University of Southern California Dissertations and Theses

Spatial delineation of market areas: a proposed approach

(USC Thesis Other)

Spatial delineation of market areas: a proposed approach

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
SPATIAL DELINEATION OF MARKET AREAS:

A PROPOSED APPROACH

by

Amanda Christine Gray

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 2014

Copyright 2014 Amanda Christine Gray

ii

DEDICATION
I dedicate this document to my parents, grandmother, sisters and friends, for their constant
support.

iii

ACKNOWLEDGMENTS
I would like to acknowledge the excellent guidance of my thesis committee chairman, Dr.
John Wilson and thesis committee members Dr. Daniel Warshawsky and Dr. Flora
Paganelli.

iv

TABLE OF CONTENTS
Dedication ii
Acknowledgments iii
List of Tables v
List of Figures vi
List of Abbreviations viii
Abstract ix

Chapter One: Introduction 1
1.1 Defining market areas 2
1.2 A proposed framework for defining market areas 3
1.3 Thesis organization 11

Chapter Two: Related Work 13
2.1 A history of trade area delineation methods 13
2.1.1 Ring models 14
2.1.2 Gravity models 15
2.1.3 Voronoi (Thiessen polygon) models 16
2.1.4 Trend surface area models 17
2.1.5 Consumer behavior in trade area creation models 17
2.2 Esri’s Approach to Market Area Definition 18
2.3 Other Approaches to Market Area Definition 20

Chapter Three: Methods and Data 25
3.1 Description of Study Area 26
3.2 Data Sources 28
3.2.1 Grocery data 29
3.2.2 Demographic data 31
3.3 Calculation of trade areas 32
3.4 Calculation of market areas 33
3.5 Association of expenditure data with trade and market areas 35
3.6 Calculation of market share 36
3.7 Evaluation of model results 37

Chapter Four: Results 39

Chapter Five: Discussion and Conclusions 54
5.1 Analysis Results 54
5.2 Areas for framework improvement 59
5.3 Final Thoughts 60

References 62

v

LIST OF TABLES

Table 1: Descriptions of trade area creation tools available in Esri’s
Business Analyst 19

Table 2: Demographic profile of Worcester County cities chosen for
market area delineation study 30

Table 3: List of target groceries 31

Table 4: Market share calculation results using political boundaries
as market areas 46

Table 5: Market share calculations for target stores using degrees
of separation method for market definition 52

Table 6: Comparison of market share calculations between boundary
and degrees of separation methods for delineating market
areas 53

vi

LIST OF FIGURES

Figure 1: Degrees of Separation market delineation illustration:
Hypothetical store locations on a plane 7

Figure 2: Degrees of Separation market delineation illustration:
Hypothetical store locations on a plane with trade areas 8

Figure 3: Degrees of Separation market delineation illustration:
Hypothetical store locations on a plane with market area
defined 9

Figure 4: Huff and McCallum (2008) map showing market area
delineation 21

Figure 5: Boots and South (1997) map showing market area
delineation in Kitchener-Waterloo, Ontario, Canada 22

Figure 6: Project workflow 25

Figure 7: Steps to create market area using degrees of separation
method 26

Figure 8: Map of Worcester County and surrounding area 27

Figure 9: Screenshot of Esri’s Business Analyst Equal Competition
Trade Area creation tool wizard input requirements and
options 33

Figure 10: Groceries in and around Worcester County, Massachusetts 39

Figure 11: Equal Competition trade areas created for groceries in and
around Worcester County, Massachusetts 40

Figure 12: Views of the trade areas created around each target store for
each town 41

Figure 13: The trade areas for the Shaw’s and Super Stop & Shop in
Worcester, the Hannaford’s in Leominster, and the Shaw’s
in Northbridge, Massachusetts 43

Figure 14: Degrees of separation market area for a Price Chopper in
Worcester, Massachusetts 47

vii

Figure 15: Degrees of separation market definition for Shaw’s in
Worcester, Massachusetts 47

Figure 16: Degrees of separation market definition for the Super
Stop & Shop in Worcester, Massachusetts 48

Figure 17: Degrees of separation market definition for Colemo’s
Market in Worcester, Massachusetts 49

Figure 18: Larger scale view of the trade areas of the Leominster, Massachusetts
Hannaford’s market area 50

Figure 19: Degrees of separation market definition for Shaw’s in
Northbridge, Massachusetts 51

Figure 20: Comparison of results from two market area definition
methods 55

Figure 21: Closer view of the Shaw’s location in Worcester,
Massachusetts 56

Figure 22: Closer view of the Hannaford’s market area in Leominster,
Massachusetts 57

viii

LIST OF ABBREVIATIONS
BDS Business Analyst Dataset
CEX Consumer Expenditure Survey (US Census Bureau)
MCI Multiplicative Competitive Interaction
MSA Metropolitan Statistical Area
NAICS North American Industry Classification System
SSI Spatial Sciences Institute
USC University of Southern California

ix

ABSTRACT
In business sectors such as consumer staple goods, the location of a retail store is central
to a consumer’s decision to purchase items at a given location. As such, customers of
some retail segments (e.g. groceries) could be considered spatial entities, with behavior
directly tied to their location relative to the location of a retail store. A “trade area”
represents the geographical space from which a business draws customers and “market
area” refers to the larger geographic region in which several businesses compete. Market
areas are usually defined by an arbitrary spatial extent, such as a political boundary or
radius around a given point, which injects arbitrariness into any market share
calculations. In the staple goods retail sector, travel time and physical location are critical
to consumer store preference. Therefore, the accurate delineation of a store’s
geographical trade area and surrounding market is needed for precise performance
measurement and market share calculation. A framework is proposed for defining retail
market areas based on degrees of spatial separation between a store and surrounding
customers and competitors. Four degrees of customer separation are proposed, classifying
customers based on their consideration of, and/or preference for, a specific store location.
The result is a less arbitrary definition of a market. This approach was modeled for the
grocery sector of Worcester, Massachusetts and the resulting market share calculations
were compared and implications of using one method over another are discussed.

1

CHAPTER ONE: INTRODUCTION
The customer is paramount in retail, and profits are made and lost based on the ability of a
store to draw consumers to their doors. Thus, retail businesses place great emphasis on
understanding the decision-making processes which influence their customer base to shop
with them versus a competitor. These decisions are extremely complex and vary infinitely
based on the consumer and store in question; yet, in particular retail sectors, a consumer’s
spatial proximity to a store may be the dominant factor contributing to a consumer’s
willingness to shop at a given location. “How long will it take me to get to store x versus
store y for product n?” is central to deciding between competing locations, especially where
staple goods and low price differentiation are present.
The importance of the above question to particular retail sectors means that
customers are essentially spatial entities from the perspective of a store manager or
business planner. Thus, the ability of a business to accurately delineate the geographic
space from which its customers originate is crucial to accurate sales forecasting, marketing
efforts and thorough understanding of the competitive environment in which the business
operates. Under these assumptions, the calculation of market share – one of the most
critical metrics in assessing retail performance – is based on a delineation of geographic
space.
A “trade area” represents the geographical space from which a business draws
customers and “market area” refers to the larger geographic region in which several
businesses compete. Retail businesses, especially those focused on staple goods or
convenience purchases, define their trade and market areas spatially. Businesses from all
sectors can be said to have spatially-defined market areas, but the emphasis on the exact

2

size and shape of that area is of greater importance in highly competitive, staple goods
markets with low product and price differentiation. A luxury goods business may draw
customers from great distances, sporadically placed, and those customers may be highly
brand-loyal. Grocery stores or gas stations, however, tend to draw customers from
nearby, who prioritize the convenience of the business’s location over the brand of the
business itself.
A store must understand both its trade area and the trade areas of its competitors
in order to compete effectively in their market. Familiarity with their own trade area is
useful in learning about the characteristics and preferences of their existing customer
base. Understanding the trade areas of their competitors informs them of both their
competitors’ relative strength versus their own, and where opportunities exist for
capturing additional customers or market share (Huff, 1963).

1.1 Defining market areas
Much attention has been given to methods for determining a single store’s trade area.
Multiple techniques exist and have been in use for decades (e.g. Applebaum, 1966; Huff,
1964). Less attention has been given to methods for spatially defining the larger market
in which a store competes. Historically, political boundaries such as a city, county or
Metropolitan Statistical Areas (MSA) served as a sufficient proxy for a store’s
competitive market, or markets were defined as space a given distance away from a
central location (Huff and McCallum, 2008).
While one or more of these proxies is satisfactory for most market share analyses,
they lack accuracy and customization to a particular market’s nuances, and in certain

3

situations, can be misleading. Should a political boundary exclude important players in a
store’s competitive environment, or include businesses within a boundary who do not
compete for any of the same customers as the target store, inaccurate conclusions may be
drawn. Both of the aforementioned outcomes constitute the “edge effects” accompanying
the market area definition and will be discussed in more detail below.

1.2 A proposed framework for defining market areas
A more accurate method for defining a store’s market would be to classify the customers
and competitors surrounding a target store location based on their likelihood to impact
the performance of the target store. Such a classification system could group customers
regardless of any political boundary or distance threshold they do or do not fall within,
as:
1. First-degree customers: Customers falling within the target store’s trade
area;
2. Second-degree customers: Customers falling within a trade area directly
adjacent to, or touching, the target store’s trade area (these trade areas
would belong to “direct” competitors);
3. Third-degree customers: Customers falling into the trade areas directly
adjacent to the trade areas of “direct” competitors (these trade areas would
belong to “indirect” competitors); and
4. Fourth-degree customers: Customers not included in the trade areas of
the target store or its direct or indirect competitors.

4

This classification begins to separate customers according to their “degree of
separation” from a given grocery store. Those customers who shop with a target store
could be said to be “first-degree customers” because the location and offerings of the
target store are preferred over any other competitor. Other customers who may not
consider a target store as their primary source for groceries are separated from the target
store by other competing groceries, or by one spatial “degree” or more from the target
store’s location. Fourth-degree customers are conceptually more complex than the other
degrees. They serve as a “catch-all”, in that they encompass all customers existing
beyond the third degree. Without a fifth degree of separation, fourth-degree customers
expand continuously into space outside the boundaries of third-degree customers, with no
outer limit. The utility of defining fourth-degree customers is to find the new boundary of
a market area. Where third-degree customers (i.e. customers of a target store’s indirect
competitors) begin abutting fourth-degree customers, a target store’s competitive market
area ends.
A similar type of classification could be applied to competing grocery stores to
emphasize the degree of competitive threat each one poses to a target store. Groceries
competing with the target store for the same customers, or whose trade areas overlap or
are adjacent to the trade area of the target store, could be considered direct competitors.
Since the performance of those competing groceries may result in the target store either:
(1) losing some of its customers to the direct competitor; or (2) “stealing” customers from
the direct competitor and converting them to customers, the competition is direct. If
instead, a nearby grocery store only competes for customers with the target store’s direct
competitors, and not with the target store itself, competition could be said to be indirect.

5

Such indirect competitors do impact the performance of a target store’s direct
competitors, and are therefore worthy of consideration in a market analysis, but their
performance carries less weight relative to the immediate success of a given grocery
store. Existing customers are unlikely to be gained or lost as a result of the presence of an
indirect competitor, unless the indirect competitor is capable of putting a direct
competitor out of business or they force the direct competitor to focus more of its
marketing efforts on the store of interest.
The final classification of competitors serves as a catch-all, much like the concept
of fourth-degree customers. If political boundaries are no longer treated as the outer
spatial limit at which additional groceries should be considered a part of a market area,
the new outer limit becomes that point at which competing groceries no longer impact the
performance of the target store. In other words, once neighboring groceries are neither
direct nor indirect competitors, they fall into a “non-competitive” degree of separation
and the market area boundary is drawn around the outermost indirect or direct
competitors surrounding a target store.
As presented, the degrees of separation approach to market area definition uses
binary relationships to describe consumer behavior across space. The definitions of each
degree of separation assume a hard line between customer “degrees”, corresponding to
the presence or absence of adjacent trade areas. This may be too simplistic an approach,
and does not address issues like trade area overlap or the “fuzzy boundary” phenomena
identified by Huff (1964) in retail trade areas. The definitions of each spatial degree of
separation are meant as a foundation upon which more nuanced definitions can be built.
For example, in dense urban areas with much trade area overlap, the question of how the

6

overlap will be classified into degrees must be answered. Perhaps overlap will be
classified as its own degree (e.g. overlap between the target store’s trade area and its
next-door competitor might be considered second-degree customers, and the direct
competitor’s customers classified as third-degree customers); or perhaps overlap will be
classified as belonging to the degree of either the nearest or farthest competitor (e.g. as in
the previous example, overlap between a target store’s trade area and that of its direct
competitor would either be classified as only part of the target store’s trade area, or as
part of the direct competitor’s trade area.) In either case, the framework for identifying
degrees of separation gives the exercise of market area delineation a common language
and robust conceptual approach .
Figures 1-3 illustrate the concept of degrees of spatial separation in market area
determination using a series of admittedly simplistic examples.

7

Figure 1: A series of hypothetical store locations shown on a plane; a target store for
which a market area is desired is shown in red; nearby stores are shown in white. An
attempt will be made to define a market area for the target store irrespective of any
political boundaries or distance thresholds.

To execute the degrees of separation market delineation method, a target store and
surrounding stores must first be identified (Figure 1). Once store locations have been
confirmed, trade areas should be created around each business. This allows for each store
to be classified as direct, indirect, or non-competitive, and allows each store’s trade area
and customers to be classified according to their degree of separation from the target
store. In Figure 2, the target store’s trade area has been shaded red, and all trade areas
which touch the target store’s trade area have been classified as second-degree customers,
with their corresponding stores labeled direct competitors. Next, all trade areas touching

8

the trade areas of direct competitors, but not the target store’s trade area, are classified as
third-degree customers, with their corresponding stores classified as indirect competitors.
Non-competitive stores are those whose trade areas touch those of indirect competitors or
other non-competitors, and they contain fourth-degree customers.

Figure 2: Store locations from Figure 1, with hypothetical trade areas generated for each
store location; stores have been re-classified as direct, indirect or non-competitors;
customers within each trade area have been shaded according to the degree of separation
between them and the target store.

To calculate the market area for the target store, the trade areas of the target store,
the target store’s direct competitors, and the trade areas of the target store’s indirect
competitors are merged. In Figure 3, these trade areas are shaded in black. An alternate

9

way to define the boundary of the trade area would be the boundary between all third-
and fourth-degree customers, or between indirect and non-competitive store trade areas.

Figure 3: The market area for the target store has been defined as the combination of the
trade areas of the target store’s trade area and the trade areas of the target store’s direct
and indirect competitors.

This alternate definition of a market area prioritizes the degrees of separation
between a store and its surrounding competitors and customers and results in a market
area consisting of the sum of the trade areas belonging to the target store and its direct
and indirect competitors (Figure 3). There are two primary benefits to this “degrees of
separation” method over the use of administrative boundaries:

10

 This definition of a market captures all of the consumers and competitors who
impact the success of the target business, and none who do not.
 This method lessens the impact of edge effects on market share calculations; it
improves market analysis accuracy for businesses located near the edge of a
political boundary or distance threshold, where competition with stores across
the boundary may be ignored, or where competition with stores near the
boundary may be overstated.
This method of market area definition is not dependent upon the method of trade
area creation used. Any of the methods included in Esri’s Business Analyst, or a
customized method, could be used to create the individual trade areas. The method, data
and parameters chosen for the creation of the trade areas will be dependent upon the
individual scenario being mapped. The purpose here is not to investigate the most
appropriate method of trade area delineation, but rather the best way to use already-
created trade areas to inform the delineation of a wider market area.
That is not to say that this method of market area delineation would be applicable
to, or useful in, all scenarios. There are clear strengths and weaknesses of this approach
which render it more helpful in some and not other situations. The utility of degrees of
separation may be weakened, for example, in scenarios involving dense, urban areas with
a high number of competing businesses in a small area; in those cases, there may be such
great overlap between trade areas that defining customers by degrees is impossible. As
discussed earlier, in situations of great urban density, the definitions of each “degree”
may take on nuances according to the situation. Degrees of separation would still exist in
dense environments, but the number of direct and indirect competitors might be quite

11

high, and careful consideration of the treatment of overlapping trade areas would be
required.
The overarching goal of the thesis was to implement the new approach described
in Section 1.2 in Worcester, Massachusetts and to evaluate its performance across a range
of stores in varied geographic settings. An experiment was conducted to illustrate the
pitfalls of using political boundaries and threshold distances as market boundaries, by
comparing a market share analysis using political boundaries and threshold areas to
define a market area against an analysis using the proposed “degrees of separation”
method. Using the city of Worcester, Massachusetts as a test site, trade areas were
created for the grocery stores in and around the city. The market areas of several target
stores in the city were delineated, using both the degrees of separation method and the
city limit boundary of Worcester as a market area. Market share was calculated for each
target store under each market area delineation method to test the relative and absolute
sensitivity of market share calculations according to how market areas are spatially
defined.

1.3 Thesis Organization
The remaining chapters of this thesis attempt to build the case for the use of a standard
framework when discussing market area delineation, and propose a degrees of separation
approach as a logical solution given the lack of an existing framework. Chapter two
situates the need for common language and conceptualization of market area delineation
in the context of the history of trade area delineation methods. Trade area delineation has
a lengthy history of research, experimentation and discussion, which is in contrast to the

12

dearth of dialogue around market area delineation. Chapter two highlights the few cases
in which market area delineation is discussed, or a lack of a solution is noted, in
professional writings. This leads to chapters three and four, which detail the setup of a
market area delineation experiment using the proposed degrees of separation approach
and report the results using this new method and one based on Theissen polygons and
political boundaries. Chapter five discusses the implications of using the proposed market
area delineation framework, noting its strengths and weaknesses, and by imagining how
this new approach would play out in wider use.

13

CHAPTER TWO: RELATED WORK
This chapter explores the historical and current treatment of market areas as spatial
entities primarily based on political boundaries or distances from a given center point.
The conceptualization of a market area as a political boundary or distance threshold is not
inherently wrong or misguided; often, it is sufficient to achieve desired results. When
used consistently over time, all things being relative, the results from these methods can
still be useful. However, it is worth noting the general dearth of investigations into the
spatial definition of market areas; especially when methods of spatial delineation of
individual trade areas have been researched and refined for decades. Multiple researchers
have noted this gap in trade area research, but few have suggested structured approaches
for defining and delineating market areas.

2.1 Trade Area Delineation Methods
While the focus of this thesis is the delineation of market area boundaries, a brief review
of the techniques used to define individual store trade areas is provided below. The
proposed degrees of separation approach is not dependent on a particular method for
delineating trade areas, but because it uses trade areas to compose a market area, an
understanding of commonly used trade area definition methods is valuable to the reader.
Additionally, this section points out the typical undervaluing of market area delineation
in favor of individual trade area definition, as an important part of retail trade analysis.
Trade area delineation followed a general evolution from the early 20th century of
thinking about singular trade areas based on simple circles, to irregular polygons taking
into account multiple environmental factors, to deep concern with consumer decision-

14

making and probabilities. Generally, the discussion moved from a very black-and-white
conceptualization of trade area space, to an understanding of trade areas as flexible,
irregular and highly sensitive to consumer behavior. In the latter half of the 20th century
and early 2000s, research focused more on quantification of various factors influencing
consumer choice, such as pricing, store appearance, driving behavior and brand loyalty.
However, the impact of competitive forces on trade areas has remained relatively
unexplored, or simplistically treated, throughout this period. Research focused either
entirely on how to define a single store’s trade area, as if in a vacuum void of
competitors; or it did not adequately explain how or why particular stores were chosen as
competitors to a given store, while others were not. In other words, there was no
discussion of how to define the competitive market in which a single store operates.

2.1.1 Ring models
One of the simplest ways to conceptualize a store’s trade area is by drawing a circle
around the store location, with the store as the center point, assuming that all individuals
or households falling within the circle are customers of the store. Applebaum and Cohen
(1961) and Applebaum (1966) were some of the first researchers to formalize methods
for determining circular, or ring, trade areas. Generally, trade areas were delineated by
drawing concentric circles around a given store location until 80 percent of the total
number of the store’s assumed customers were included in the circle. This method
assumed that an analyst would have prior knowledge of the store’s customer base, and
assumed that customer behavior and customer attributes were uniform across the space
encompassed by the trade circle.

15

Decades later, Patel, Fik and Thrall (2008) presented a novel and clever update to
the idea of ring models with their wedge-casting trade area tool. They proposed defining
trade areas based on wedges which independently expand around a store location until a
given percentage of the customer base is accounted for. There are major flaws in this
design, such as ignoring competitive forces and as usual, assuming a homogenous
customer dispersion; but the concept breaks away somewhat from the the idea of trade
areas as simple, smooth polygons by allowing for very varied trade area shapes.

2.1.2 Gravity models
The most well-known and heavily imitated method of trade area delineation is the gravity
model, first adapted by David Huff (Huff 1964) for use in retail trade area delineation
from earlier versions introduced by William Reilly (Reilly 1931). Huff’s original gravity
model assumed that customers choose one retail site over another based on two factors:
(1) Euclidian distance to the store; and (2) the physical size of the store. Huff later refined
his model to include probabilities, so that boundaries between one retail site’s trade area
and a competitor’s could be blurred, or gradual, and that the surface of the trade area
polygon varied in saturation based on the probability that consumers would shop there
(Huff 1964).
Later researchers adapted Huff’s gravity model by experimenting with the
influence of factors other than store square footage on consumer store choice. Bucklin
(1971), for example, made the important point that multiple trade areas might exist for a
single store, depending upon the type of product being examined. Many retail businesses
sell to a wide range of customer bases, meaning that the trade areas for certain product

16

groups may be different from the trade areas for other products. Bucklin (1971) also
reiterated that boundaries between trade areas should not be thought of as distinct, clear
lines, but rather as gradations. Huff (1984) later responded to Bucklin’s ideas on multiple
trade areas for individual stores by proposing a “best fit” trade area which averages
multiple trade areas for a store into a single trade area.
The gravity model of trade area delineation has achieved widespread use due to
its ability to include multiple gravitational factors (i.e. store size, consumer
demographics, consumer preference factors, travel distance, or other metrics) to best suit
the situation being studied.

2.1.3 Voronoi (Thiessen polygon) models
Voronoi, or Thiessen polygon, trade area methods provide a quick and simple way of
defining trade areas. Based on the mathematical concept of Thiessen polygons, they
essentially assume no other influencing factor than Euclidian distance on the formation of
trade areas. In practical terms, they assume that customers frequent the store closest to
them. Boots and South (1997) did propose a Thiessen polygon method which was
capable of accounting for consumer preferences, bringing it closer to the gravity model
method. Of note in their research was the exploration of consumer tendency to split
purchasing between multiple locations on a regular basis. In their case, this splitting
behavior was quantified as a probability that a consumer would shop at a given location
at any given time. This too, is similar in thinking to Huff’s proposal that borders between
trade areas are gradual, and based on likelihoods rather than binary relationships.

17

2.1.4 Trend Surface Area models
Trend Surface Area models introduced an important debate on the heterogeneity of trade
area surfaces under pre-existing models. MacKay (1973) and Peterson (1974) discussed
the unlikelihood that trade area polygons were consistent throughout in terms of
consumer behavior and shopping choices. Because the concept of trade areas had at that
time been that of enclosed, homogenous polygons, MacKay and Peterson’s arguments
were ground-breaking in that they argued for non-homogenous, and even non-contiguous,
trade areas. Unfortunately, the methods they proposed for determining inter-trade area
variation were not practical for widespread use, as surveys and primary, household-level
data collection were presented as the most effective methods for data capture. Huff’s
previous assertions about trade area border gradation and his statistical approach to
variation in customer behavior were perhaps less realistic, but more practical, than the
methods of MacKay (1973) and Peterson (1974).

2.1.5 Consumer behavior in trade area models
The topic of trade area creation became a heavily mathematical and statistical one when
consumer behavioral factors became more important to trade area definitions. Nakanishi
and Cooper (1974), Davies (1977), Fik (1988) and Fotheringham (1988) initiated
discussions around the many factors influencing consumer choices, and Drezner (2006), a
mathematician, directed trade area delineation toward much more statistical analysis than
had been suggested by Huff and others.
Nakanishi and Cooper (1974) developed the multiplicative competitive interaction
(MCI) model. The MCI model added a level of consumer-oriented complexity to Huff’s

18

gravity model by allowing for independent weighting of gravity variables according to
individual consumer preference. Fik (1981) and Fotheringham (1981) went much deeper
into specific decision criteria for consumers who are selecting retail locations. Fik (1988),
for example, attempted to quantify the relative weight of price differences against travel
distances, and as such, provided a simple update to Huff’s original gravity model (by
essentially adding price as an additional gravitational factor). Fotheringham (1988), on
the other hand, argued that consumer decisions are hierarchical, and introduced the idea
of amenity clusters as an influencing factor on consumer behavior.
Drezner (2006) attempted to quantify the “soft” quality of store “attractiveness” in
consumer decisions. Her methods included direct customer surveys to validate her
findings, and are very specific to her particular case study of shopping malls, but the
metrics used and her approach to defining more complex consumer behavior was a
breakthrough in the field. Drezner also wrote on the need for the use of uncertainty in
trade area modeling, such that users of a trade area delineation tool could understand the
impacts of various scenarios on the shape and size of a trade area (Drezner 2009, 2011).
It was yet another step away from the older idea of trade areas as static, homogenous and
binary, and an endorsement of the use of trade area delineation as a continuous business
decision-making tool.

2.2 Esri’s Approach to Market Area Definition
Esri’s Business Analyst software provides an extensive toolkit with which to build trade
areas. Ten trade area creation methods are available in the software, some of which have
additional, branching options for the inclusion of particular variables or sub-methods

19

(Table 1). The software allows for the inclusion of custom data obtained by the user, and
is also integrated with Esri’s Business Analyst’s repository of demographic data,
spending data and business listings. As such, it is an example of a thoroughly-researched
trade area creation toolkit capable of handling substantial complexity.

Table 1: Descriptions of the ten trade area creation tools available in Esri’s Business
Analyst 10.1.

If a user desired to calculate the market share commanded by a particular store’s
trade area, Esri’s Business Analyst Market Penetration tool could be used. (Market
penetration and market share are not always interchangeable concepts, but the structure
of Esri’s Market Penetration tool effectively allows for market share calculation.) The

20

tool takes a customer layer (i.e. trade area layer, or customers known to “belong” to a
given store) and a base market layer as input, and divides the customers into the market
layer to determine penetration. Available options for the market layer are any layer
existing in the table of contents, or one of the preset Business Analyst Data Source (BDS)
layers. BDS layers are demographic and boundary files based on political and census
divisions, such as census blocks or ZIP codes. So, while it would be feasible for a user to
create a workaround by defining their own custom market layer in the Table of Contents,
there is no preset method for selecting multiple trade areas as a market area. Political
boundaries are emphasized instead.

2.3 Other Approaches to Market Area Definition
Examples of confusion surrounding how market areas should be spatially defined are
evident in both scholarly research and real-life case studies. David Huff, creator of the
Huff model for trade area creation (Huff 1966), wrote an Esri White Paper with Bradley
McCallum in 2008 (Huff and McCallum 2008) which treated the concept of a market
area as simplistically as drawing a freeform boundary around a pre-selected set of store
trade areas, and further reducing the area based on the placement of “barriers”, or major
roadways (Figure 4). Admittedly, Huff and McCallum (2008) were primarily focused on
demonstrating how to calibrate the Huff Model available through Esri’s Business
Analyst. However, the notion that competitive market areas could be defined that way
illustrates the lack, even decades after his groundbreaking research into trade area
creation, of tools for simply concatenating trade areas into competitive market areas in a
methodical and repeatable manner.

21

Figure 4: Drawing a freehand boundary around the trade areas of the outermost stores
and further restricting the polygon according to roadway barriers (in image, the multi-
colored lines intersecting the polygon) to form a competitive market area for a proposed
store location (Huff and McCallum 2008).

In other cases, it is simply taken as an assumption that the most appropriate
market boundary is a political one. For example, Boots and South (1997) proposed an
improved version of the Thiessen polygon (Voronoi diagram) method for delineating
trade areas. They improved on the original model by allowing for consumer choice
around a set of options in a market and the assumption that consumers could patronize
more than one supermarket location at a time. However, in demonstrating their method

22

using the supermarket chains in the towns of Kitchener and Waterloo in Ontario, Canada,
it was assumed that the administrative boundaries of the two towns were the hard
boundary at which competition and trade areas stopped. Figure 5 shows some of the trade
area definitions resulting from their research, and the sharp cutoffs of the trade areas
where they meet town boundary lines.

Figure 5: Trade areas created with the multiplicatively weighted Voronoi diagram trade
area calculation method of Boots and South (1997). The outer shape of the image is made
up of the city boundaries of Kitchener and Waterloo in Ontario, Canada. Inside the city
boundaries, trade areas are created (some are shaded) and cut off where they intersect the
city boundaries. The results are particularly jarring in the bottom right of the image,
where two curved trade area boundaries intersect the irregularly-shaped city limits.

The trade area creation methods proposed by Boots and South (1997), as well as
many methods proposed by other researchers, are adding impressive levels of detail and

23

customization to our ability to define trade areas. However, when trade areas sit within an
ill-defined or arbitrary market area, the value of their accuracy is seriously denigrated. In
Figure 5 above, the abrupt chopping off of some of the trade area boundaries would result
in inaccurate assumptions about customer counts, revenue potential and market shares. It
is highly probable that the actual trade areas in Kitchener-Waterloo extend over the
administrative boundaries and include some of the customers of competitors on the other
side of these boundaries as well.
Several researchers have discussed the potential error introduced by arbitrary
market boundaries and the need for a more formal framework for defining geographic
market areas. Weiss (1972) noted the incompatibility of political boundaries with
empirically-verified market areas. Twenty-three years later, Brooks (1995) observed that
“market definition is an important and unresolved problem in studies of market structure-
performance relationships.” Brooks (1996) was one among multiple authors to relate the
question of spatial market definition to the hospital industry and patient services (Ill et. al.
2004; Gaynor et. al. 2013). Brooks calculated market areas for hospitals in the San
Francisco, California area, using his “enacted market” approach (based on defining a
market using locations of past activities; in this case, where services have historically
been rendered by a set of hospitals) to calculate market boundaries and shares, and then
comparing results to the same example using political boundaries to define the market
areas and shares. Significant variability existed between his enacted market share
calculations and those rendered by the political boundaries approach. Brooks’ (1995)
work is a strong argument for developing and validating more thoughtful methods for
defining market areas. His “enacted market” approach, however, relies on the availability

24

of historical sales or services records for the target location and its competitors. This
thesis will suggest options for market area definitions when such data is not available.
In his 1995 Defining Market Boundaries article, Geoffrey Brooks noted “market
definition is an important and unresolved problem in studies of market structure-
performance relationships”, and he stressed the relationship between market share
calculations and an accurate geographic market definition, saying, “calculations of
market share or market concentration, terms typically used to assess conditions of
competition, have little meaning if the boundaries of the market have not been correctly
defined.” Bronnenberg and Albuquerque (2002) also referenced the lack of a “formal
definition” of a market area, but proposed to define it based on “consumer arbitrage”, or
the idea that markets could be separated and defined based on whether or not consumers
would invest in traveling to one or the other to benefit from price differentiation. This
was a meaningful step in placing the burden of market definition on consumer behavior
rather than store placement, and is relevant when considered at a national or international
scale, but falls short of providing a detailed framework for defining market areas under
various, smaller scale (i.e. small area) conditions.
This thesis provides criteria for classifying a given location in space as part of one
market area or another, but stops short of offering a method for defining a continuous
and/or fuzzy boundary between adjacent market areas. The next chapter documents this
method and the experiment that was implemented to test its efficacy.

25

CHAPTER THREE: METHODS AND DATA
A case study was undertaken to evaluate the performance of the degrees of separation
approach for defining market areas that was introduced in Chapter 1 using a variety of
political boundaries and distance thresholds. The grocery markets of Worcester County,
Massachusetts were used for this case study. Trade areas for groceries in the county were
created using Esri’s Business Analyst, and the market areas for multiple groceries located
in varied settings throughout the county and city of Worcester, were defined. The
complete workflow is summarized in Figure 6 and the various sections that follow
describe the geographic setting, data sources, and methods used.

Figure 6: Project workflow
The objective of the case study was to create a geographic market area for a given
grocery store market using two different techniques, and to compare the resulting market
share calculations under each method for a particular set of retail stores. Under the
boundary method, store trade areas were created independently of the market area, which
was defined as the town boundary surrounding the stores. Under the new and innovative
degrees of separation approach, store trade areas were categorized into “degrees of
separation” from the target store’s trade area, based on whether or not they shared a
boundary with the trade area of the target store. First-degree customers were defined as
those within the target store’s trade area; second-degree customers were those in trade

26

areas bordering the target store’s trade area; and third-degree customers were defined as
those in trade areas bordering the trade areas of second-degree customers. The entire
market area for a target store was then defined as the space occupied by first, second, and
third-degree customers. Figure 7 summarizes the progression of steps taken to determine
market area under this new degrees of separation approach.

Figure 7: The progression of spatial analysis tasks completed to determine a market area
using a degrees of separation approach. Stores (upper left) are assigned individual trade
areas (upper right); stores with trade areas bordering the trade area of the target store are
categorized into “degrees of separation” from the target store (bottom right); finally, the
space occupied by each group of trade areas is considered the market area for the target
store (bottom left).

3.1 Description of Study Area
The experiments focused on three cities in Worcester County, Massachusetts. As a
medium-sized county with several small urban centers, and suburban and rural areas at its

27

edges, it provided a mix of built environments. It also has a mix of large, national
supermarket chains, some regional grocery chains and many independently-owned
markets, groceries and bodegas.
Worcester County covers approximately 1,513 square miles, and had a population
of 798,552 as of the 2010 US Census. The county extends the full length of the state, and
is bordered on the north by New Hampshire, and on the south by Connecticut and Rhode
Island (Figure 8).

Figure 8: Map of Massachusetts and surrounding states, with Worcester County
highlighted

28

While the choice of Worcester County offered advantages in terms of the number
and diversity of grocery stores it provided for analysis, the unique character of the county
and town places it on the declining socio-economic side of the average American county
and town. This warrants special consideration of the impacts to the study which may arise
from the particular socio-economic conditions in the area. Being an older northeastern
town, Worcester has a denser physical infrastructure, making travel over short distances
more time-consuming, but more heavily populated with retail establishments than what
can be found in other parts of the country, particularly the sprawling American Midwest
and West. Being more economically depressed than other regions of the country may also
lead to consumer behavior outside “the norm” for American grocery shoppers. Lower
levels of disposable income may lead to greater price sensitivity and higher willingness to
travel for lower prices on grocery items; conversely, the same limitations in spending
may lead to a dependence upon proximity for purchasing decisions, to reduce travel
costs. Urbanized northeastern towns like Worcester tend to be more walkable than cities
in other parts of the country, which may also skew the behavior of customers towards
shopping based on proximity and access relative to more automobile-dependent areas.

3.2 Geographic Data Sources
Data for this project was collected primarily through Esri’s Business Analyst Data
Repository, which contains a database of business listings and extensive demographic
data, as well as shapefiles for boundary features such as counties, ZIP codes and roads.

29

3.2.1 Grocery data
“Groceries” in this analysis were defined as businesses with a 2012 North American
Industry Classification System code (NAICS) of 445110, representing “commissaries,
primarily groceries”, “delicatessens primarily retailing a range of meats”, “Food (i.e.
groceries) stores”, “grocery stores” and “supermarkets”.
Data about each grocery store was collected by Infogroup, and accessed through
Esri’s Business Analyst. The dataset was current through 2012. To verify that the
business listings were up-to-date, all groceries listed as being in the city of Worcester
were either phoned to verify that they were still open for business, and self-identified as a
place to buy groceries, or in the case of chain retailers, company websites were visited
and business listings from Infogroup were cross-checked with the locations listed on the
web page. This additional verification step resulted in the dataset being reduced from 171
to 150 listings located in the county.
For those locations which were phoned, the question, “Do you identify as a
grocery store, bodega, specialty foods store, or convenience store?” was asked, and if the
employee responded affirmatively to being a grocery store, bodega, or specialty foods
store, their business listing was retained in the study dataset. If the person answering the
phone indicated that the number called was no longer in business, or if an employee
identified the business as a convenience store, the listing was removed from the study
dataset. If the employee was unsure how to respond to the question, they were asked the
explanatory question of, “do customers mostly shop with you for basic foods like milk
and bread for their weekly meals?”. If the response was affirmative, the record was
retained.

30

Of the 21 records which were removed from the dataset, nine were found to be
closed, four self-identified as convenience stores, three identified as clothing stores, one
identified as a restaurant, one was discovered to be a duplicate listing for the same store,
one self-identified as “not a food store”, and two listings could not be reached by phone
after three attempts at various times of the week.
Six grocery stores in Worcester County were chosen as “targets” to be
investigated more closely, and for which market areas would be formed. Four were
located within the densely urban Worcester city boundary; one was located in suburban
Leominster, and one was located in suburban Northbridge, MA. Table 2 profiles each city
chosen as part of the study, and Table 3 profiles the six target groceries.

Table 2: Demographic profile of Worcester County cities chosen for market area
delineation study
Town
Area
(mi
2
)
Population
(2010 Census)
Population
Density (per mi
2
,
2010 Census)
Median household
income (2010
Census)
Northbridge 18.1 15,707 868 $50,457
Leominster 29.8 40,759 1,400 $44,893
Worcester 38.6 181,045 4,678 $61,212

31

Table 3: List of target groceries
Grocery Name Town
Price Chopper
Worcester
Shaws
Super Stop & Shop
Colemo’s Market
Hannaford’s Leominster
Shaw’s Northbridge

3.2.2 Demographic data
Demographic data provided by Esri’s Business Analyst database includes estimates for
“food at home” expenditures by block group. The “food at home” data combines
consumer expenditure diary-keeping surveys and quarterly expenditure interviews from
the Bureau of Labor Statistics’ Consumer Expenditure Surveys (CEX). The data is
derived from survey respondents and interviewees either keeping a detailed journal of
daily expenses, or by estimating a monthly total during a quarterly interview for food
eaten at home. In 2011, the Bureau of Labor Statistics cited that:
…the weighted calendar period estimated mean expenditure for total food by all
consumer units in 2011 is $7,143.84. The standard error for this estimate is $50.30. A
95 percent confidence interval can be constructed around this estimate, bounded by
values 1.96 times the standard error less than and greater than the estimate, that is,
from $7,045.25to $7,242.43. We could conclude with 95 percent confidence that the

32

true population mean expenditure for food for all consumer units in 2011 lies within
the interval $7,045.25 to $7,242.43.
This level of confidence by the Bureau of Labor Statistics was an acceptable threshold for
this project.

3.3 Calculation of Trade Areas
Trade areas were calculated for all groceries in Worcester County, using the trade area
creation tools available in Esri’s Business Analyst. For this project, the Equal
Competition (Thiessen polygon) method was used to create the trade areas. Since the
focus of the project is on the creation of a market area boundary, rather than the creation
of individual trade areas, this method was chosen for its relatively simple input
requirements. It is worth noting again, however, that the degrees of separation approach
is intended to function similarly, regardless of the exact method chosen for trade area
creation.
The input required to create trade areas using the Equal Competition method were
grocery store locations in a points shapefile layer. The Trade Area Creation window
required the identification of this layer, and a unique identifier for each feature in the
layer (Figure 9).
As described in Table 1, the Equal Competition (Thiessen polygon) method
functions by creating polygons around points on a plane, such that any point within the
created polygon is closer to the point at its center than any other point on the plane, in
terms of Euclidian distance.

33

Figure 9: Esri’s Business Analyst Equal Competition Trade Area Creation wizard window
in which a groceries layer is identified, along with a unique identifier for each grocery
location; selecting “All stores” creates trade areas for each feature in the groceries layer

3.4 Calculation of Market Areas
Market areas were calculated for each of the six target stores using two different
methods. First, market areas were calculated using an administrative boundary as the
assumed market area boundary. For the target stores, the city boundary for the city in
which they reside was used as the market boundary. Shapefiles for the city boundaries
were obtained from the MassGIS website, a repository of GIS data and shapefiles
maintained by the state government of Massachusetts.

34

Market areas were next calculated for each of the target stores using the degrees
of separation approach that was proposed in Chapter 1. The objective was to form a
market area composed of each store’s individual trade area, the trade areas of its direct
competitors, and the trade areas of its indirect competitors. To identify the various
degrees of separation, the SELECT command in ArcMap was used to progressively
select each new set of customer degrees of separation. First, the trade area of the target
store was identified, and saved as its own layer. Second, a new selection was created in
ArcMap which selected those trade areas which were adjacent to the target store’s trade
area (thereby selecting all second-degree customers). The BOUNDARY_TOUCHES
command under the Select By Location tool was used to execute the selection, although
INTERSECT would have achieved the same result. This new selection was saved as a
layer, named according to its status as the set of second-degree customers belonging to
the target store. Next, a new selection was created which selected all trade areas adjacent
to the second-degree customers layer. This operation was carried out with the same
BOUNDARY_TOUCHES command that was used for the selection of second-degree
customers. The additional step of removing the target store’s own trade area (i.e. first-
degree customers) was needed here, because the first-degree customers are also adjacent
to the second-degree customers. This was accomplished manually, by removing the first-
degree customers polygon from the selection with the REMOVE FROM SELECTION
interactive selection option. Then, this new selection was saved as its own layer, named
according to its target store association and selection of third-degree customers.
Once the first-, second-, and third-degree customers were saved in their own
layers for a given target store, the three layers were combined, to account for the entire

35

market area of the target store. This was executed by selecting the entire contents of each
layer, and saving the selection as a new layer. It would also have been feasible to use the
DISSOLVE command to create a new market area layer consisting of a single polygon.

3.5 Associating Expenditure Data with Trade and Market Areas
The trade and market areas as originally created, did not contain any attribute information
other than their own location, shape and size. To carry out market share calculations
using the trade and market areas, it was necessary to associate Food at Home expenditure
data (stored in the block group polygon layer) with each trade area. However, because the
block group polygons and trade area polygons did not match in size, and had many
overlapping boundaries, a method was needed to proportionally associate Food at Home
data with the trade area polygons overlaying the block groups. Business Analyst’s
Append Data tool, found under the Analysis toolset, was used for this task. Its function is
to overlay two shapefile layers, in this case, two polygon layers, and transfer attribute
data from one layer to the other. It is sophisticated enough to apportion quantities in the
attribute table of one dataset according to how much overlap exists between it and the
other layer. This apportionment can be accomplished in one of two ways. First, the
apportionment could be based solely on the proportion of overlapping space between the
two layers. For example, if exactly half of one block group overlapped a trade area, then
exactly half of the Food at Home expenditures in that block group would be added to the
new Food at Home expenditure attribute in the trade area attribute table. The second
method for apportionment makes use of Business Analyst’s awareness of the locations of
individual block centroids within each block group. Blocks within a block group may

36

have centroids which are unevenly distributed across a block group, and the Append tool
can use this information to apportion more realistically. For example, if exactly half of a
block group’s total area overlapped a given trade area, but 75% of its block centroids
were contained in that overlap, the tool would apportion 75% of the block group’s Food
at Home expenditures to the trade area it overlapped. This second method was used for
this project, as it mirrors population density and from this vantage point is probably more
reflective of reality.
Using block centroid apportionment, each trade area in Worcester County was
given a new attribute of total Food at Home Expenditures. This information was used to
calculate market shares, as described in the next section.

3.6 Calculation of Market Share
The goal of creating trade and market areas for each target store was to ultimately
calculate the market share of each store under both market area delineation techniques.
The basic market share calculation formula is:

Market Share = Sales captured by target store * 100 (1)
Total sales in market area

where Market Share is reported in percent. Specifying this process in terms of this project,
the formula used to calculate the market share using city boundaries as market areas, was:

Market Share = Food at home spending captured by target store * 100(2)

37

Total sales in market area

To calculate market share under the degrees of separation method, the formula was:

Market Share = Food at home spending captured by target store * 100(3)
Total food at home expenditures captured by
first-, second-, and third-degree customers

3.7 Evaluation of Model Results
The case study was purposefully structured such that the only data used in the experiment
is readily available from Esri’s Business Analyst data repository and the US Census
Bureau. Market area definitions inherently involve assumptions about the competitors in
a market, and primary data on competitor sales and performance are typically not
available to business managers wishing to model trade area scenarios. For this reason, it
was deemed most relevant to use data that any manager could obtain with relatively little
investment.
Given those parameters, the market shares of the target stores were calculated and
compared to one another to assess the level of sensitivity of a market share calculation to
the spatial character of the market area, or market share denominator. The biggest
inconsistency coming out of the formation of the trade and market areas was the
conundrum posed by first-degree customers in a target store’s trade area which
overlapped political boundaries. This occurred for three of the six target stores, and posed
an interesting question with respect to the treatment of similar scenarios. When some of a
target store’s primary customers reside outside the politically-based market boundary,
how are those customers to be treated? As noted in the discussion of the Kitchener-

38

Waterloo problem in Chapter 2, one solution which has been used in the past, is to simply
remove the portion of the trade area located outside the political boundary. This action
would have less impact if it were not the target store’s primary customer base being
altered. If the customers being removed were third- or fourth-degree customers, their loss
would be arguably less critical. However, removal of any customers from a trade area,
merely for the sake of aligning with a political boundary, is highly questionable. While
such a scenario may not occur in all situations, it highlights again the motivation that
guided the choice of this thesis topic – there is a need for a more thoughtful approach to
market area definition.
For the sake of consistency, however, the target store trade areas which spilled
over a political boundary were clipped to fit within city limits for the political boundary-
based market area calculations. Otherwise, market share calculations would have
displayed drastic inconsistencies.

39

CHAPTER FOUR: RESULTS[JW1][JW2][JW3]

Figure 10: Groceries in the Worcester County area shown with Massachusetts county
boundaries; Worcester County outlined in blue; groceries were identified for surrounding
states to allow for trade area creation across the Worcester County border.

The distribution of groceries in Worcester County is summarized in Figure 10. There are
150 total groceries, ranging from large, national supermarkets like Shaw’s, to regional

40

chains like Market Basket, and Hannafords, to local stores like D’Errico’s Market.
Groceries for the entire state of Massachusetts, New Hampshire, Vermont, Connecticut,
and Rhode Island were collected, since Worcester county borders those areas. This
inclusion of the groceries at the border of the county allowed the creation of more
realistic trade areas, since spillovers across boundaries are inevitable.
The resulting Thiessen polygon trade areas for groceries located in or near to
Worcester County are shown in Figure 11.
Figure 11: Thiessen polygon trade areas created for groceries in or near Worcester
County; the city boundaries of Worcester, Leominster and Northbridge are shown in
pink, with Worcester in the center, Leominster to the north, and Northbridge to the south.

Six groceries were chosen as “targets” to illustrate the various methods of market
area delineation. Four were within the densely urban Worcester city boundary; one was in
suburban Leominster, and one was in suburban Northbridge, MA. Figure 12 shows

41

the trade areas created around each target grocery store. Four of the six target stores had
trade areas which overlapped the relevant city boundary.

Figure 12: Views of the trade areas created around each target store (represented by a
red star) for each town: (a) Price Chopper, Shaw’s, Super Stop & Shop, and Colemo’s
Market groceries in the City of Worcester; (b) Hannafords grocery in Leominster; and
(c) Shaw’s in Northbridge (next page).

42

Figure 12: Continued

43

The market areas for each grocery were defined in two separate ways. First, the
political boundary of the town in which the grocery resided was used as the market
extent. Second, the market areas were defined by identifying the trade areas of the direct
and indirect competitors of the grocery, as defined by the degrees of separation method.
When defining market areas based on the boundary of a town, the overlap created
by many of the target store’s trade areas were resolved by clipping the trade areas by the
town boundaries, and reapportioning food at home expenditures. This kept market share
calculations consistent and prevented expenditures from outside the town boundary from
factoring into the market share calculations. Figure 13 shows the clipped portions of the
trade areas for the target stores.

Figure 13: The trade areas for the Shaw’s and Super Stop & Shop in Worcester (a) were
clipped for the market share analysis using the Worcester town line as a market area
boundary. The portions of the trade areas that were retained are outlined in red. The trade
areas for the Hannaford’s in Leominster (b) and the Shaw’s in Northbridge (c) were also
clipped.

44

Figure 13: Continued

45

Once all trade areas were clipped by the town boundary they resided within, the
food at home expenditures within the clipped trade areas of each target store were
calculated and divided by the total food at home expenditures for each town to calculate
market share using Equation (2).
The results of using the political boundary market definition method to estimate
the market share for each target store are summarized in Table 4. The third and fourth
columns in the table represent the total food at home expenditures contained in the trade
areas of the target stores, first unclipped, and second, clipped by the political boundary
such that only those records inside the political boundary were counted as part of the
target store’s trade area. The fifth column in Table 4 represents the denominator for the
market share calculation, or the total food at home expenditures contained within the city
boundary. The sixth column calculates market share by dividing the target store’s clipped
food at home expenditures into the town’s total food at home expenditures, and
represents market share in percent. For context, the final column notes the total number
of groceries inside the relevant town boundary.
For the degrees of separation approach, first degree customers were identified by
the location of the target store’s trade area. For second-degree customers, a new layer was
created containing all those trade areas which touched the target store’s trade area. Third-
degree customers were identified in a new layer containing all those trade areas which
adjoined the trade areas of the second-degree customers, as was explained in Chapter 3.
In Figures 14 through 19, first- and second-degree customers for each target store are
shown together in one color, and third-degree customers are shown in a separate color.

46

Table 4: Market share calculation results using political boundaries as market areas
Town
Target Store
Name
Target store
food at home
expenditure
(unclipped)
Target store
food at home
expenditure
(clipped)
Total town
food at home
expenditure
Market
share of
target
store
(clipped
where
needed)
Total
number of
groceries
in town
Worcester
Price
Chopper
9,337,804 n/a
280,963,154
3.32%
37
Shaws 23,141,849 17,329,772 6.17%
Super Stop &
Shop
10,022,388 9,908,091 3.53%
Colemo’s
Market
2,212,495 n/a 0.79%
Leominster Hannaford’s 22,317,781 21,911,428 74,076,162 29.58% 8
Northbridge Shaw’s 15,472,076 12,931,965 29,190,462 44.30% 3

As seen in Figure 14, the Price Chopper’s trade area in Worcester is
contained entirely within the Worcester city limits, but some of the store’s direct and
indirect competitors are drawing revenue from outside the city. Also, several of the
groceries in the southeastern part of the city are not considered part of the competitive
market under the degrees of separation method. The Worcester Shaw’s location’s first-
and second-degree customers actually spill over the city boundary (Figure 15), and much
of the eastern half of the town is not considered part of the store’s competitive market.
Similar to the Worcester Shaw’s location, the Super Stop & Shop’s first- and
second-degree customers spilled over the town limits (Figure 16). The store’s position in
the northwestern portion of the Worcester town limits means that, under the degrees of
separation method, nearly half the groceries in the city – those in the southern half of the
city – are no longer considered part of the Super Stop & Shop’s market area.

47

Figure 14: Degrees of separation market area for a Price Chopper in Worcester,
Massachusetts

Figure 15: Degrees of separation market definition for Shaw’s in Worcester,
Massachusetts.

48

Figure 16: Degrees of separation market definition for the Super Stop & Shop in
Worcester, Massachusetts.

Located in the urban center of the city, the Worcester Colemo’s first- and second-
degree customers cover a smaller area than those of the other target Worcester stores
(Figure 17). Colemo’s location in a dense downtown area highlights the weakness of
defining markets based on only two or three degrees of separation. It would be reasonable
to assume that the non-competitive stores to Colemo’s north, also in the downtown
Worcester area, are in fact active competitors.

49

Figure 17: Degrees of separation market definition for Colemo’s Market in Worcester,
Massachusetts.

The Leominster Hannaford’s location had a trade area with portions outside the
city boundary, and was located in a tight cluster of groceries near the town center (Figure
18a). The dense grouping of groceries around the Hannaford’s store renders the strict
definition of degrees of separation less relevant than if the stores were more evenly
distributed (Figure 18b). For a more accurate view in dense areas, there may be minimum
distance thresholds inside which neighboring groceries should be considered direct
competitors.

50

Figure 18: Degrees of separation market definition for Hannaford’s in Leominster: (a)
Map showing full solution; and (b) large-scale map showing partial solution

51

The customer degrees shown in Figure 19 for the market area of the Northbridge
Shaw’s are an example of a more dispersed grocery network resulting in a market area
which is much larger than the administrative boundary in which the target store sits. If the
Shaw’s location was considered only in the context of the other two groceries in the
town, the competitive impact from the clusters of stores to the east and south would be
completely ignored.

Figure 19: Degrees of separation market definition for Shaw’s in Northbridge,
Massachusetts.

The market shares for the target stores under the degree of separation market
definition method are summarized in Table 5. Note that no clipping of trade areas was
required in this analysis. The Worcester stores resulted in a range of market shares, from

52

3% for the smaller Colemo’s Market, to 8% for Shaw’s supermarket. The Hannaford’s in
Leominster held nearly 10% market share against its seven other competitors in the city.
The Northbridge Shaw’s earned nearly 4% market share in its area. Table 4 summarized
the market share calculation results under the political boundary approach. The Worcester
groceries had a range of 1-6% market share using the political boundary market
definition, while the Leominster and Northbridge groceries held 30% and 44% market
shares, respectively.
Table 5: Market share calculations for target stores using degrees of separation method for
market definition
Town
Target Store
Name
Target store
food at home
expenditure
Food at home
expenditure of
first-, second-,
and third-
degree
customers
Market
share of
target
store
Total
number of
groceries
in town
Worcester
Price Chopper 9,337,804 227,101,434 4.11%
37
Shaws 23,141,849 288,851,577 8.01%
Super Stop &
Shop
10,022,388 257,594,485 3.89%
Colemo’s
Market
2,212,495 86,101,290 2.57%
Leominster Hannaford’s 22,317,781 225,112,570 9.91% 8
Northbridge Shaw’s 15,472,076 428,478,709 3.61% 3

Market share calculations for both methods are compared in Table 6. For the
Worcester stores, small differences exist between the market share results of the two
methods. All Worcester groceries received a higher market share under the degrees of
separation method; the Price Chopper and Super Stop & Shop locations gained less than
one percentage point in market share, while the Shaw’s and Colemo’s Market groceries
each gained nearly two percentage points.

53

Table 6: Comparison of market share calculations between boundary and degrees of
separation methods for delineating market areas

Town
Target Store
Name
Market share (boundary
method)
Market share (degrees of
separation method)
Worcester
Price Chopper 3.32% 4.11%
Shaws 6.17% 8.01%
Super Stop &
Shop
3.53% 3.89%
Colemo’s
Market
0.79% 2.57%
Leominster Hannaford’s 29.58% 9.91%
Northbridge Shaw’s 44.30% 3.61%

The differences between the two methods are much larger in the less-dense
neighborhoods of Leominster and Northbridge, with both groceries experiencing a
decrease in market share under the degrees of separation approach. The Hannaford’s in
Leominster declined from a 30% market share using political boundaries, to a 10%
market share using degrees of separation. Likewise, the Shaw’s in Northbridge lost 40
market share percentage points, moving from 44% under the political boundaries method,
to 4% using the degrees of separation method.

54

CHAPTER FIVE: DISCUSSION AND CONCLUSIONS
While the calculations undertaken in this project point toward the sensitivity of market
share calculations to market area definitions, and begin to suggest alternative methods for
delineating market areas, the data used here would be more valuable were it validated
through primary research into actual grocery store sales for a competing set of groceries.
That said, the project results demonstrated the sensitivity of the market area estimates to
the delineation approaches that are chosen, and they point to ways to refine the degrees of
separation-like framework that was the focus of this thesis for market area delineation.

5.1 Broader Significance of the Worcester County Results
As seen in Table 6, the absolute differences in market share between the two methods are
relatively small for groceries in the town of Worcester. However, the results for the four
groceries in the Town of Worcester do show that the degrees of separation method tended
to estimate a higher market share for urban area groceries compared to the boundary
method (Table 6). This could be due to the high number of groceries inside the city
boundary. The degrees of separation method resulted in a smaller set of groceries with
which to compete than in the boundary method, so resulting market shares were generally
higher. Figure 20 illustrates the absolute differences in overall market size for each
grocery under each method. For three of the four urban groceries, the market sizes were
relatively similar under both approaches, but wide variation can be seen in the rural
markets. The Hannafords market in rural/suburban Leominster, for example, was three
times larger under the degrees of separation approach than the boundary method.

55

Figure 20: Comparison of results from two market area definition methods; dollars in
food at home expenditure per year on Y-axis, by-store results under boundary and
degrees of separation method shown on X-axis

The Worcester stores also illustrated where the degrees of separation method can
break down and therefore be less useful than in situations with a wider dispersion of
competitors. Figure 21 shows a close up view of multiple downtown Worcester grocery
stores relative to the target Shaw’s location. Many of the urban groceries are excluded as
competitors, yet other groceries north of the city, whose consumers would presumably
drive through downtown Worcester to reach the Shaw’s location, are considered more of
a competitive threat than the stores located (more or less) between them and Shaw’s.
A similar situation is seen in the Leominster example, as reproduced in Figure 22.
Groceries, in this instance, that are located within a mile of the target store are considered

56

only “indirect” competitors, a function of the model’s strict definition of what constitutes
a degree of separation.

Figure 21: Closer view of the Shaw’s location in Worcester. The cluster of downtown
groceries complicate the degrees of separation approach by treating different distances
between competitors as effectively equal in their impact on competitive threat.

The degrees of separation model might be improved by adding minimum distance
thresholds between stores as grounds for automatic direct competition. Additionally, the
fact that the examples provided used non-overlapping trade areas likely oversimplifies
the extent of competitive areas. Had overlap been included in the trade area creation
method, it is likely that clusters of close stores as in Figures 21 and 22 would have all

57

been considered direct competitors. These ideas are taken up in more detail in Section 5.2
below.

Figure 22: Closer view of the Hannaford’s market area in Leominster.
Another potential improvement in the degrees of separation method could be to
consider different thresholds for degrees of separation according to the density of the
competitive environment. For example, if instead of a minimum distance threshold, a
maximum threshold of four or five degrees of separation were considered in urban areas,
and perhaps only two degrees in sprawling rural areas, the nuances of different
competitive densities might be handled better.
In contrast to the Worcester examples, the results of the Leominster and
Northbridge analyses appear to result in more sensible trade areas through the degrees of
separation method. In Northbridge, a town with only three groceries within its limits, it

58

becomes clear that political boundaries can be quite misleading. The trade areas of all
three groceries extended well beyond the city limits. Using the degrees of separation
method, it appears that groceries nearly within the Worcester city limits are actually
competing indirectly with the Northbridge Shaw’s. The major highway linking
Northbridge and Worcester likely adds to this competition. Furthermore, with an initial
calculation of 40% market share within the city boundary, it would be simple for such a
metric to understate the competitive force of the groceries directly south, and just outside,
the city of Northbridge. It could be that the presence of these stores threatens the Shaw’s
existing market share in the city limits, or conversely, they might represent an
opportunity for additional market share and revenue gain. Either way, their potential
impact could be underestimated with a poorly defined market area.
Using the degrees of separation approach in Leominster, the market area does in
fact cover the city limits, but it also extends south to include groceries like Appletown
Market, Next Door Market and another Hannaford’s location. Also linked directly to
Worcester by highways, the competition extending southward toward the city nearly
intersects the Worcester city limits.
Leominster’s Hannaford’s reached a 30% market share estimate using its city
limit as a market area. While this is positive news for the grocery, and a useful statistic, it
ignores the market potential existing just over the boundary which may already be a part
of the store’s market share, or may be prime for the taking. The three stores directly south
of the Hannaford’s Leominster location lie outside the city limits, but within 10 miles of
Hannaford’s. With a relatively sparse grocery market to the south of Leominster, it is
unlikely that these stores do not vie for similar customers. On the other hand, there is a

59

cluster of stores in the town of Fitchburg 10 miles to the north of the Leominster
Hannaford’s. While a similar distance away, the higher number of competitors in the
Fitchburg micro-market would be less likely to compete with the Hannaford’s to the
south. This demonstrates the utility of using a degrees of separation approach over a
method like a ring definition for delineating market areas. Assuming a simple buffer
distance around stores to delineate a market area, as Huff and McCallum (2008) did,
relies on the seemingly risky assumption that competitive forces act at a constant rate
across space. In the Leominster example, it does not appear that the competitive picture is
the same 10 miles north versus 10 miles south of the target store, and the degrees of
separation method used helped to point out that nuance in a simple but effective way.

5.2 Opportunities to Improve the Degrees of Separation Framework
There are several possible ways to improve upon the degrees of separation method that
was conceived and implemented in this thesis.
First, it would be useful to test the results of the model using validated store and
competitor sales data, as well as the perceptions of managers regarding the spatial extent
of the market in which they operate.
Second, while the degrees of separation framework is intended to be flexible
enough to work with any type of trade area delineation method, it would be beneficial to
test it using methods other than the Equal Competition (Thiessen polygon) approach.
Threshold rings, drive time analyses, and trade areas with overlapping segments would
provide a different set of nuances to test the applicability of the degrees of separation
method. Also, only one type of political boundary was used in this analysis. Town and

60

city limits are easy to relate to, but ZIP codes, census block groups or some other political
boundaries could be used as market boundaries and compared against the degrees of
separation results.
Third, it would be interesting to test the method at a larger scale, perhaps with an
entire regional retail chain, or using a retail sector with low-order goods but high brand
loyalty, such as apparel.
Finally, it would be necessary to test the method in a diverse set of geographic
areas. As mentioned, the particular character of Worcester county as a northeastern,
historically industrial, and socio-economically declining area lends a particular set of
assumptions and practices to the creation of accurate trade areas. A different set of best
practices may exist for creating trade areas in sprawling cities, rural areas, or coastal
towns. While the methods chosen for creating trade areas do not fundamentally alter the
degrees of separation market area delineation framework, it would be interesting to assess
the varying shapes and sizes of markets that result from separate sets of underlying trade
area creation practices.

5.3 Closing Thoughts
The degrees of separation market delineation approach sought to take advantage of the
wealth of research already underway on trade area delineation. There should be no cause
for “re-inventing the wheel” for market area definitions, and certainly no cause for using
arbitrary market definitions based on a different spatial unit than pre-existing trade areas
(such as using census block groups to define trade areas, and then town boundaries to
define a market area). The approach proposed here suggests that three degrees of

61

separation between a target store and customers is appropriate for the grocery market; but
the intent is also to suggest flexibility in how “degrees” are handled on a case-by-case
basis. What works for groceries may not work for gas stations; but a common language
for what separates competitors from one another and ultimately forms a spatial market
extent is just as important as accurately defining an individual store’s trade area. If a
common language and theory can be established around the task of spatially defining
retail markets, the robust methods which support individual trade area delineation may be
applied to market area delineation as well.
Taken as a whole, the results presented in this thesis suggest that we have some
way to go before we can fully exploit location in retail business analytics and economic
decision-making. This discovery points to the need for fundamental geospatial research to
complement the business applications touted by Esri and many others in their business
location analytics promotional materials.

62

REFERENCES
Applebaum, W. 1966. Methods for Determining Store Trade Areas, Market Penetration,
and Potential Sales. Journal of Marketing Research 3:127-141.
Applebaum, W., and S. B. Cohen. 1961. The Dynamics of Store Trading Areas and
Market Equilibrium. Annals of the Association of American Geographers 51:73-101.
Boots, B., and R. South. 1997. Modeling Retail Trade Areas Using Higher-order,
Multiplicatively Weighted Voronoi Diagrams. Journal of Retailing 73:519-536.
Bronnenberg, B., and Albuquerque, P. 2002. Geography and Marketing Strategy in
Consumer Packaged Goods. Advances in Strategic Management 20:215.
Brooks, G. R. 1995. Defining Market Boundaries. Strategic Management Journal
16:535-549.
Bucklin, L. P. 1971. Trade Area Boundaries: Some Issues in Theory and Methodology.
Journal of Marketing Research 8:30-37.
Davies, R. L. 1977. Store Location and Store Assessment Research: The Integration of
Some New and Traditional Techniques. Transactions of the Institute of British
Geographers 2:141-157.
Drezner, T. 2006. Derived Attractiveness of Shopping Malls. IMA J Management Math
17:349-358.

63

———2009. Location of Retail Facilities Under Conditions of Uncertainty. Annals of
Operations Research 167:107-120.
———2011. Cannibalization in a Competitive Environment. International Regional
Science Review 34:306-322.
Fik, T. J. 1988. Spatial Competition and Price Reporting in Retail Food Markets.
Economic Geography 64:29-44.
Fotheringham, A. S. 1988. Consumer Store Choice and Choice Set Definition. Marketing
Science 7:299-310.
Gaynor, M. S., S. A. Kleiner, and W. B. Vogt. 2013. A Structural Approach to Market
Definition With an Application to the Hospital Industry. The Journal of Industrial
Economics 61:243-289.
Huff, D. L., 1963. A Probabilistic Analysis of Shopping Center Trade Areas. Land
Economics 39:81-90.
———1964. Defining and Estimating a Trading Area. Journal of Marketing 28:34-38.
———1984. Measuring the Congruence of Market Areas. Journal of Marketing --(pre-
1986) 48:68.
Huff, D., and B. M. McCallum. 2008. Calibrating the Huff Model Using ArcGIS
Business Analyst. Redlands, CA, Esri White Paper.

64

Ill, H. E. F., J. Langenfeld, and R. F. McCluer. 2004. Elzinga-Hogarty Tests and
Alternative Approaches for Market Share Calculations in Hospital Markets. Antitrust
Law Journal 71:921-947.
MacKay, D. B. 1973. Spatial Measurement of Retail Store Demand. Journal of Marketing
Research 10:447-453.
Nakanishi, M., and L. G. Cooper. 1974. Parameter Estimation for a Multiplicative
Competitive Interaction Model: Least Squares Approach. Journal of Marketing
Research 11:303-311.
Patel, A. R., T. J. Fik, and G. I. Thrall. 2008. Direction Sensitive Wedge-Casting for
Trade Area Delineation. Journal of Real Estate Portfolio Management 14:125-139.
Peterson, R. A. 1974. Trade Area Analysis Using Trend Surface Mapping. Journal of
Marketing Research 11:338-342.
Reilly, W. J. 1931. The Laws of Retail Gravitation. New York: Knickerbocker Press.
Weiss, L. W. 1972. The Geographic Size of Markets in Manufacturing. The Review of
Economics and Statistics 54:245-257.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

Guiding business oriented volunteered geographic information through geotigger services: a case study of CrossFit affiliates

PDF

Evaluating spatial changes in the rate of insurgency‐violence in Central Africa: the Lord's Resistance Army 2008-2012

PDF

Explorations of American churchscape diversity

PDF

Does the Bay Area have a social center? Delimiting the postmodern urban center of the San Francisco Bay Area

PDF

Spatial analysis of urban built environments and vehicle transit behavior

PDF

Geographic information systems and marketing: a transdisciplinary approach to curriculum development

PDF

Development of a Web GIS for urban sustainability indicators of Oakland, California

PDF

Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

PDF

Predictive habitat distribution modeling of sperm whale (Physeter macrocephalus) within the central Gulf of Alaska utilizing passive acoustic monitoring

PDF

Implementing spatial thinking with Web GIS in the non-profit sector: a case study of ArcGIS Online in the Pacific Symphony

PDF

Visualizing email response data to improve marketing campaigns

PDF

The role of precision in spatial narratives: using a modified discourse quality index to measure the quality of deliberative spatial data

PDF

Modeling the spatio-temporal variability of solar radiation on buildings: a case study of Lewis Hall

PDF

Social media to locate urban displacement: assessing the risk of displacement using volunteered geographic information in the city of Los Angeles

PDF

Semi-automated visualization of spatial information in unstructured text

PDF

A fire insurance map geocoder for pre-earthquake San Francisco

PDF

Spatial narratives of struggle and activism in the Del Amo and Montrose Superfund cleanups: a community-engaged Web GIS story map

PDF

Filling in the gaps: 3D mapping Arizona's Basin and Range aquifer in the Prescott Active Management Area

PDF

Spread global, start local: modeling endemic socio-spatial influence networks

Asset Metadata

Creator Gray, Amanda Christine (author)

Core Title Spatial delineation of market areas: a proposed approach

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 03/31/2014

Defense Date 11/13/2013

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

Tag GIS,gravity model,Huff,market delineation,OAI-PMH Harvest,spatial,trade area

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wilson, John P. (committee chair), Paganelli, Flora (committee member), Warshawsky, Daniel N. (committee member)

Creator Email amandacg@usc.edu,mandycgray@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-369996

Unique identifier UC11296624

Identifier etd-GrayAmanda-2305.pdf (filename),usctheses-c3-369996 (legacy record id)

Legacy Identifier etd-GrayAmanda-2305.pdf

Dmrecord 369996

Document Type Thesis

Format application/pdf (imt)

Rights Gray, Amanda Christine

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