accessibility in Redan, Georgia Web GIS application

Asset Metadata

Creator Cisneros, Alfredo David (author)

Core Title Population disaggregation for trade area delineation in retail real estate site analysis

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 02/05/2015

Defense Date 12/19/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag census,dasymetric mapping,disaggregation,distance,Land use,OAI-PMH Harvest,Population,retail real estate,San Diego County,site analysis,trade area

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kemp, Karen K. (committee chair), Bennett, Victor (committee member), Oda, Katsuhiko (Kirk) (committee member)

Creator Email adcisner@usc.edu,alfredo.cisneros@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-529792

Unique identifier UC11297889

Identifier etd-CisnerosAl-3162.pdf (filename),usctheses-c3-529792 (legacy record id)

Legacy Identifier etd-CisnerosAl-3162-0.pdf

Dmrecord 529792

Document Type Thesis

Format application/pdf (imt)

Rights Cisneros, Alfredo David

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA