Retail Location Analysis:

A Case Study of Burger King & McDonald’s in Portage & Summit Counties, Ohio

A thesis submitted to the College of Arts

of Kent State University in partial fulfillment of the
requirements for the degree of
Masters of Arts

by

Niti Duggal

December, 2007
 Thesis written by
Niti Duggal

B.A. (Hons), University of Delhi, India 1996

M.A., Jawaharlal Nehru University, New Delhi, India 1998
MPhil, Jawaharlal Nehru University, New Delhi, India 2001
M.A., Kent State University, 2007

Approved by

____________________________________, Advisor
Dr. Jay Lee
____________________________________, Chair, Department of Geography
Dr. Jay Lee
____________________________________, Dean, College of Arts and Sciences
Dr. Jerry Feezel

ii
 Table of Contents

Table of Contents ……………………………………………………………….. iii

List of Maps and Figures …………………………………………….………….. v

List of Tables …………………………………………………………………… viii

Acknowledgments ……………………………………………………………… ix

Chapter

1: Introduction …………………………………………………………..…… 1

1.1 Research Objectives ............................................................................ 2

1.2 Summary …………………………………………………..…..……. 4

2: Problem Statements ………………………………………………...……. 6

2.1 Size and Shape of the Retail Trade Area………….……………....…. 6

2.2 Summary………………………………………..………………...….. 9

3: Literature Review ………………………………………………………… 11

3.1 GIS for Business and service Sector Planning ……………………….11

3.2 GIS as a Tool for Retail Location Decisions……………………….... 12
3.3 GIS Methodologies for Retail Location Studies…………………...… 13
3.4 Analysis of Trade Areas………………………………………………19
3.4.1 Simple or Basic Methods of Trade Area Analysis .…………… 19
3.4.2 Gravitational Methods for Trade Area Analysis ……………… 20
3.5 Forecasting the Fast Food Restaurant Sales ………………………… 21
3.6 Identifying the Trading Area of the Fast Food Restaurants ……………….. 26
3.7 Retail Marketing Strategies …………………………………………. 29
3.9 Eating Facilities at Fast Food Restaurants ………………………….. 31
3.9 Pricing Strategies …………………………………………………… 32
3.10 Summary …………………………………………………………... 33

iii
 4: Data and Research Methodology ………………………………….. 35

4.1 Data Preparation ……………………………………………… 36

4.2 Research Methodology ………………………………………. 39
4.2.1 Geocoding …………………………………………………. 39
4.2.2 Catchment Area Analysis …………………………………. 47
4.2.3 Regression analysis ………………………………………… 59
4.3 Study Area ……………………………………………………. 66
4.4 Summary ……………………………………………………... 67

5: Analysis and Discussions …………………..……………………… 69

5.1 Regression Analysis: Enter Method ………………………….. 76

5.2 Regression analysis: Stepwise method ……………………….. 78
5.3 Summary ……………………………………………………... 88

6: Conclusion …………………………………………………………. 90

Appendix: Maps and Figures …………………………………………………. 97

References ……………………………………….…………………………… 130

iv
 List of Maps and Figures:

Chart 4.1: Research Methodology ………………………………………………….. 38

Figure: 1 Burger Kings and McDonalds’ QSR: Portage and Summit Counties ……. 41

Figure 2: Burger Kings and McDonalds’- Portage County ………………………… 42

Figure 3: Burger King and McDonald’s- Summit County …………………………. 43

Figure 4: Distance of Burger King & McDonald’s Restaurants from Major Roads 46

Figure: 5 Burger King and McDonald’s-Portage and Summit Counties

(Thiessen Polygons) ………………………………………………………………… 50

Figure 14: Burger King & McDonald’s-Portage & Summit Counties (Buffer 1-mile) 55

Figure 15: Burger King & McDonald’s Portage & Summit Counties (Buffer 2 miles) 56

Figure 16: Burger King & McDonald’s Portage & Summit Counties (Buffer 5 miles) 57

Figure: 6 Burger King and McDonald’s in Portage County (Thiessen Polygons) …… 98

Figure: 7 Burger King and McDonald’s in Summit County (Thiessen Polygon) ……. 99

Figure: 8 Burger Kings in Portage County (Thiessen Polygons) ……….………….. 100

Figure: 9 Burger Kings in Summit County (Thiessen Polygons) …………………… 101

Figure: 10 Burger King in Portage and Summit Counties(Thiessen Polygons) .….… 102

Figure 11: McDonald’s in Portage County (Thiessen Polygons) …………………… 103

Figure 12: McDonald’s in Summit County (Thiessen Polygons) …………………… 104

Figure 13: McDonald’s in Portage and Summit Counties (Thiessen Polygons) ….… 105

Figure 17: Burger King & McDonald’s in Portage County (Buffer 1 mile) ………... 106

v
Figure 18: Burger King & McDonald’s in Portage County (Buffer 2 miles) ………. 107

Figure 19: Burger King & McDonald’s in Portage County (Buffer 5 miles) ………. 108

Figure 20: Burger King & McDonald’s in Summit County (Buffer 1 mile) ……….. 109

Figure 21: Burger King & McDonald’s in Summit County (Buffer 2 miles) ………. 110

Figure 22: Burger King & McDonald’s in Summit County (Buffer 5 miles) ………. 111

Figure 23: Burger King in Portage County (Buffer 1 mile) ………………………… 112

Figure 24: Burger Kings in Portage County (Buffer 2 miles) ……………………… 113

Figure 25: Burger Kings in Portage County (Buffer 5 miles) ………………………. 114

Figure 26: Burger Kings in Summit County (Buffer 1 mile) ……………………….. 115

Figure 27: Burger Kings in Summit County (Buffer 2 miles) ………………………. 116

Figure 28: Burger Kings in Summit County (Buffer 5 miles) ………………………. 117

Figure 29: Burger Kings in Portage & Summit Counties (Buffer 1 mile) ………….. 118

Figure 30: Burger Kings in Portage & Summit Counties (Buffer 2 miles) …………. 119

Figure 31: Burger Kings in Portage & Summit Counties (Buffer 5 miles) …………. 120

Figure 32: McDonalds’ in Portage County (Buffer 1 mile) ………………………… 121

Figure 33: McDonalds’ in Portage County (Buffer 2 miles) ………………………... 122

Figure 34: McDonalds’ in Portage County (Buffer 5 miles) ………………………... 123

Figure 35: McDonalds’ in Summit County (Buffer 1 mile) ………………………... 124

Figure 36: McDonalds’ in Summit County (Buffer 2 miles) ………………………. 125

Figure 37: McDonalds’ in Summit County (Buffer 5 miles) ………………………. 126

Figure 38: McDonalds’ in Portage & Summit Counties (Buffer 1 mile) …………... 127

Figure 39: McDonalds’ in Portage & Summit Counties (Buffer 2 miles) ………….. 128

vi
Figure 40: McDonalds’ in Portage & Summit Counties (Buffer 5 miles) ………….. 129

vii
 List of Tables:

Table 1: Distance of Burger King & McDonald’s QSR from Major Roads ……….. 45

Table2: Spatial Configuration (Regression Analysis: Enter Method ………………. 70

Table 3: Restaurant wise (Regression Analysis: Enter Method) …………………… 71

Table 4: County wise (Regression Analysis: Enter Method) ……………………… 72

Table 5: Restaurant wise and County wise (Regression Analysis:Enter Method) … 73

Table 6: Spatial Configuration (Stepwise Regression) ……………………………. 80

Table 7: Restaurant wise (Stepwise Regression) …………………………………. 81

Table 8: County wise (Stepwise Regression) ……………………………………… 82

Table 9: Restaurant wise and County wise (Stepwise Regression …………………. 83

Table 10: Spatial Configuration (Stepwise Regression) ……………………………. 84

Table 11: Restaurant wise (Stepwise Regression) …………………………………. 85

Table 12: County wise (Stepwise Regression) ………………………………..……. 86

Table 13: Restaurant wise and County wise (Stepwise Regression) ………………. 87

viii
 Acknowledgements

I would like to gratefully acknowledge the enthusiastic supervision of Dr. Jay Lee during

this work. His seemingly endless enthusiasm and constant support helped me thru-out my

program at Kent State University. I thank my other committee members Dr. Shawn

Banasick and Dr. Chuanrong Zhang for their valuable comments and suggestions. I thank

Dr. Milton Harvey and Mrs. Mary Lou Church for their affection and concerns. I am

grateful to them and all my friends at McGilvrey Hall, for being the surrogate family

during my years at Kent and their continued moral support thereafter. I thank Dr. Munro-

Stasiuk, Dr. Schmidlin, Dr. Sheridan, Dr. Kaplan, Dr. Haley, Dr. Dymon, Dr. Bhardwaj

and other faculty members in the Department of Geography for making the atmosphere in

the department stimulating for research and academics.

The Kent State University Library staffs are acknowledged for their efficiency

and availability. A particular thanks to Edith Scarletto, Head of the Map Library, who

helped me, gather the initial data required for the research.

I would like to thank my friend Sathy for his help in formatting the entire text.

I am forever indebted to my family, for their blessings and love and who have

supported and encouraged me to do my best in all matters of life. Particular thanks, to

my husband, Harsha, for his tireless support, love and affection and without whom I

would have struggled to find the inspiration and motivation needed to complete this

thesis.

ix
 Last but not the least, I dedicate my thesis to my Grandmother “Jhaiji” who’s

Blessings and loving support has encouraged me throughout my academic career and life.

Sadly, Jhaiji left for her heavenly abode just a few days before the thesis was submitted.

x
 Chapter 1

Introduction

There has been a growing interest among the academia and the private sector for the use

of GIS techniques in the analysis and planning of retail store network. Almost without

exceptions, various retail organizations need to plan for complex consumer markets and

keep up with competitions. Over the past few decades the methodologies used for

research of sighting of retail outlets have become more sophisticated as a result of

applicable modeling procedures being developed with GIS. This study conducts a retail

location analysis of the relationship between the fast-food store performance of

McDonald’s and Burger King and the various spatial and socio-economic factors of

their respective catchment areas.

Analytical procedures in GIS and statistical techniques have been applied to

carry out the analysis in this study. In particular, study areas have been partitioned into

a set of Thiessen polygons and into various spatial configurations using variable buffer

polygons to emulate various spatial configurations of catchment areas (i.e., trade areas)

associated with each fast food store. The socio-economic profiles in the partitioned

polygons have been analyzed with a series of regression models. The result of the study

brought out a better understanding of how location factors influence the performance of

1
 2

the stores as well as how the socio-economic attributes of the catchment areas affect the

store revenues.

1.1 Research Objectives:

The main objective of this retail location analysis is to develop and apply methodology

for analyzing the relationship between fast food store performance and the various

socio-economic and demographic factors with various spatial configurations of their

catchment areas in Portage and Summit Counties.

The traditional role of GIS in retail demand-and-supply analysis has been to

analyze market characteristics such as consumer demand, geodemographics, traffic

flow, competitor locations, etc. and to search for an optimal location for a new retail

outlet or to close retail outlets in over crowded markets. Knowing the geographical

distributions of retail demand and supply is important in conducting marketing analysis

using GIS analytical tools. GIS can overlay different data sets onto one another in an

integrated environment. GIS analytical tools have been widely applied for exploring the

relationships between demand and supply in many types of business practices, including

operations of fast food restaurants.

However, perhaps due to relatively low real estate costs and flexible rentals or

perhaps due to the all too often time lag in adopting newly emerging technology, many

retailers do not make use of sophisticated location analysis methods that are now

available. Many a times, retailers follow the location decisions previously made by

anchor retailers. The choice of a store location has a profound effect on the entire
 3

business of a retail operation. For picking an optimal store-site, it is necessary to utilize

data of the demographics of that area (income, family size, age, ethnic composition, etc

of the population), traffic patterns, and similar kind of retail outlets or competition in

the area. These factors are basic to all retail location analysis. GIS tools can help to find

the right site along with market penetration, market share and trade areas by combining

aerial photos/maps, competitors’ locations, geodemographic factors, customer surveys

and census data.

GIS market analysis tools can also help to determine whether the products match

the lifestyle and buying patterns of the customers. In this study- Retail Location

Analysis: A Case Study of Burger King & McDonald’s in Portage & Summit Counties,

Ohio, an analysis of catchment areas of the analyzed restaurants has been done using a

series of regression models to analyze socio-economic and demographic factors in

various spatial configurations of the study area. The study area has been partitioned to a

set of Thiessen Polygons and also to sets of spatial configurations by using different

buffering zones surrounding the retail outlets to create different proximity polygons for

further analysis.

Thiessen polygons define individual areas of influence around each service

center, or in this case each fast food restaurant, in a set of points/locations of fast food

outlets geocoded in such a way that any locations within a Thiessen polygon are closer

to the polygon’s centroid (the retail outlet used to make up the polygon) than to any

other retail outlet. Buffer polygons have been constructed around the fast food locations
 4

based on various assumptions of how far the distances consumers may be willing to

travel to receive fast food services.

With the various spatial configurations of Thiessen polygons and buffer

polygons as defined by the locations of retail outlets, regression models have been

constructed to examine the importance of a set of selected socio-economic and

geodemographic factors. The different regression models that use different independent

variables as structured by both the Thiessen polygons and Buffer polygons have been

done to see how well or poorly either of the two approaches capture the variations in the

sales volumes of fast food stores.

In today’s world of highly competitive market environment, it has become

imperative that retailers must make use of spatial analytical technology to acquire new

clientele, retain the existing/current customers, to enable market expansion, and to stay

abreast with changing consumer tastes and requirements. Advances in GIS technology

reiterates the fact that the future success of retail, real estate and restaurants will be

determined to get a great extend by using this smart technology.

1.2 Summary:

Many successful businesses in the United States make use of GIS software to integrate,

view and analyze data using geography. Use of GIS techniques enables retailers to

understand and visualize spatial relationships and improves productivity and

effectiveness of the business processes. The use of multiple regressions modeling in this

study has been done to identify how the ethnic composition of population and median
 5

household income in the service areas of Burger King and McDonald’s restaurants

interact with one another to produce a specific sales outcome.

 Chapter 2

Problem Statements

Retail location analysis is an important part in site selection of a retail store. “A trade area

of a retail store is the geographical area from which it draws most of its customers and

within which its market penetration is the highest”(Ghosh and McLafferty, 1987). Retail

location analysis also helps to determine the focus areas for marketing promotional

activities, highlights geographic weaknesses in the customer base and projecting future

growth and expansion of the retail services (Berman and Evans, 2001).

2.1 Size and Shape of the Retail Trade Area:

The size of the retail trade area often depends on the nature of goods and services

rendered at the retail outlets, along with the geographical distribution of other competing

retail outlets. For instance, fast food restaurants like Burger King and McDonald’s sell

goods and services that are popular, easily substituted and affordable by the majority of

consumers create a smaller retail trade zone as compared to a specialty restaurant.

Usually, retail trade zones are not geometrically regular, i.e., a circle, a square or a

polygon. Rather, the shape of the trade zone is based on road networks, geology and

topography of the area, land use of the neighboring areas, etc.

6
 7

When examining the way customers travel to make retail purchases, it is always

necessary to take into consideration the distance that a customer has to travel. The

distances that customers may be willing to travel are different, depending upon the type

of object to be purchased. The number of trips undertaken by consumers and the travel

time will be different based on specialty or commodity product (Salvaneschi, 1996). For

purchasing a specialty product, which is generally expensive, unique or long lasting, the

consumer is willing to travel over a longer distance. This tends to expand the trading area

of that good or service. On the other hand, to purchase everyday supplies or common

items consumers often prefer convenience, as the trips for such goods are frequent,

distances are short and travel time is brief. For instance, people typically will not drive to

another town for fast food, unless they are on way to or back from other destinations.

According to consumer behavior studies the time availability of consumers is an

important variable in the convenience and fast food market. Therefore, it should be an

important part of market strategy (Darian and Cohen, 1995). In this thesis research, the

study area is partitioned into polygons representing trade areas for further analysis.

Several different approaches to creating trade areas are used. These include trade areas

defined as buffer polygons surrounding fast food restaurants with widths of 1, 2 and 5

miles. In addition, partitioning the study area into a set of collectively inclusive but

mutually exclusive Thiessen polygons with the restaurants as polygon centroid also

generates trade areas.

Generating buffers around features is a commonly used analytical procedure in

GIS. Most buffering methods create simple-distance bound geometric buffers around the
 8

features. Buffers surrounding retail outlets(or other service-rendering establishments) are

also known as service areas, hinterlands or market areas and have useful in many

geographical applications (Shaw, 1991; Sierra et al., 1999; Van Wee et al., 2001). A

buffer delineates the area within a specified distance of a feature. It can be created from

points, lines or polygons. The output buffers may be lines or polygons depending upon

the features and their distance are specified in map units (Price, 2004).

Concentric buffers represent the delineation of multiple levels of proximity. For

example, different distances of 1 mile, 2 miles and 5 miles from the store can be used to

generate buffer polygons around retail outlets. This type of concentric buffers may

reveal patterns of market penetration in which the inner buffers often account for the

largest proportion of customers while the density of customers decreases as one moves

away from the outlet to the subsequent buffers. This distance-decay effect reflects the

impact of geographic accessibility on store patronage. The actual size of the trade area

for each store varies, depending on the location of the store. The sharper the distance-

decay effect, the smaller would be the trade area for each of the fast food store.

For this study a regression models are applied that relates sales outcomes

(dependent variable) to many factors such as ethnic composition and median household

income (independent variable) of population in the retail trade zones of the Burger King

and McDonald’s in Portage and Summit Counties. These regression models show that

Burger King’s annual sales are better explained by the included independent variables for

buffers with widths of 1 and 2 miles than those of McDonald’s sales by the same set of

variables. For a 5-mile buffer and Thiessen polygons, sales are better explained for
 9

McDonald’s. Ethnic population and median household income for buffer polygons of 1

and 2 miles around the restaurants better explain annual sales for Burger King and

polygons of 5-miles for McDonald’s.

2.2 Summary:

Retail location analysis helps in site selection for a business outlet and in determining the

performance of retail outlets in the trade area of the store. The trade area of the store

reflects the socio-demographic characteristics of the clientele and is thus useful in

determining the marketing strategies. The size and the shape of a retail trade area are

determined by the nature of goods and services offered.

Since fast food restaurants sell goods that can be easily substituted, majority of

consumers form a small retail trade area. Ethnic composition of population and the

median household income within the buffer polygons constructed around the fast food

restaurants indicate how much time and distance consumers drive or travel to patronize

these restaurants.
 Chapter 3

Literature Review

During the past three decades, several important advancements have taken place in

spatial-data analysis, data storage, retrieval and mapping. Geographic Information

Systems have been very useful in tackling spatial analytic approaches and in forming an

interface with the field of location science (Church, 2002). Several studies give an

overview of the major impacts of GIS on works done in the field of location science in

terms of model application, development and various methods that can be used for land-

use suitability modeling (Malczewski, 2004). For example: GIS is now the most widely

used software for analyzing, visualizing and mapping spatial data such as retail location

analysis, transport networks, land-use patterns and census track data.

Since GIS can be used to assemble large volumes of data from various sources with

different map scales and in different coordinate systems, it is considered an important

tool in location analysis. GIS can combine and simultaneously use several databases by

transforming them into a common set of database (Pettit and Pullar, 1999). However, the

use of GIS in location analysis involves the aspect of accuracy of representing real world

situations in a GIS database. The notion of accuracy is the representation of geographical

objects and representing socio-economic, cultural and political elements of the

environment within which location analysis is done (Church, 2002). Not only is GIS used

10
 11

as the source of input data for a location model, it has also been used as a means to

present model results (Malczewski, 2004).

3.1 GIS for Business and Service Sector Planning:

The growing consumer orientation in business and service planning along with

advances in GIS and spatial analysis techniques, have led to the promotion of the use of

GIS in the area of business and service planning (Longley and Clark, 1995). Several

books and articles assess the use of GIS for supporting business and service planning at

the level of tactical and strategic decision-making (for example: Davies and Clarke,

1994; Benoit and Clarke, 1997; Clarke, 1998; Birkin, et al., 2002). These studies aim to

further explore and promote the use of GIS in the area of business and service planning

by demonstrating the benefits of both methodological advances and evidence of benefits

in GIS applications and spatial models in GIS. Business planning requires a critical

review of geodemographic features and paying attention to requirements posed by end-

users (Longley and Clark, 1995).

By linking GIS and spatial analysis software, proprietary GIS can be applied to

solving problems in several applications like retail location analysis, localized

marketing, etc. This involves the integration of spatial models and GIS customized to

the specific information needs of retail organizations for specific localities. Thus spatial

modeling is used in the explanation and prediction of interaction between demand and

supply for retail facilities and the search for suitable locations for retail outlets in an

area. The major theme of these studies is the evolution of GIS towards a more flexible
 12

and powerful spatial decision support system (DSS) or intelligent GIS (IGIS), applied in

several service sectors, including retailing, financial services and health care. Marketing

information systems (MKIS) are decision support systems targeted at marketing-

specific decisions (Birkin, Clark and Clark, 1996). There is a realizable benefit in

integrating GIS with MKIS because of its ability to provide map-based data

presentation considered most effective for decision-makers (Ronald and Lawrence,

2004).

3.2 GIS as a tool for Retail Location Decision:

A dynamic and uncertain environment characterizes retailing and retail organizations as

needing to plan for the complex consumer markets, while anticipating and reacting to

competitions. This competitive nature of retail environment and the large number of

techniques made use of by the retailers in locational planning, has led GIS to be used as

an aid in strategic retail decision making and applications (Davies and Clark, 1994).

GIS is used not just for location and catchments analysis but also for other retail sector

issues such as category management, merchandising, marketing communications and

relationship marketing (O’Malley, Patterson and Evans, 1997).

Existing literature contains a practical framework and other important issues

involved in retail network planning. GIS has contributed immensely in improving the

efficiency and precision of retail planning and marketing. Since the 1960s

methodologies used for retail outlet location research have become more sophisticated
 13

as a result of modeling procedures brought about by GIS (Birkin, Clark and Clark,

2002).

The US experience shows that the effective utilization of geospatial databases,

and the development of decision support systems (DSS), is becoming a significant

source of competitive advantage for retailers over those without. Some retailers further

explore information opportunities afforded by GIS technology for their business

practices. Rather than relying on customer information alone, they are now combining

data from several sources simultaneously in a bid to better support their process of

decision-making (Birkin, Clark and Clark, 2002).

3.3 GIS Methodologies for Retail Location Studies:

For analyzing the spatial structure of retail activities with location data at micro scale, a

number of technologies are now widely available and utilized. These include

application of methods such as Probability Density Function (PDF), Decision Support

Systems (DSS), Spatial Interaction Models, Network Huff Model, Analysis of Variance

(ANOVA) (Byrom, 2005), MATISSE (“Matching Algorithm, A Technique for

Industrial Site Selection and Evaluation”), and RASTT (Retail Aggregate Space Time

Trip Model) (Baker, 2003), and others.

The Probability Density Function (PDF) of the retail stores is a function of how

densities of the subject matters vary over specified dimension. If the specified

dimension is time, the probability density function describes how such matter changes

their frequencies and distribution over time. Alternatively, if the specified dimension is
 14

locations (or space), the probability density function then describes how such matters

vary in their spatial patterns. The PDF has been used to analyze the spatial structure of

retailing (Sadahiro, 2001). Sadahiro tested the validity of this method by applying this

method to the locational data of retail stores in Yokohama. This approach helps to

measure the degree of agglomeration, spatial patterns, the relationship between the size

and function of retail agglomerations and analyzes the spatial structure of retail

agglomeration.

Retailers for sales promotion activities and long-term strategic decision-making

are increasingly developing GIS as DSS. GIS merges endogenous database by retailers

and the exogenous databases sources to introduce retail decision- making and systems

implementation (Nasirin and Birks, 2003). As an example, the examination of the

experiences of some of the UK based retailers reflecting GIS implementation in retail

location analysis shows a highly organized series of process management that has

resulted as a result of this application.

The Network Huff Model is formulated on a network with the shortest-path

distance as an extension of the ordinary Huff (based on Euclidean distance) (Okabe and

Okunuki, 2001). This computational method can be used for estimating the demand of

retail stores on a street network in a GIS environment. Extending from the gravity

model, the original and network Huff models use distances (Euclidean or shortest

distance over a network) between retail outlets as inverse weights to estimate divisions

of the entire market area into individual trade areas of the retail outlets. The benefits of
 15

these models are the ability to meaningfully divide the studied space into a set of trade

areas to support retail business operations.

MATISSE is a knowledge-based decision support system (KBDSS) based on

decision tables that can be used by industrial decision-makers and planners to assess the

suitability of potential sites (Witlox, 2003). Witlox explains how a relational approach

to the modeling of the site suitability concept can be implemented and tried to find all

possible locations that meet the spatial production requirements based on the

organizational characteristics of the firm. The growing interest of urban geographers

and economic geographers in applying KBS, DSS and integrated system has been

largely attributed to the development of computer systems. Computers are able to store,

organize and process enormous amount of data as well as make possible the availability

and accessibility of the domain-specific knowledge underlying the spatial problem.

Witlox has identified three major categories of location factors at the highest

level of decision-making. These three conditions are site conditions, investment and

operating considerations and make up MATISSE’S head decision table. He points out

that the experience with the construction of the system indicates that the developed

procedure of knowledge in acquisition worked quite well, however, there are some

problems with capturing of compensatory decision-making in terms of the decision

table formalism. Nevertheless, the system is at a stage where it can be used in a

straightforward manner. Locational requirements can be activated and even changed

during consultation sessions and the system can identify and satisfy these requirements.
 16

As another effective means of partitioning space into meaningful sections based

on a set of focal points, Thiessen trade area models have been used to generate

theoretical trade areas based upon stored characteristics and consumer behavior

assumptions. Thiessen diagram models are used for delimiting trade areas for a set of

similar and competing facilities in an area (Jones and Simmons, 1993). This method is

also useful in those studies where detailed consumer patronage data is unavailable or

difficult to acquire. Thiessen polygons can be used to identify the impact and changes

of the existing set of facilities, as well as identify potential sites for new facilities

(Ghosh and McLafferty, 1987). Thiessen models do not require complex statistical

calculations and provide a quick and inexpensive appropriation of real trade areas

(Jones and Simmons, 1993).

Two kinds of Thiessen diagrams can be used to model retail trade areas (Boots

and South, 1997). The ordinary Thiessen or Voronoi diagrams (OVD), takes into

consideration the location of the stores and assumes that the consumers patronize the

nearest store (Ghosh and McLafferty, 1987). The second method is the multiplicatively

weighted Thiessen.

The multiplicatively weighted Voronoi diagram (MWVD) takes into

consideration both the locational as well as non-locational factors. According to this

approach, the consumers select stores on the basis of distance (time) and attractions of

stores (Boots, 1980). Boots and South used this approach to delineate the trade areas

and project sales estimates of supermarkets owned by the Zehrs chain in the twin cities

of Kitchener-Waterloo in Ontario, Canada. Thiessen polygons are contiguous but

 17

mutually exclusive. They cover all of the available area and are collectively exhaustive.

They have been used to estimate individual breeding densities for great tits in Wytham

Woods, United Kingdom (Wilkin, et al, 2006). The concept of Voronoi diagram has

been in use since antiquity. Voronoi-like diagrams were used to show cosmic

fragmentation and disposition of matter in the solar system and its environs in the works

of Descartes in 1664 (Mahoney, 1979).

A more comprehensive use of Voronoi diagrams appeared in the works of Peter

Dirichlet (1805-1859) and Georgy Voronoy (1868-1908) (Dirichlet, 1850). In their

context, sets of points was considered to be regularly placed in the space generated by

linear combinations of linearly independent vectors with integer coefficients. This

contains many points and the Voronoi diagram generated by these points partitions the

space into congruent polyhedra (Delone, 1961). Since the initial concept of Voronoi

diagram involved sets of regularly placed points in space, it was first applied in

crystallography (Nowacki, 1933, 1976). At this time Voronoi diagram was also applied

to areas involving spatial interpolation. Thiessen used Voronoi regions to compute

estimates of regional rainfall averages (Thiessen, 1911). Horton also developed the

procedure in the same context (Horton, 1917).

In 1912, Whitney referred to the procedure as ‘Thiessen’s method’ and since

then this application has been widely used in geography, meteorology and other social

science disciplines (Whitney, 1912). By 1960s, the knowledge of Voronoi diagram was

being applied in various studies of natural and social sciences. Somehow, the empirical

application of the concept was limited as it lacked a simple and efficient means of
 18

construction. The researcher had to rely on methods that involved tools such as the

compass and the ruler (Kopec, 1963). Gradually with the developments in the field of

computer sciences, algorithms were developed to help in constructing Voronoi diagram

in two and three dimensions (Shamos and Hoey, 1975).

Thiessen polygons have been extensively used in a wide variety of studies. For

example, Thiessen polygons have been applied to calculate mean areal precipitation

computation (Dartiguenave and Maidment, 1996), derivation of Road centerlines

(Ladak and Martinez, 1996), distribution of health services (Taylor and Carmichael,

1980; Brabyn and Skelly, 2002) and watershed delineation

(http://www.barrodale.com/watershed/tour.htm), climate change on forestry

(http://www.pfc.forestry.ca/climate/change/spatial_e.html), snow investigations

(http://www.ph.ucla.edu/epi/snow/mapmyth/mapmyth2_a.html), hydrodynamics

(Hargrove, Winterfield and Levine, 1995).

Voronoi diagrams are useful for spatial analysis and also for spatial optimization

(Okabe, et al., 2000). They are used to discuss the locational optimization of points in a

plane or space where points represent certain facilities or services. These points are

located in a continuous plane where demand arises at any point in a plane and the

feasible locations of these facilities are these points in the place. Computational

methods have been developed to solve locational optimization problems of a large

number of points in a continuous plane or space with the help of Voronoi diagram

Okabe and Suzuki, 1997). Voronoi diagram have also been used to minimize the total
 19

travel time and average travel cost to the nearest point facilities and locational

optimization of lines.

3.4 Analysis of Trade Areas:

According to the available literature, some other methods have also been used for

delineating retail areas. These methods have been classified into the following

categories:

1. Simple or Basic methods for trade area analysis

2. Gravitational methods for trade analysis

3.4.1 Simple or Basic Methods of Trade Area Analysis:

William Applebaum pioneered the analog method in 1932, for developing systematic

retail forecasting model based on empirical data. This method is commonly used by

retail and consulting firms to quantify the performance characteristics of existing stores

in order to forecast sales at new sites (Rogers and Green, 1979). The analog method is

non-geographic and is often implemented by regression analysis (Wang, 2006).

Proximal Area Method is a geographic approach for delineating trade areas. This

method assumes that consumers choose the nearest store to visit among the similar kind

of outlets (Ghosh and McLafferty, 1987). This method also assumes that customers also

consider travel distance and time while choosing a store. Once the trade area is defined,

the store sales can be projected by analyzing demographic variables and spending habits

of the perspective customers (Wang, 2006). By using GIS techniques, proximal area
 20

method can be studied by two approaches. The first approach is the consumer based and

the second is store based (Wang, 2006).

3.4.2 Gravitational Methods for Trade Analysis:

The consumer based approach looks for the nearest store location in relation to the

consumers. The store based approach constructs Thiessen polygons around each store in

order to define the proximal area. The Thiessen polygon layer can be overlaid by

demographic variables in order to obtain consumer information. This method takes into

consideration distance and time traveled to delineate trade areas. Other sales-forecast

methodologies have been developed and applied that consider distances (or time) and

attraction of stores (Reilly, 1931 and Converse, 1949). One of these techniques that

have been used for many years to delineate retail-trading areas is based on the law of

retail gravitation. This law establishes the relationship between two cities on the basis of

their relative populations and distances between them (Reilly, 1929). The statistical

formula applied for establishing this relationship as given by Converse (1943):

 21

The Dist. from B to the breaking po int of the retail trade area between two cities

Dist. from A to B
= 1 + Population of A / Population of B

Where:

Α = first city

Β = second city

The law of retail gravitation has been used for marking off the areas/ zones from which

a retail outlet gets its patronage.

D. Huff, T.R. Lakshmanan and W.G. Hansen later improved a more general

gravity based model in a probabilistic framework to define trade areas of multiple

stores. This model is based on the assumption that the probability of a customer offered

a set of alternatives, selecting a particular item/service is directly proportional to the

perceived utility of each alternative. This model can be used for predicting consumer

spatial behavior, delineating trade areas, locating retail and service facilities, analyzing

market performance and forecasting sales, etc.

3.5 Forecasting the Fast Food Restaurant Sales:

Retailers always seek growth and expansion of their revenues and profits. In order to

achieve this, they adopt various strategies like opening up new outlets, diversification of

goods and products, increasing marketing efforts, etc. The scope for increasing the
 22

revenues of existing outlets also depends upon the size and economic potential of the

geographic area served by the outlets. An essential feature of retail outlets is the spatial

orientation of their markets (Ghosh and McLafferty, 1987). Each store may have a

geographic area from where most of its customers originate. Market potential of any

store is determined by the expenditure pattern of residents in the trade area. The market

potentials are however, not static and may change over time due to changes in economic

trends, population size, age and ethnic composition and other socio-economic indicators

(Ghosh and McLafferty, 1987). Therefore, an understanding of the orientation of

customers is the basis on which the retail outlets should make their target market

decisions.

The population size, its demographic composition, expenditure potential and

customer orientation has to be related to the competitive environment of a retail chain.

The level and quality of direct competition is an important assessment that has to be

considered in any store location strategy (Mercurio, 1984). Retailers measure

competition levels by including stores per capita, square footage per capita as well as

the degree of market share concentration, etc.

A retail chain develops its store location strategy keeping in view the future

retail setting of the market. The population numbers, demographic set-up and the

expenditure potential of the catchment area are important factors of the competitive

environment (Davies and Rogers, 1984). Knowledge of the regional shopping area and

consumer orientation provide a general orientation to the retail environment.

 23

Sales forecasts/ projections can be made for future store locations. A gravity

model uses three key factors such as: size of the store, distance traveled to get to the

store and the retail image of the store based on its products, easy accessibility, visibility,

parking, etc. (Mercurio, 1984). One of the prominent features of retail revolution over

the past few decades has been the transformation from independent, family-run firms to

large-scale, professionally managed multiple retail organizations (Dawson, 1991). In

most developed countries large multiples (i.e. with ten or more retail outlets) account

for a major part of the total retail sales (Burt and Dawson, 1990).

The extent of chain store dominance may vary from sector to sector, but the

food retailing exhibits the greatest degree of concentration. The retail trade market is

characterized by a small number of extremely large retail organizations and a large

number of small-scale outlets (Brown, 1992). Therefore, it becomes imperative for the

fast-food retail outlets to not ignore their competitors when taking locational decisions.

Micro-scale retail location cannot be studied in complete isolation. Hence, the broader

locational literatures like Central Place theory, Spatial Interaction theory, bid-rent

theory, etc. that has been applied at the macro level can be adapted to the micro-scale

retail location settings. Similarly, cognitive mapping and methodologies exercises

conducted at micro-scale can be employed in wider macro-scale context (Brown, 1992).

Retail site selection is the most important aspect of any business. An accurate

projection of sales often helps to determine the right amount to invest in order to get

maximum returns. For opening retail business hard costs, such as real estate,

construction, equipment, interior decoration and furniture and soft costs such as zoning,
 24

professional fees, training and personnel relocation are taken into consideration

(Salvaneschi, 1996). Many a times for estimating the sales projection figures of a future

store, similar kinds of stores or outlets offering similar kind of goods and services are

evaluated- also called analogues.

Statistical or mathematical models of analysis have also been used to forecast a

store’s future sales. Regional and trade related data can be used as inputs for these

models for sales forecasting. Three major statistical methods used are:

1. The Regression Model

2. The Gravity Model

3. Reilly’s Law

Many real estate consultants prefer regression model. Volume Shoe Corporation

uses four types of multiple regression models to predict new stores based on

quantitative information derived from existing stores. These models contain

demographic data from new store location: population density, median household

income, percentage of low-middle income households, age data, percentage of non-

white population, occupation, etc. (Wood, 1986).

A regression model can be applied in retail location decisions, that relates sales

output (dependent variables) to one or more factors (independent variables) positively

or negatively related to sales. The results can be compared to existing similar stores for

future development of the retail business (Thompson, 1982; Green, 1986).

Regression models have also been used to identify location variables related to

store sales performance. For state-owned liquor stores in Charlotte, North Carolina, the
 25

regression model related annual sales volume for each existing store (dependent

variable) and population within 1.5 miles of the store site, mean household income,

distance from subject store to next nearest liquor store, daily traffic volume,

employment within 1.5 miles of the store, etc. (independent variables) (Lord and Lynds,

1981).

Regression analysis can be used to determine the factors that influence the

performance of retail outlets at a particular site. The performance of an individual store

may depend upon a number of factors and regression analysis can help to identify the

factor that has the greatest impact for a particular retail outlet. The development of

regression model is based on two assumptions:

1. Performance of a store affected by its location characteristics, socio-economic

composition of the trade area, level of competition and store characteristics

2. These factors can be isolated by systematic analysis (Ghosh and McLafferty,

1987).

There are many empirical studies done that have used regression models on retail

performance of a variety of outlets. Some of these studies include convenience stores

(Hise et al., 1983; Jones and Mock, 1984), hospitals and health services (Erickson and

Finkler, 1985), grocery stores (Cottrell, 1973), banks and financial institutions (Martin,

1967; Clawson, 1974; Olsen and Lord, 1979; Lynge and Shin, 1981), etc. Most of these

studies infer that the population numbers and demographic composition of an outlet’s

trade area affect performance to a great extent. Composition has a complex influence on
 26

store performance but merchandizing, promotion, customer services and sales also have

a positive relationship.

The regression model can be explained as:

Υ = b0 + b1x1 + b2 x 2 + .... + bnxn

Where:

Υ = Dependent variable: measure of performance of a store (store sales revenue/

profitability)

x = Independent variables: that may influence store performance

Parameters b0 = intercept term

b1, b2,...., bn = Regression coefficients corresponding to independent variables

(measure the relative impact of each variable on performance).

3.6 Identifying the Trading Area of the Fast Food Restaurants:

Every retailer seeks growth and expansion of the business. This growth can be

accomplished by either increasing sales/revenues from existing stores/establishments

and expansion by adding more establishments. For physical expansion into new or

existing markets, an analysis and selection of the broad geographical region should be

done. This process of site selection for a new retail outlet follows a hierarchy from a

macro-analysis at the regional and market levels, to trade area analysis and finally down

to a micro-analysis of a particular site (Anderson, 1993).

A region consists of several geographic markets, which in turn consist of several

trading areas encompassing metropolitan regions, cities and towns. Within each trade
 27

area there might be a number of potential sites for establishing retail outlets. According

to the United States Census Bureau, there are nine census regions in the country, which

were established in 1910 for the presentation of census data. These regions are: New

England, Mid-Atlantic, South Atlantic, East North Central, East South Central, West

North Central, West South Central, Mountain and Pacific (US Census Bureau, 2002).

This Census Bureau classification has been used by a number of data sources for

reporting statistics on retailing, manufacturing and other economic analysis. Many large

chains like JC Penny, have divided the United States into regions based on their

operations (JC Penny, 1987). These nine regions of the United States can also be

conceptualized as the nine nations within North American. Each of these divisions

depicts a distinct cultural and anthropological entity rather than mere political divisions

(Garreau, 1981). Each regional market may exhibit varying retailing opportunities.

Large retailing chains adapt their marketing strategies to fit the requirements of the

individual regional market based on its physical, geographical, consumer, economic and

competitive characteristics (Anderson, 1993). A regional market further consists of

diverse geographic markets like a metropolitan region, city or town. Retailers have to be

constantly aware of the changing boundaries of market area and conduct market

analysis as an ongoing process. The changes in the boundary of the market areas may be

brought about by customer locations and characteristics, modifications in major traffic

arteries and entry of a new competition in the market, etc. (Huff and Rust, 1984).

Market areas are further subdivided into trade areas that contain target market

populations from which a particular retail outlet draws its customers. The point, at
 28

which the competition edge of a particular retail establishment is lost in favor of an

alternative establishment, determines the trade area boundaries of a retail outlet or

service (Anderson, 1993). While carrying out this trade area analysis, customer

movement based on points of origin (residential or employment areas), preferred retail

destinations and nature of outlets bypassed on their way to the destination store, must be

evaluated. The trade area analysis is carried out on the basis of geography,

demographics, economic, administrative and competitive characteristics prevailing in

the trade area.

Although there have been different interpretations for the concept of a retail

trading area, generally it refers to a geographical area from where retail patronage or

clientele is derived. But the question arises as to how much clientele contribute from an

area should be included within a retail trade zone of a retail outlet. Theoretically it has

been accepted that if there are any customers of a retail outlet living in the study area,

then it should be included as a part of the retail trade zone (Fine, 1954).

A retail trade area surrounds the retail outlets and within which 75 to 80% of

customers more from one point to another (Salvaneschi, 1996). The retail trade area is

formed by people who live, work and move there and also a percentage of customers

who come from outside. According Salvaneschi, a retail-trading zone has three main

sections of customers. The core area that accounts for about 50% of customers, the

secondary area that accounts for up to 25% of customers and the tertiary area that

accounts for 10-15% of customers (including those from beyond 4-5 miles). The

customers who come from beyond the tertiary area are sometimes referred to as the off-
 29

map customers. These customers vary in frequency and numbers according to the type

of goods or services offered, the type of road and the topography of the area.

A retail outlet’s retail trade zone can fall into one or a combination of areas such

as: downtown, urban, rural or suburban. Each of these categories can reflect different

types of land uses such as residential, industrial, commercial, professional or

recreational. The profile of the customers will differ according to the zone in which they

are located.

3.7 Retail Marketing Strategies:

The retail marketing strategy is greatly dependent upon the firm’s value platform. The

value platform deals with the manner in which a firm distinguished itself from its

competitors (Ghosh and McLafferty, 1987). Therefore, in order that consumers

patronize a particular fast-food outlet, a value platform has to be established by the

outlets. Creating and maintaining this value platform will allow the store to achieve a

differential advantage over its competitors.

The other functional strategies followed by the store such as merchandizing,

advertising, store atmosphere, customer service, easy accessibility, visibility, parking,

etc. have to be kept consistent with the value platform (Ghosh and McLafferty, 1987;

Mercurio, 1984). The store management has to understand the customer needs and

wants and also be aware of competitions faced by the outlet. Customers may have

varying needs and wants and how they perceive the value of eating at a fast-food joint.
 30

These differences may create the potential for segmenting the trade area based on

different customer expectations.

The fast food stores can select the segment of consumers it can serve best and

orient its services to meet the needs of that group. In delimiting its trade market, the

store must also take into consideration the strengths and weaknesses of the potential

competitors as well as its own resource base (Ghosh and McLafferty, 1987). In any

market area, there may be a variety of stores offering similar products and services.

Under such circumstances, the value platform of each store will determine where it

stands in the extent of competition. And if the value platform of two competing stores is

more or less similar, then an overlap in their target markets is greater.

According to Ghosh and McLafferty, retail stores can be grouped into

competitive clusters based on how similar their value platforms are. Firms belonging to

each of these groups follow similar marketing strategies, target the same segment of

consumers and compete directly with each other. The competitive edge of the store can

thus be maintained by price of goods and services as well as the outlet location.

Therefore, if a consumer has to choose among similar retail alternatives, he or she is

likely to choose the store that is more conveniently located. Price of products and

services and store location are the most important factors at this stage of competition

within strategic groups, when consumers have to make a choice (Ghosh and

McLafferty, 1987).
 31

3.8 Eating Facilities At Fast Food Restaurants:

Fast food restaurants like Burger King and McDonald’s generally have three types of

eating facilities like: Dine-in, Carryout and Drive-thru. The types of customers for each

of these facilities are different and therefore, their expectations are also based

accordingly. In today’s fast-paced life, where customers are looking for convenience,

the Drive-thru concept has become very popular. Customers on the move generally

prefer the Drive-in facility for pure convenience.

A substantial part of any quick service restaurant’s (QSR) revenue comes from

the Drive-thru. According to an article in QSR magazine, “Technology as a whole, now

plays an indispensable role in quick-serve. From consumer applications that enhance the

diner experience to operational applications that improve the restaurants’ day to day

functioning, experts say quick-serves that do not embrace technology are missing both

revenue-boosting and cost cutting opportunities that could put them at a competitive

advantage”. (QSR Magazine, 2002).

Drive-thru facilities have existed since the late 1920s and until 1970s most QSR

customer orders were taken inside the restaurants. Over the last twenty-five years there

has been a spurt in drive-thru sales. Some newer QSR chains have reported a 50-75%

share of their revenues from the drive-thru restaurant facilities (USA Today, 2002). In

1975, McDonald’s opened its first drive-thru window in Sierra Vista, Arizona. This

service gave the fast food consumers a convenient way to get a quick meal. The

company’s goal was to provide service in 50 seconds or less. Drive-thru sales

eventually accounted for about half of all McDonald’s restaurant sales in the United
 32

States (www.mcdonalds.ca and www.fundinguniverse.com). Even though the drive-thru

concept originated in the QSR history, today we can find drive-thru video-stores, banks,

photo-processing shops, coffee outlets and grocery stores, etc.

3.9 Pricing Strategies:

Although both McDonald’s and Burger King were established in the same year (1954)

and by the 21st century, McDonald’s has been considered as the undisputed hamburger

king especially in terms of sheer size. By 2000, more than 43 million people visited one

of McDonald’s 26,000 restaurants in 120 countries every single day. Which is about

more than 15 billion customers a year with system-wide sales of over $40 billion.

Burger King, by comparison, served nearly 15 million customers daily. Burger King

Corporation and its franchisees operated over 11,000 restaurants in the U.S. and 57

countries and international territories around the world, producing 2000 system wide

sales of about $11 billion.

While both companies sold burgers and fries, their nearly 50-year battle had

been waged around two competing concepts. McDonald’s aimed to be the world’s best

‘quick service restaurant experience’. Being the best meant providing outstanding

quality, service, cleanliness and value. Burger King based its strategy around customer

satisfaction through flexibility. Since the company’s founding in Miami, the Burger

King brand had become recognized for its flame-broiled taste and ‘Have it Your Way

Food Customization’ (Johnson & Pyke, 2004).

 33

Most fast food restaurants offer discounted prices on their products in order to

attract more customers. McDonald’s has a $1 Billion marketing budget, almost twice

that of Burger King Corp. But in order to compete on the basis of price discounts,

McDonald’s is forced to sell Quarter Pounder at 99 cents (Regular price $1.90)

(Bremner and DeGeorge, 1989). Even though McDonald’s is continuously fighting for

market share, it is also worried about the negative effect of price slashing of its products

in the eyes of consumers.

McDonald’s have been introducing new products on its menu (pizzas, low-fat

milkshakes) and also opening up outlets in diverse markets like hospitals, airports and

museums. McDonald’s have also extended their menu choices by including healthier

platter like salads, soups and yogurts. But they still maintain discounted burgers on their

menu in order retain their clientele and win new customers (Gibson, 1990).

3.10 Summary:

Much technological advancement has taken place in the field of GIS over the past few

years and it is widely used software for spatial data analysis. The review of available

literature related to the study topic in this chapter provides the necessary background

understanding for the interplays between retail outlets and various location and socio-

economic factors. This literature review has helped to identify a set of useful articles

and books by accredited scholars and researchers, which have applied GIS principles of

analysis to retail location studies.

 34

GIS has been helpful in accurately representing real world situations and has

therefore been used for business and service planning. GIS offers a large number of

techniques that have been successfully made use of by the retail businesses in location

planning and strategic retail decision-making and applications. Over the past few

decades, the methodologies used for retail outlet location research have become more

sophisticated as a result of modeling procedures brought about by GIS.

Retailers are always trying to grow and expand their business. This is done by

increasing sales revenues of the existing stores and by adding more establishments.

Looking into the size and economic potential of the trade area often does this. The

market potential of a store is determined by the ethnic composition and the expenditure

pattern of residents in the trade area. Selecting site for a new retail outlet requires an

accurate projection of sales in order to determine the right amount of investments for

maximum returns.

A good marketing strategy has to be integrated into the organization’s marketing

policies and goals. Retail marketing strategies can allow the businesses to concentrate

their resources on the greatest opportunities to increase sales and at the same time

achieve a competitive advantage.

 Chapter 4

Data and Research Methodology

The concept of Trade Area analysis has always been quite appealing and popular with

retailers because it helps them to gain much needed understanding of the business

potential and the competition with other retailers. In its simplest form, a circle of certain

radius around a retail outlet can function as the outlet’s trade area. In the circle around a

retail outlet, it is possible to analyze certain geodemographic factors and other variables

related to that particular business.

GIS can easily calculate the size and the potential of the market based on the

socio-economic profile of the area in the identified trade area, for example, the circle

around a store. However, Trade Area analysis in the past was limited to rough estimates

and mental arithmetic because of the time consuming manual calculations involved. With

laborious manual calculation required, analysts often assumed that people travel in

straight lines, instead of using actual distance or traveling time. They also had to work

with the notion that those residing within the radius of the circle are the customers and

nobody outside the circle shops at that store, ignoring traffic patterns and the additional

attracting factors by any nearby attractions. Moreover, they often assumed away

competition because of the extensive time and labor involved in including all these in

their calculation and analysis. With GIS, however, it is now possible to handle larger

35
 36

volumes of data, with higher levels of precision/details, and shorter turn around time

(speed).

Competitive Location Evaluation and Optimization (CLEO)

(http://dgrc.ca/services/retail/index.html) is a suite of retail location models that push the

frontiers of location science to identify the very best sites in a multi-store market. It is

retail location software developed by a GIS consulting firm Digital Geographic Research

Corporation (DGRC) (http://dgrc.ca/index.html). Some of the recent retail location

analyses have tried to improve upon the aforementioned limitations by considering non-

circular trade areas; by detailing trade areas into primary or secondary trade areas and

even by constructing Thiessen polygons to partition the space with a given pattern of

retailers. In today’s consumer society, shopping behavior and store choices depend upon

many factors such as advertisement, complementary shopping opportunities and

proximity to one’s place of residence or work (CLEO) and others.

4.1 Data Preparation:

In this study, GIS and statistical analysis techniques have been applied as an applicable

tool for retail business. This study examines the spatial patterns, and the dynamics of the

fast food restaurant chains of Burger King and McDonald’s in Portage and Summit

Counties in Northeastern Ohio.

The data used in this study include data layers of Block Groups 2000 –

TIGER/Line® Shapefiles for Portage and Summit Counties provided by U.S. Census

Bureau. The data layers were in Geographic Coordinates NAD83 and re-projected to
 37

Geographic Coordinate System GCS_Clarke_1866. Geodemographic data for ethnic

composition of population and median household income has been taken from Census

2000 Summary Files (SF 3) and joined with the Block group layer. The annual sales data

for midyear 2004, for Burger King and McDonald’s QSR (Quick Service Restaurants) in

Portage and Summit Counties of Ohio, has been obtained from Restaurant TrendsTM, a

research company in the United States that tracks the performance of the chain restaurant

industry.
 38

TIGER files
Data Layers Census Data
Preparation Restaurant Data

1
Mile
Geocoding of
Restaurants Buffers
2
Miles
Proximity Analysis
Thiessen
5
Geoprocessing Miles

Overlay Analysis Union

Regression Analysis Enter and Stepwise

Chart 4.1: Research Methodology

 39

4.2 Research Methodology:

Research methodology includes the following GIS techniques and statistical analysis.

The analytical procedures are outlined in Chart 4.1 as well as in the discussion below.

4.2.1 Geocoding:

This is the method that converts addresses to specific point locations on the street

network based on locations of the addresses as defined in a reference data layer of street

information. Taking into consideration the various types of address parameters one can

do Geocoding by matching the house number and street number against those in the

referenced street database. In this study the fast food stores have been geocoded using

their street address.

Geocoding combines map information with street address so that a point can be

located uniquely on a base map for each corresponding address. The process of address

matching, or Geocoding, required a database of properly formed addresses of the studied

restaurants, a reference database of streets for Portage and Summit Counties and a

Geocoding service for matching them.

In ArcCatalog, a new Address Locator was created by choosing the style of the

Geocoding service as US Streets. The US Streets address locator style helps to create

address locators for common addresses. Another advantage of using this style is that it

provides a range of house or property number values for both sides of a street segment.

This style also provides a location along the street and the side of the road segment where

the address is located (Crosier, 2004). The US Census Bureau maintains a database of
 40

streets with address ranges that have been enhanced in the various commercial

Geocoding products/ software. This database, referred to as the TIGER/Line files, can be

easily downloaded from the Bureau of the Census website (http://www.census.gov).

Moreover, private data vendors also make this database available for downloads. For

example, ERSI’s website has the TIGER/Line files formatted to shapefiles so that their

software can work with them directly. TIGER/Line files in shapefile format can be

downloaded from http://www.esri.com/data.

Since each of the Burger King and McDonald’s restaurants in Portage and

Summit Counties has a properly formed address, ArcGIS estimated the position of each

address as a point location and stored them in a point data layer that has the same

geographic projection as that of the street reference data layer. This estimation is done

with the linear interpolation between the two endpoints of the street segment that contains

the house number of the matched address. Figure 1, Figure 2 and Figure 3 show the

spatial distribution of the fast food restaurants included in the study.

 41

Fig: 1 Burger Kings and McDonalds’ QSR: Portage and Summit Counties
 42

Fig 2: Burger Kings and McDonalds’- Portage County

 43

Fig 3: Burger King and McDonald’s- Summit County

 44

Sixty-five of the studied restaurants fall under the Urbanized areas of Cleveland

(66.98 square miles), Akron (273.43 square miles), Windham (1.69 square miles) and

Youngstown (0.12 square mile). The other four restaurants that are situated outside of these

urbanized areas are in Garrettsville (McDonald’s in Portage County), Mantua (McDonald’s

in Portage County) and Richfield (Burger King and McDonald’s in Summit County).

Figure 4 shows the distance of the studied restaurants from four types of roads as

classified by the US Census Bureau has been taken into consideration. According to the

Census Feature Class Codes (CFCC) these roads are A1- Interstate Highways and Toll

Highways, A2- nationally & regionally important highways, A3- State Highways and

A4- Local Roads.

 45

Table 1: Distance of Burger King & McDonald’s QSR from Major Roads

Number of Burger King QSR

Road Type 1 Mile 2 Miles
A1 12 16

A2 3 3

A3 23 24

Number of McDonald’s QSR

Road Type 1 Mile 2 Miles
A1 24 33

A2 2 4

A3 41 44

Above table shows that about 36 restaurants are situated within one mile of the interstate

highways and toll highways. Out of these 12 are the Burger King restaurants and 24 are

the McDonald’s restaurants. About 49 of the restaurants are situated within two miles of

the interstate highways. Out of these, 16 are the Burger King and 33 are McDonald’s

restaurants.
 46

Fig 4: Distance of Burger King & McDonald’s Restaurants from Major Roads
 47

Five fast food restaurants are situated within a mile of the A2 roads. Three of

these are Burger King and 2 are McDonald’s QSR. And within 2 miles of the A2 roads,

seven fast food restaurants are situated. Three are the Burger King and four are the

McDonald’s restaurants.

About 64 of the 69 studied restaurants are situated within a mile of A3 roads.

Burger King makes up about 23 and McDonald’s 41 of these QSR. Sixty-eight

restaurants are situated within 2 miles of the state highways (Burger Kings 24 and

McDonalds’ 44 restaurants).

All the sixty-nine studied restaurants are accessible by local (A4) roads.

4.2.2 Catchment area analysis: has been done by using:

(a.) Thiessen Polygon

Thiessen polygons, also known as Voronoi polygons, represent areas of influence around

a set of focal points (or retail outlets as in this study). They can be generated based on a

set of points as centroid for the polygons. Thiessen polygons have been so constructed

that each polygon contains exactly one of the outlets and that any location with a

Thiessen polygon is closer to that point outlet (centroid) than to any other point outlets.

This polygon structure is constructed by using the perpendicular bisectors between

neighboring points (in this case, retail outlets) as the boundaries of the resulting Thiessen

polygons.

In this study, GIS techniques have been used to predict the catchment boundary

along a direct transact between each pair of neighboring fast food outlet in the sample set
 48

based on customer choice patterns. Once the location of a catchment area boundary has

been established along a given transact, this has been compared to that predicted by the

theoretical polygon boundary. This process helps to assess the validity of the Thiessen

boundary, and if its position is found to be inappropriate, the position of a more realistic

boundary can be identified (Gething et. al, 2004). Many commercially available GIS

software packages now offer such functions at only a few keystrokes to produce Thiessen

polygons for a given set of points.

The Thiessen tool in ArcGIS can proportionally divide and distribute point

coverage into Thiessen or Voronoi polygons (ArcGIS Desktop Help 9.2).

The Thiessen polygons are constructed as follows:

a.) All geocoded restaurants are triangulated into a triangulated irregular

network (TIN).

b.) The perpendicular bisectors of each triangle edge are generated that form

the edges of the Thiessen polygons. The locations at which the bisectors

intersect determine the locations of the Thiessen polygons vertices.

c.) The Thiessen polygons generate polygon topology.

Each Thiessen polygon contains only one input point (geocoded restaurant) and

any location within the polygon is closer to its point than to the point of any other

polygon. Thiessen polygons have been generated for Burger King and McDonald’s

Restaurants in Portage and Summit Counties, Burger King and McDonald’s in Portage
 49

County, Burger King and McDonald’s in Summit County, Burger King in Portage

County, Burger King in Summit County, Burger King in Portage and Summit Counties,

McDonald’s in Portage County, McDonald’s in Summit County and McDonald’s in

Portage and Summit Counties. Figure 5 to Figure 13 show the spatial structures of the

Thiessen polygons calculated using different subsets of the fast food restaurants included

in this study. Figure 5, shows the Thiessen polygons around the geocoded Burger King

and McDonald’s restaurants in Portage and Summit Counties. (Please refer to Figure 6 to

Figure 13 in the Appendix).

Using Thiessen polygons as trade areas of a set of retail outlets, analysts

can evaluate the retail outlets in terms of socio-economic and/or demographic attributes

of each outlet’s trade area. For example, census data can be summarized by Thiessen

polygons to provide hints to business planners the size of the market, the potential

customer base, the demographic composition of the population in a particular trade area,

or the estimated purchasing power or consumption patterns of the population in the trade

area.
 50

Fig: 5 Burger King and McDonald’s-Portage and Summit Counties

(Thiessen Polygons)
 51

(b.) Buffer Polygons

Rather than simply being interested in the locations of a retail store, an analyst may be

interested in the locations within a pre-defined distance from a store. For instance,

wanting to know about geodemographics of all areas within or outside of 1 mile from a

particular fast-food store, or within 5 miles of a highway exit or within 2 miles of an

urban area. Where information of this type is required, a buffering technique is made use

of. Buffers may be generated around a point, a line segment, or a polygon, with a

specified width. Following the establishment of buffer polygons, analytic steps similar to

those used in analyzing Thiessen polygons have been taken to examine how buffer

polygons behave similarly or differently from those of Thiessen polygons in this study.

Geoprocessing is the process of applying geographic analysis and modeling to data

to produce new information. The ArcINFO Geoprocessing environment has many tools

for processing all types of data. Some of these tools that have been used for this study

are: Overlay analysis (Union tool) and Proximity analysis (Buffer tool).

Buffer analysis tool creates a new feature class of buffer polygons around the

geocoded fast food restaurants. Buffers work in Euclidean space and use a two-

dimensional algorithm (ArcGIS 9.2 Desktop Help).

 52

The width of the buffer can be specified by two methods:

a.) Fixed distance - when a constant distance is specified and applied to the

Input Features.

b.) From Field- when the name of a numeric distance is chosen from the

specified feature class and each feature is buffered according to the chosen

value.

These buffer polygons allow us to see how they are related to the performance of

the fast food restaurants. The buffering function generates polygons by encircling a point

with a specified distance, or the buffer width. Geometric buffers (sometimes also referred

to as geometric outlines, geodesic offsets, or geodesic parallel) are applied for various

spatial analyses in GIS (Gombosi and Zalik, 2005). Theoretically, geometric buffer is

defined as a Minkowski sum by:

Α+Β= {x + y / x ∈ A, y ∈ B}

Where: - Α represents a disk, and Β a polyline (Gombosi and Zalik, 2005).

In order to calculate store revenue and optimum locations of retail outlets, GIS

literature often suggest the techniques of buffer and overlay analysis (Beaumont, 1991;

Elliot, 1991; Howe, 1991; Reid, 1993; and others). These studies first estimate how far

customers are willing to travel to a store and the result will be either a travel time or

distance. The next stage delimits an area around the store with the help of a buffer to

mark the limit of that time or distance from the store. The revenue generated from that
 53

store can be estimated by overlaying the consumer spending power that resides within

that buffer (Benoit and Clarke, 1997).

Three buffer zones generated around each fast food restaurant with the widths of

1-mile, 2-mile and 5-mile, are chosen to emulate consumers’ willingness to walk over to

a fast-food restaurant for a quick lunch in a congested area (1-mile), in a less congested

area (2-mile), or with a car they can drive to the restaurant (5-Mile). Demographic data

and median household income data of each polygon serves as the independent variable

for regression analysis and store sales volume as the dependent variable.

However, when the buffers around the fast-food outlets overlap one another, it

may become difficult to distribute the attributes of the population between different

buffers. This problem would be more apparent when there are two or more stores located

closer together and have very similar catchment areas. When it becomes difficult to

estimate how much revenue exists for each store in such a situation, the fair-share method

can be employed (Beaumont, 1991). According to this method the number of stores in the

buffer divides the total amount of revenue generated within the buffer.

In this study, fixed distance method of Buffering has been applied to generate

buffers of 1 mile, 2 miles and 5 miles around each of the Burger King and McDonald’s

fast food restaurants in Portage & Summit Counties. Figure 14 to Figure 40 provide the

layout of the buffer polygons with various subsets of the fast food restaurants included in

this study. Among these maps, I wish to point out that Figure 14, Figure 15 and Figure 16

show the Buffers with radii of 1mile, 2 miles and 5 miles around the geocoded Burger
 54

King and McDonald’s restaurants in Portage and Summit Counties (Please refer to Figure

17 to Figure 40 in the Appendix).

 55

Fig 14: Burger King & McDonald’s-Portage & Summit Counties (Buffer 1-mile)
 56

Fig 15: Burger King & McDonald’s Portage & Summit Counties (Buffer 2 miles)
 57

Fig 16: Burger King & McDonald’s Portage & Summit Counties (Buffer 5 miles)
 58

(c.) Overlay Analysis:

Map overlay is one of the fundamental operations of GIS (Sadahiro, 2004). Overlay

Analysis tools allow the user to apply weights to several inputs and combine them into a

single output. In other words, it applies a common scale of values to diverse and

dissimilar input to create an integrated analysis (ArcGIS Desktop Help 9.2). Overlay

operation not only merges all line work, but also the attributes of the features taking part

in the overlay are carried through to create a new polygon.

Thiessen polygon and Buffer polygons (1, 2 and 5 miles) layers are overlaid on

the data layer of Census Block groups. The Block group level data layer is associated

with information of percentage of Whites, Blacks, Native Americans, Asians, Pacific

Islanders, Other Races, Two Plus Races and Average Median Household Income. The

Union tool has been used to carry out an overlay of generated Thiessen polygons and

buffer polygons over the Census block group map (reporting demographic and household

income data).

Union calculates the geometric intersection of any number of feature class and

layers. For doing a union analysis, all inputs must be of a common geometry type and the

output will be of that same geometry type. The output features have the attributes of all

the input features that are ‘unioned’ together (ArcGIS Desktop Help 9.2).
 59

Union tool in ArcGIS does the following:

(i.) It determines the spatial reference of the output. All the input feature

classes and layers are projected into this spatial reference.

(ii.) Union cracks and clusters the features. The cracking process inserts

vertices at the intersection of feature edges and clustering snaps together

vertices that are within the XY tolerance.

(iii.) A geometric relationship is formed between features of the various feature

classes and layers that are overlaid.

4.2.3 Regression Analysis:

Assuming that Thiessen polygons or Buffer polygons are a reasonable representation of

trade areas, it is then possible to summarize the socio-economic and other demographic

attributes for the fast-food restaurants available in the study over Thiessen polygons and

Buffer polygons.

Regression is a statistical technique that has been used to quantify the level of

change in the outcome of dependent variable that would be expected based upon a given

level of change in one or more independent variables (Skrepnek, 2005). Linear regression

describes the trend of changes in values of a dependent variable as a straight-line function

with respect to one or more independent variables. Simple linear regression refers to a

case where in a linear relationship is analyzed between one dependent variable and one

independent variable, whereas multiple regressions involve more than one explanatory

(independent) variable.
 60

Multi-variate Regression models the degree to which the variation within values

of a dependent variable measured at ratio scale (for example, Annual Sales of Burger

King and McDonald’s Restaurants in Portage and Summit Counties) as explained by

predictors such as the percentage to total population of Whites, Blacks, Native

Americans, Asians, Pacific Islanders, Others, Two Plus races and median household

income. The multivariate linear regression model assumes that there is a direct linear

relationship between the dependent variable and each predictor. This relationship is

described as:

Y = b0 + b1 x1 + b2 x 2 + ... + b p x p + e
Where: -

Y = the value of the ith case of the dependent scale variable

p = no. of predictors

bi = the ith regression coefficient, i = 0,…,p

xi = the ith predictor

e = error term

This regression model is referred to as linear because increasing the value of the

ith predictor by one unit increases the value of the dependent by bi units. The intercept, b0 ,

is the value of dependent variable when all predictors are set to 0.

The validity of a regression model is often evaluated by the value of the

coefficient of determination, or know otherwise as R 2 . It gives the degree to which

variation among values in the dependent variable explained by the combined variation of
 61

the predictor variables (independent variables). For example, if R 2 = 0.53 , the regression

model is said to have 53% of the variability in the dependent variable being explained by

the variation among the independent variables.

In terms of statistical testing, regression models are often tested with a hypothesis

of no statistical significance. This is often translated to

Ho : R2 = 0

Which means that the variation among values of the dependent variable does not have a

statistically significant relationship with the variation among values in the predictor

variables.

Beyond testing the statistical significance of R 2 , it is possible to test the

significance of the model parameters. These would include the regression coefficients

and the error term. The concepts of this statistical testing would also follow the same

general format as that of the hypothesis for testing R 2 .

For the purpose of testing hypotheses about the values of model parameters, the

linear regression model assumes the following:

a.) The error term has a normal distribution with a mean of zero.

b.) The variance of the error term is constant across cases and independent of the

variables in the model

c.) The value of the error term for a given case is independent of the values of the

variables in the model and of the values of the error term for other cases.

In addition to examining the relationship between dependent and independent variables

by regression models, it should be noted that regression models could take one of many
 62

forms. For this study, both Enter and stepwise regression models are used. The stepwise

regression model is used here for two reasons:

a.) That no pre-conceptualized relative importance of each independent

variable will be assumed or used. This is to avoid any bias that may be

introduced by the analyst.

b.) That the stepwise regression model reveals the importance of each

individual independent variable one at a time. This will allow analysis be

carried out to compare how each individual independent variable perform

in different regression models run for this study.

The statistical analysis in this thesis research uses the SPSS (Statistical Package

for Social Science) of SPSS Inc. (233 S. Wacher Dr, Chicago, IL 60606). SPSS is widely

used among academic and professional analysts for its robustness and accuracy of

statistical calculations. Furthermore, Kent State University holds a site license of SPSS

that enables faculty and students to access the package for academic studies. The

paragraphs below discuss the output from multivariate regression analysis by SPSS. For

the purpose of illustration, a sample output is included here to facilitate the discussion.

The following tables of the statistical output from SPSS are considered for this study:
 63

Model Summary Table:

Model Summary(b)

Adjusted R Std. Error of

Model R R Square Square the Estimate
1
.341(a) .117 .014 523.998

b Dependent Variable: ANN_SALE

a Predictors: (Constant), Ave_MEDHHI, PerASIAN, PerPACIF, PerNATIVE, PerBLACK,
PerOTHRS, PerTWOPL

In the model summary R gives the strength of correlation among included variables. As

mentioned earlier, R-squared, or R 2 is R that has been squared. It represents the total

amount of variance in the dependent variable accounted for by the independent variables.

The value of R 2 can be interpreted as the proportion of variance explained by moving the

decimal point two places to the right and expressing this value as a percentage. The

coefficient of determination is R-squared. R 2 gives an indication of how good a choice

the independent variables are in predicting the dependent variable. The larger the value

the better the regression line describes the data. Adjusted R square is computed using the

formula:

1−
(1 − R )(N − 1)
2

N − K −1
When the number of observations is small and the number of predictors (k) is

large, there will be a much greater difference between R Square and Adjusted R Square

(because the ratio of (N-1)/ (N-k-1) and will be much less than 1). By contrast, when the
 64

number of observations is very large as compared to the number of predictors, the value

of R square and adjusted R square will be much closer as the ratio of (N-1)/ (N-k-1) will

approach 1.

Analysis of Variance (ANOVA) Table:

ANOVA(b)

Sum of
Model Squares df Mean Square F Sig.
1 Regression 2174266.862 7 310609.552 1.131 .356(a)
Residual 16474463.770 60 274574.396
Total 18648730.632 67
a Predictors: (Constant), Ave_MEDHHI, PerASIAN, PerPACIF, PerNATIVE, PerBLACK,
PerOTHRS, PerTWOPL
b Dependent Variable: ANN_SALE

ANOVA is a multivariate regression output that examines the variability, which can be

applied to look at the total amount of variance in the variance among the dependent

variables. It also gives clues to how much of that variance is accounted for by the

independent variables. The significance of the value F (called Sig. In the table) is the

probability associated with R 2 . This probability can be regarded of as a significance

value for the whole model or a significant value of R 2 (Miles and Shevlin, 2001).
 65

Coefficient Table (Parameter Estimates):

Coefficients(a)

Unstandardized Standardized
Coefficients Coefficients

Model B Std. Error Beta t Sig.

1 (Constant) 1276.802 582.794 2.191 .032
PerBLACK .630 6.909 .015 .091 .928
PerNATIVE -714.970 583.791 -.175 -1.225 .225
PerASIAN -42.460 104.981 -.075 -.404 .687
PerPACIF 2122.583 1549.094 .195 1.370 .176
PerOTHRS -435.873 402.795 -.181 -1.082 .284
PerTWOPL 160.076 162.492 .234 .985 .329
Ave_MEDHHI .003 .009 .077 .332 .741

a. Dependent Variable: ANN_SALE

In the Coefficient Tables the following parameters have been used for the study:

t and Sig.:

These columns provide the t-value and 2-tailed p-value used in testing the null hypothesis

that the coefficient/ parameter is 0. The p-value is the calculated probability level of the

tested parameter. If using a 2-tailed test, each p-value is compared to the pre-selected

value of α – a pre-defined level of statistical significance. Coefficients having p-values

less than the pre-defined α are statistically significant. Following a conventional level of

statistical significance, this study also uses α = 0.05 . Therefore, regression coefficients

having a p-value of 0.05 or less are said to be statistically significantly different from 0

(i.e., the null hypothesis can be rejected and it can be said that the coefficient is

statistically significantly different from 0).

 66

A comparative analysis of the R 2 of the Thiessen polygons and Buffer polygons with

three different buffering limits has been done to see how well or poorly either of the two

approaches captures the variations in the sales volumes of fast food stores in the study

area.

In addition a comparative analysis of the R2 of three buffer limits has also been

performed to determine a near ideal buffering limit for each of the fast food retail stores

4.3 Study Area

A total of 65 fast-food restaurants in Summit County and Portage County of Ohio are

included in this study. These 65 restaurants are all the McDonalds and the Burger King

restaurants in these two counties. These two brands are chosen because they are the top

two leaders in fast food industry in the region and the food items they serve are similar

and, to a great extent, substituted easily. There have been various reasons why this

particular set of counties has been chosen to apply the retail location analysis model. A

variety of socio-economic and demographic attributes are found to be consistent with

those of the state as well as those of the national averages. These counties also represent a

good mix of urban, rural as well as suburban lifestyles (Joseph, 2005). Hence, a retail

location analysis done for these counties can also be applied to several other retail

markets across the country.

The use of McDonalds and Burger Kings as study subjects is based on the assumption

that quality of food at both the chains is expected to be standard at each of their stores

(McDonalds’ Burger will be the same quality at all McDonalds outlets and similarly for
 67

Burger King as well). With the quality of products controlled, the comparative studies of

how socio-economic and demographic factors have led us to understand how markets

with different location factors perform. These results can be applied to other types of

businesses as well, if product quality control can be achieved.

4.4 Summary

Trade area analysis has been an important and popular concept with retail businesses.

GIS tools and statistical analysis can be used to calculate the size and potential of the

market based on the geodemographic profile of the trade area. This study aims at

developing a methodology for analyzing the relationship between fast food store

performance and the demographic and socio-economic variables in the outlets’ catchment

areas in Portage and Summit Counties, Ohio. These counties represent a good mix of

urban, rural as well as suburban lifestyles and therefore, the location analysis can be

applied to several other retail markets across the United States.

The locations of Burger King and McDonald’s QSR have been geocoded and a

catchment area analysis has been done for these restaurants by constructing Thiessen and

buffer polygons around the locations. Geoprocessing environment including overlay and

proximity analysis has been made use of to see how the ethnic composition of population

and the median household income in trade areas are related to the performance of the

QSR. The statistical technique of regression analysis has been used to quantify the level

of change in the outcome of Annual Sales of the restaurants that would be based upon a

given level of change in percentage of ethnic populations and the median household
 68

income of the trade areas. In the end, a comparative analysis of the coefficient of

determination ( R 2 ) of Thiessen and buffer polygons of different buffering units show

how well or poorly either of these two catchment analysis approaches capture variations

in sales volumes of the QSR. This comparative analysis also helps to determine an ideal

buffering limit for each of the fast food restaurants.

 Chapter 5

Analysis and Discussions

2
The Coefficient of Determination, or R is the percent of the Total Sum of Squares that

is explained: regression Sum of Squares divided by Total Sum of Squares. The

denominator is fixed and the numerator can only increase. Each additional variable used

2
in the equation, increases the numerator at least slightly, resulting in a higher R even

when the new variable causes the equation to become less efficient.

2
The adjusted R represents the result of adjusting both the numerator and

2
denominator by their respective degrees of freedom. Adjusted R can decline in value if

the contribution to the explained deviation by the additional variable is less than the

2
impact on the degrees of freedom. The adjusted R will react to alternative equations for

the same dependent variable in a manner similar to the Standard Error of the estimate, i.e.

the equation with the smallest Standard Error of the Estimate will most likely also have

2 2 2
the highest adjusted R . Unlike R , adjusted R is referred to as an index value and

not a percept (Jensen, 2006). The statistical models have been tabulated based on spatial

configuration of Buffers 1, 2, 5 miles and Thiessen, Restaurants within each County,

Restaurant-wise and County-wise (Tables2-5).

69
 70

Table 2: Spatial Configuration (Regression Analysis: Enter Method: S-Summit, P-Portage, M-McDonald’s,
B-Burger King)
(. – Excluded Variables)
Buffe r 1 m ile
M ode l Buffe r N R2 Adj_R2 F Prob(F) Cons tant %White %Black %Native %As ian %Pacific %Othe rs %Tw opl Avg_M e dInc
BP 1-mile 5 1.00 . . . 785.85 . -9.81 . . . 564.77 -149.34 0.01
BS 1-mi l e 19 0.75 0.59 4.70 0.01 1020.52 . 2.90 412.86 58.52 1196.38 75.21 72.13 0.00
BPS 1-mi l e 24 0.58 0.39 3.10 0.03 583.42 . 6.22 131.59 -49.98 1285.68 271.74 105.85 0.01
BMP 1-mile 14 0.67 0.29 1.76 0.25 12.69 . -43.62 -690.53 172.77 5406.49 -1142.84 645.68 0.02
MP 1-mile 9 0.89 0.08 1.10 1.10 1003.46 . -171.09 1793.71 -37.19 5446.70 910.34 434.63 0.01
MPS 1-mile 41 0.11 -0.08 0.56 0.78 1498.37 . 3.34 -309.47 -106.25 134.18 166.64 23.17 0.01
BMPS 1-mile 64 0.15 0.04 1.38 0.23 1059.84 . 1.39 -105.10 -119.13 1064.29 9.52 135.64 0.01
BMS 1-mile 51 0.05 -0.10 0.36 0.92 1307.01 . 4.20 -41.33 -54.41 1452.19 6.62 47.71 0.00
MS 1-mile 32 0.10 -0.16 0.37 0.91 1840.61 . 5.97 -270.08 -35.69 265.56 -1.82 -79.29 0.00
Buffe r 2 m ile s
M ode l Buffe r N R2 Adj_R2 F Prob(F) Cons tant %White %Black %Native %As ian %Pacific %Othe rs %Tw opl Avg_M e dInc
BP 2-mile 5 1.00 . . . -340.33 . -5.28 790.08 . . 465.54 0.02
BS 2-mi l e 19 0.78 0.64 5.55 0.01 1805.82 -5.18 . 352.26 77.62 1041.00 -427.44 105.27 -0.01
BPS 2-mi l e 24 0.65 0.50 4.34 0.01 1390.14 -6.01 . -469.05 -4.70 1852.82 -301.94 207.40 0.00
MP 2-mile 9 0.93 0.46 1.99 0.50 1902.26 . -227.38 3460.77 -856.14 -20051.19 4209.38 80.15 0.00
MPS 2-mile 41 0.12 -0.06 0.67 0.70 1907.97 -4.99 . -285.72 -155.52 -145.62 275.00 -5.90 0.01
BMP 2-mile 14 0.73 0.41 2.31 0.16 -1172.98 . 5.97 -1293.89 -284.92 14240.75 237.58 877.64 0.04
BMS 2-mile 51 0.07 -0.08 0.46 0.86 2487.62 -8.32 -203.00 -50.69 -184.05 -399.43 -32.10 0.00
BMPS 2-mile 64 0.17 0.07 1.68 0.13 1586.19 -6.33 -182.40 -176.96 1553.99 -399.05 161.59 0.01
MS 2-mile 32 0.14 -0.11 0.57 0.78 1819.46 7.46 -420.01 -76.28 209.04 27.71 -111.01 0.00
Buffe r 5 m ile s
M ode l Buffe r N R2 Adj_R2 F Prob(F) Cons tant %White %Black %Native %As ian %Pacific %Othe rs %Tw opl Avg_M e dInc
BP 5-mile 5 1.00 . . . 33.19 . . -816.71 . . -1251.61 1077.02 0.00
MP 5-mile 9 0.99 0.95 25.17 0.15 10683.60 . -2130.85 20024.62 2905.38 -116642.80 -3407.87 -3040.52 -0.08
MPS 5-mile 41 0.18 0.01 1.03 0.43 -27.03 . -5.13 1266.09 -202.24 1309.98 -702.16 485.61 0.03
BPS 5-mile 24 0.43 0.18 1.73 0.17 416.40 . -5.25 178.55 116.92 933.62 107.62 278.30 0.00
BS 5-mile 19 0.42 0.05 1.12 0.41 400.44 . -5.43 9.22 158.08 470.97 904.65 145.20 0.00
MS 5-mile 32 0.25 0.03 1.15 0.36 842.03 . 5.13 1494.91 13.74 1701.17 -2476.06 503.33 0.01
BMPS 5-mile 64 0.10 -0.01 0.91 0.51 2524.21 . 3.05 -566.60 173.71 38.91 -1393.99 27.93 -0.02
BMS 5-mile 51 0.11 -0.04 0.76 0.63 2287.86 . 4.54 -242.55 225.52 1036.14 -1799.99 115.35 -0.01
BMP 5-mile 14 0.48 -0.13 0.79 0.62 -866.07 . 374.09 -3673.27 -666.50 69201.60 -2242.17 456.17 0.04
Thie s s e n
M ode l Buffe r N R2 Adj_R2 F Prob(F) Cons tant %White %Black %Native %As ian %Pacific %Othe rs %Tw opl Avg_M e dInc
BP Theissen 5 1.00 . . . 559.51 . . . . -8595.10 -157.85 55.24 0.01
MP Theissen 9 0.95 0.60 2.72 0.44 2065.87 . -716.37 -1316.26 1327.26 -51145.21 1789.96 1387.65 -0.01
MPS Theissen 41 0.16 -0.02 0.86 0.55 2232.98 . 5.93 -622.88 19.19 727.01 -436.58 -126.28 -0.01
BS Thei ssen 19 0.72 0.54 4.02 0.02 842.65 . 8.87 482.39 1.08 1251.79 -305.08 81.89 0.00
BPS Thei ssen 24 0.66 0.52 4.50 0.01 750.27 . 10.94 295.41 -20.82 1517.22 -192.28 71.05 0.00
BMP Theissen 14 0.76 0.47 2.65 0.13 3752.01 . -36.51 -3067.44 175.59 -36458.61 -1082.73 -270.41 -0.02
BMPS Theissen 64 0.17 0.06 1.59 0.16 1503.71 . 5.63 -579.09 -99.25 1037.93 -191.28 51.56 0.00
BMS Theissen 51 0.10 -0.05 0.69 0.68 1417.84 . 7.42 -414.44 -85.96 832.62 -124.20 39.75 0.00
MS Theissen 32 0.19 -0.05 0.78 0.61 1918.09 . 6.55 -634.38 -77.36 759.64 -222.54 -62.37 0.00
 71

Table 3: Restaurant wise (Regression Analysis: Enter Method: S-Summit, P-Portage, M-McDonald’s, B-Burger King)
Burger King & McDonald's
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BMP 1-mile 14 0.67 0.29 1.76 0.25 12.69 . -43.62 -690.53 172.77 5406.49 -1142.84 645.68 0.02
BMP 2-mile 14 0.73 0.41 2.31 0.16 -1172.98 . 5.97 -1293.89 -284.92 14240.75 237.58 877.64 0.04
BMP 5-mile 14 0.48 -0.13 0.79 0.62 -866.07 . 374.09 -3673.27 -666.50 69201.60 -2242.17 456.17 0.04
BMP Theissen 14 0.76 0.47 2.65 0.13 3752.01 . -36.51 -3067.44 175.59 -36458.61 -1082.73 -270.41 -0.02
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BMPS 1-mile 64 0.15 0.04 1.38 0.23 1059.84 . 1.39 -105.10 -119.13 1064.29 9.52 135.64 0.01
BMPS 2-mile 64 0.17 0.07 1.68 0.13 1586.19 -6.33 -182.40 -176.96 1553.99 -399.05 161.59 0.01
BMPS 5-mile 64 0.10 -0.01 0.91 0.51 2524.21 . 3.05 -566.60 173.71 38.91 -1393.99 27.93 -0.02
BMPS Theissen 64 0.17 0.06 1.59 0.16 1503.71 . 5.63 -579.09 -99.25 1037.93 -191.28 51.56 0.00
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BMS 1-mile 51 0.05 -0.10 0.36 0.92 1307.01 . 4.20 -41.33 -54.41 1452.19 6.62 47.71 0.00
BMS 2-mile 51 0.07 -0.08 0.46 0.86 2487.62 -8.32 -203.00 -50.69 -184.05 -399.43 -32.10 0.00
BMS 5-mile 51 0.11 -0.04 0.76 0.63 2287.86 . 4.54 -242.55 225.52 1036.14 -1799.99 115.35 -0.01
BMS Theissen 51 0.10 -0.05 0.69 0.68 1417.84 . 7.42 -414.44 -85.96 832.62 -124.20 39.75 0.00
Burger King
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BP 1-mile 5 1.00 . . . 785.85 . -9.81 . . . 564.77 -149.34 0.01
BP 2-mile 5 1.00 . . . -340.33 . -5.28 790.08 . . 465.54 0.02
BP 5-mile 5 1.00 . . . 33.19 . . -816.71 . . -1251.61 1077.02 0.00
BP Theissen 5 1.00 . . . 559.51 . . . . -8595.10 -157.85 55.24 0.01
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BPS 1-mile 24 0.58 0.39 3.10 0.03 583.42 . 6.22 131.59 -49.98 1285.68 271.74 105.85 0.01
BPS 2-mile 24 0.65 0.50 4.34 0.01 1390.14 -6.01 . -469.05 -4.70 1852.82 -301.94 207.40 0.00
BPS 5-mile 24 0.43 0.18 1.73 0.17 416.40 . -5.25 178.55 116.92 933.62 107.62 278.30 0.00
BPS Theissen 24 0.66 0.52 4.50 0.01 750.27 . 10.94 295.41 -20.82 1517.22 -192.28 71.05 0.00
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BS 1-mile 19 0.75 0.59 4.70 0.01 1020.52 . 2.90 412.86 58.52 1196.38 75.21 72.13 0.00
BS 2-mile 19 0.78 0.64 5.55 0.01 1805.82 -5.18 . 352.26 77.62 1041.00 -427.44 105.27 -0.01
BS 5-mile 19 0.42 0.05 1.12 0.41 400.44 . -5.43 9.22 158.08 470.97 904.65 145.20 0.00
BS Theissen 19 0.72 0.54 4.02 0.02 842.65 . 8.87 482.39 1.08 1251.79 -305.08 81.89 0.00
McDonald's
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
MP 1-mile 9 0.89 0.08 1.10 1.10 1003.46 . -171.09 1793.71 -37.19 5446.70 910.34 434.63 0.01
MP 2-mile 9 0.93 0.46 1.99 0.50 1902.26 . -227.38 3460.77 -856.14 -20051.19 4209.38 80.15 0.00
MP 5-mile 9 0.99 0.95 25.17 0.15 10683.60 . -2130.85 20024.62 2905.38 -116642.80 -3407.87 -3040.52 -0.08
MP Theissen 9 0.95 0.60 2.72 0.44 2065.87 . -716.37 -1316.26 1327.26 -51145.21 1789.96 1387.65 -0.01
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
MPS 1-mile 41 0.11 -0.08 0.56 0.78 1498.37 . 3.34 -309.47 -106.25 134.18 166.64 23.17 0.01
MPS 2-mile 41 0.12 -0.06 0.67 0.70 1907.97 -4.99 . -285.72 -155.52 -145.62 275.00 -5.90 0.01
MPS 5-mile 41 0.18 0.01 1.03 0.43 -27.03 . -5.13 1266.09 -202.24 1309.98 -702.16 485.61 0.03
MPS Theissen 41 0.16 -0.02 0.86 0.55 2232.98 . 5.93 -622.88 19.19 727.01 -436.58 -126.28 -0.01
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
MS 1-mile 32 0.10 -0.16 0.37 0.91 1840.61 . 5.97 -270.08 -35.69 265.56 -1.82 -79.29 0.00
MS 2-mile 32 0.14 -0.11 0.57 0.78 1819.46 7.46 -420.01 -76.28 209.04 27.71 -111.01 0.00
MS 5-mile 32 0.25 0.03 1.15 0.36 842.03 . 5.13 1494.91 13.74 1701.17 -2476.06 503.33 0.01
MS Theissen 32 0.19 -0.05 0.78 0.61 1918.09 . 6.55 -634.38 -77.36 759.64 -222.54 -62.37 0.00
 72

Table 4: County wise (Regression Analysis: Enter Method: S-Summit, P-Portage, M-McDonald’s, B-Burger King)
Portage County
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BMP 1-mile 14 0.67 0.29 1.76 0.25 12.69 . -43.62 -690.53 172.77 5406.49 -1142.84 645.68 0.02
BMP 2-mile 14 0.73 0.41 2.31 0.16 -1172.98 . 5.97 -1293.89 -284.92 14240.75 237.58 877.64 0.04
BMP 5-mile 14 0.48 -0.13 0.79 0.62 -866.07 . 374.09 -3673.27 -666.50 69201.60 -2242.17 456.17 0.04
BMP Theissen 14 0.76 0.47 2.65 0.13 3752.01 . -36.51 -3067.44 175.59 -36458.61 -1082.73 -270.41 -0.02
BP 1-mile 5 1.00 . . . 785.85 . -9.81 . . . 564.77 -149.34 0.01
BP 2-mile 5 1.00 . . . -340.33 . -5.28 790.08 . . 465.54 0.02
BP 5-mile 5 1.00 . . . 33.19 . . -816.71 . . -1251.61 1077.02 0.00
BP Theissen 5 1.00 . . . 559.51 . . . . -8595.10 -157.85 55.24 0.01
MP 1-mile 9 0.89 0.08 1.10 1.10 1003.46 . -171.09 1793.71 -37.19 5446.70 910.34 434.63 0.01
MP 2-mile 9 0.93 0.46 1.99 0.50 1902.26 . -227.38 3460.77 -856.14 -20051.19 4209.38 80.15 0.00
MP 5-mile 9 0.99 0.95 25.17 0.15 10683.60 . -2130.85 20024.62 2905.38 -116642.80 -3407.87 -3040.52 -0.08
MP Theissen 9 0.95 0.60 2.72 0.44 2065.87 . -716.37 -1316.26 1327.26 -51145.21 1789.96 1387.65 -0.01
Summit County
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BMS 1-mile 51 0.05 -0.10 0.36 0.92 1307.01 . 4.20 -41.33 -54.41 1452.19 6.62 47.71 0.00
BMS 2-mile 51 0.07 -0.08 0.46 0.86 2487.62 -8.32 -203.00 -50.69 -184.05 -399.43 -32.10 0.00
BMS 5-mile 51 0.11 -0.04 0.76 0.63 2287.86 . 4.54 -242.55 225.52 1036.14 -1799.99 115.35 -0.01
BMS Theissen 51 0.10 -0.05 0.69 0.68 1417.84 . 7.42 -414.44 -85.96 832.62 -124.20 39.75 0.00
BS 1-mile 19 0.75 0.59 4.70 0.01 1020.52 . 2.90 412.86 58.52 1196.38 75.21 72.13 0.00
BS 2-mile 19 0.78 0.64 5.55 0.01 1805.82 -5.18 . 352.26 77.62 1041.00 -427.44 105.27 -0.01
BS 5-mile 19 0.42 0.05 1.12 0.41 400.44 . -5.43 9.22 158.08 470.97 904.65 145.20 0.00
BS Theissen 19 0.72 0.54 4.02 0.02 842.65 . 8.87 482.39 1.08 1251.79 -305.08 81.89 0.00
MS 1-mile 32 0.10 -0.16 0.37 0.91 1840.61 . 5.97 -270.08 -35.69 265.56 -1.82 -79.29 0.00
MS 2-mile 32 0.14 -0.11 0.57 0.78 1819.46 7.46 -420.01 -76.28 209.04 27.71 -111.01 0.00
MS 5-mile 32 0.25 0.03 1.15 0.36 842.03 . 5.13 1494.91 13.74 1701.17 -2476.06 503.33 0.01
MS Theissen 32 0.19 -0.05 0.78 0.61 1918.09 . 6.55 -634.38 -77.36 759.64 -222.54 -62.37 0.00
Portage & Summit Counties
Model Buffer N R2 Adj_R2 F Prob(F) Constant %White %Black %Native %Asian %Pacific %Others %Twoplvg_MedInc
BMPS 1-mile 64 0.15 0.04 1.38 0.23 1059.84 . 1.39 -105.10 -119.13 1064.29 9.52 135.64 0.01
BMPS 2-mile 64 0.17 0.07 1.68 0.13 1586.19 -6.33 -182.40 -176.96 1553.99 -399.05 161.59 0.01
BMPS 5-mile 64 0.10 -0.01 0.91 0.51 2524.21 . 3.05 -566.60 173.71 38.91 -1393.99 27.93 -0.02
BMPS Theissen 64 0.17 0.06 1.59 0.16 1503.71 . 5.63 -579.09 -99.25 1037.93 -191.28 51.56 0.00
BPS 1-mile 24 0.58 0.39 3.10 0.03 583.42 . 6.22 131.59 -49.98 1285.68 271.74 105.85 0.01
BPS 2-mile 24 0.65 0.50 4.34 0.01 1390.14 -6.01 . -469.05 -4.70 1852.82 -301.94 207.40 0.00
BPS 5-mile 24 0.43 0.18 1.73 0.17 416.40 . -5.25 178.55 116.92 933.62 107.62 278.30 0.00
BPS Theissen 24 0.66 0.52 4.50 0.01 750.27 . 10.94 295.41 -20.82 1517.22 -192.28 71.05 0.00
MPS 1-mile 41 0.11 -0.08 0.56 0.78 1498.37 . 3.34 -309.47 -106.25 134.18 166.64 23.17 0.01
MPS 2-mile 41 0.12 -0.06 0.67 0.70 1907.97 -4.99 . -285.72 -155.52 -145.62 275.00 -5.90 0.01
MPS 5-mile 41 0.18 0.01 1.03 0.43 -27.03 . -5.13 1266.09 -202.24 1309.98 -702.16 485.61 0.03
MPS Theissen 41 0.16 -0.02 0.86 0.55 2232.98 . 5.93 -622.88 19.19 727.01 -436.58 -126.28 -0.01
 73

Table 5: Restaurant wise and County wise (Regression Analysis: Enter Method: S-Summit, P-Portage,
M-McDonald’s, B-Burger King)
Portage County
Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
BMP 1-mile 14 0.67 0.29 1.76 0.25 12.69 . -43.62 -690.53 172.77 5406.49 -1142.84 645.68 0.02
BMP 2-mile 14 0.73 0.41 2.31 0.16 -1172.98 . 5.97 -1293.89 -284.92 14240.75 237.58 877.64 0.04
BMP 5-mile 14 0.48 -0.13 0.79 0.62 -866.07 . 374.09 -3673.27 -666.50 69201.60 -2242.17 456.17 0.04
BMP Theissen 14 0.76 0.47 2.65 0.13 3752.01 . -36.51 -3067.44 175.59 -36458.61 -1082.73 -270.41 -0.02

Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
BP 1-mile 5 1.00 . . . 785.85 . -9.81 . . . 564.77 -149.34 0.01
BP 2-mile 5 1.00 . . . -340.33 . -5.28 790.08 . . 465.54 0.02
BP 5-mile 5 1.00 . . . 33.19 . . -816.71 . . -1251.61 1077.02 0.00
BP Theissen 5 1.00 . . . 559.51 . . . . -8595.10 -157.85 55.24 0.01

Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
MP 1-mile 9 0.89 0.08 1.10 1.10 1003.46 . -171.09 1793.71 -37.19 5446.70 910.34 434.63 0.01
MP 2-mile 9 0.93 0.46 1.99 0.50 1902.26 . -227.38 3460.77 -856.14 -20051.19 4209.38 80.15 0.00
MP 5-mile 9 0.99 0.95 25.17 0.15 10683.60 . -2130.85 20024.62 2905.38 -116642.80 -3407.87 -3040.52 -0.08
MP Theissen 9 0.95 0.60 2.72 0.44 2065.87 . -716.37 -1316.26 1327.26 -51145.21 1789.96 1387.65 -0.01
Summit County
Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
BMS 1-mile 51 0.05 -0.10 0.36 0.92 1307.01 . 4.20 -41.33 -54.41 1452.19 6.62 47.71 0.00
BMS 2-mile 51 0.07 -0.08 0.46 0.86 2487.62 -8.32 -203.00 -50.69 -184.05 -399.43 -32.10 0.00
BMS 5-mile 51 0.11 -0.04 0.76 0.63 2287.86 . 4.54 -242.55 225.52 1036.14 -1799.99 115.35 -0.01
BMS Theissen 51 0.10 -0.05 0.69 0.68 1417.84 . 7.42 -414.44 -85.96 832.62 -124.20 39.75 0.00

Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
BS 1-mile 19 0.75 0.59 4.70 0.01 1020.52 . 2.90 412.86 58.52 1196.38 75.21 72.13 0.00
BS 2-mile 19 0.78 0.64 5.55 0.01 1805.82 -5.18 . 352.26 77.62 1041.00 -427.44 105.27 -0.01
BS 5-mile 19 0.42 0.05 1.12 0.41 400.44 . -5.43 9.22 158.08 470.97 904.65 145.20 0.00
BS Theissen 19 0.72 0.54 4.02 0.02 842.65 . 8.87 482.39 1.08 1251.79 -305.08 81.89 0.00

Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
MS 1-mile 32 0.10 -0.16 0.37 0.91 1840.61 . 5.97 -270.08 -35.69 265.56 -1.82 -79.29 0.00
MS 2-mile 32 0.14 -0.11 0.57 0.78 1819.46 7.46 -420.01 -76.28 209.04 27.71 -111.01 0.00
MS 5-mile 32 0.25 0.03 1.15 0.36 842.03 . 5.13 1494.91 13.74 1701.17 -2476.06 503.33 0.01
MS Theissen 32 0.19 -0.05 0.78 0.61 1918.09 . 6.55 -634.38 -77.36 759.64 -222.54 -62.37 0.00
Portage & Summit Counties
Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
BMPS 1-mile 64 0.15 0.04 1.38 0.23 1059.84 . 1.39 -105.10 -119.13 1064.29 9.52 135.64 0.01
BMPS 2-mile 64 0.17 0.07 1.68 0.13 1586.19 -6.33 -182.40 -176.96 1553.99 -399.05 161.59 0.01
BMPS 5-mile 64 0.10 -0.01 0.91 0.51 2524.21 . 3.05 -566.60 173.71 38.91 -1393.99 27.93 -0.02
BMPS Theissen 64 0.17 0.06 1.59 0.16 1503.71 . 5.63 -579.09 -99.25 1037.93 -191.28 51.56 0.00

Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
BPS 1-mile 24 0.58 0.39 3.10 0.03 583.42 . 6.22 131.59 -49.98 1285.68 271.74 105.85 0.01
BPS 2-mile 24 0.65 0.50 4.34 0.01 1390.14 -6.01 . -469.05 -4.70 1852.82 -301.94 207.40 0.00
BPS 5-mile 24 0.43 0.18 1.73 0.17 416.40 . -5.25 178.55 116.92 933.62 107.62 278.30 0.00
BPS Theissen 24 0.66 0.52 4.50 0.01 750.27 . 10.94 295.41 -20.82 1517.22 -192.28 71.05 0.00

Model Buffer N R2 Adj_R2 F Prob(F) Constant % White % Black % Native % Asian % Pacific % Others % Tw opl Avg_MedInc
MPS 1-mile 41 0.11 -0.08 0.56 0.78 1498.37 . 3.34 -309.47 -106.25 134.18 166.64 23.17 0.01
MPS 2-mile 41 0.12 -0.06 0.67 0.70 1907.97 -4.99 . -285.72 -155.52 -145.62 275.00 -5.90 0.01
MPS 5-mile 41 0.18 0.01 1.03 0.43 -27.03 . -5.13 1266.09 -202.24 1309.98 -702.16 485.61 0.03
MPS Theissen 41 0.16 -0.02 0.86 0.55 2232.98 . 5.93 -622.88 19.19 727.01 -436.58 -126.28 -0.01
 74

Examining the adjusted R 2 of the models (Tables 2-5) show that Burger King’s

annual sales revenues are better explained by the included independent variables than

those of McDonald’s sales by the same set of variables. This is apparent by looking at the

adjusted R 2 of the models for Buffer-1 mile for Burger King in Summit County and

Burger King in Portage and Summit Counties, Buffer-2 miles for Burger King in Summit

County and Burger King restaurants in Portage and Summit Counties.

However, this trend is not apparent for the Buffer-5 mile around Burger King

restaurants but sales are better explained by the variables for McDonald’s restaurants in

Portage and McDonald’s in Portage and Summit Counties. For the Thiessen buffers,

annual sales figures for McDonald’s in Portage and McDonald’s taken together for both

counties are better explained by the independent variables than those for Burger King.

The models for Burger King in Portage County have a very small sample size; hence

their statistical analysis is unable to project any relevant results.

Looking at the ethnic percentage of population in the buffers, it is seen that in

most of the cases, for every unit decrease in the percentage of Black population, there has

been a relative increase in Annual Sales of the fast food restaurants. This trend is more

pronounced in Portage County as compared to Summit County. Summit County has a

higher mean of average median household income ($ 47,176) than Portage County ($

46,138) (US Census Bureau, 2001). One can infer that the black population is not

necessarily the largest clientele for these fast food restaurants. This could also point

towards the economic status of the black population as well as their eating preferences.
 75

Taking a look at both Burger King and McDonald’s together, a decrease in the

percentage of black population and an increase in the sales revenues of these restaurants

are recorded for Portage County Buffer-1 mile and Thiessen. For the Native American

population a decrease in their numbers reflect an increase in the sales revenues of the

restaurants. However, looking at the percentage of their population to the total population

of both the counties shows that they are less than 1% in numbers for each buffer around

the restaurants.

A decrease in the percentage of Asian population and an increase in sales

revenues of the restaurants are recorded for both Burger King and McDonald’s in Portage

and Summit Counties within Buffers-1, 2, 5 miles and Thiessen. The percentage of Asian

population in these counties varies from less than 1% to 3% for the different blocks. An

increase in sales revenues as a result of decrease in Asian population reflects that either

their numbers are very small to reflect any significant impact over the sales revenues or

the other possible reason could be that the Asian population prefers to eat out at other

types of restaurants/ ethnic specialty restaurants or cook at home.

With a decrease in the population of all other races (including Hispanics), there is

an increase in sales revenues for both Burger King and McDonald’s in Buffer-1, 2, 5

miles and Thiessen. But for Burger King and McDonald’s (Portage and Summit

Counties) Buffer-1 mile, Burger King and McDonald’s (Summit County alone) Buffer-1

mile show an increase in sales revenues along with an increase in the percentage of

population of all other races (including Hispanics). The percentage of all the other races

(including Hispanics) varies from less than 1% to 1.12%.

 76

5.1 Regression Analysis: Enter Method:

Figures for Burger King Restaurants in Portage County with Buffer-1, 2 and 5

miles and Burger King in Portage and Summit County Buffer-5 miles show that a

decrease in percentage of Black population reflects an increase in sales revenue. Since the

number of Burger King stores in Portage County is only five, it does not give a very clear

picture of the actual trend prevailing in the region. However, an increase in sales

revenues for Burger King in both Portage as well as Summit County could possibly mean

that these restaurants draw clientele from mainly 1 to 2 miles around the restaurants.

Customers either walk to the restaurants or drive for a distance of 1~2 miles.

Burger King in Summit County and Burger King and McDonald’s for Portage and

Summit County for Buffer-2 miles show an increase in sales revenues with a decrease in

percentage of white population. These buffers also show a decrease in percentage of

Native Americans, Asians and all other races (including Hispanics). Such a trend

suggests that most of these stores are probably located away from the residential areas, in

commercial establishments, near office areas or along the highway exits.

McDonald’s restaurants show a rather clear-cut trend in terms of an increase in

Annual Sales as a result of a decrease in percentage of black population in Portage

County for Buffer-1, 2, 5 miles and Thiessen. For McDonald’s in Summit County a

decrease in Asians, Native Americans and all other races, an increase in annual sales is

observed.
 77

From the above observations one can infer that:

• Ethnic population and median household income for Buffers-1 and 2 miles around

the restaurants better explain Annual Sales for Burger King.

• Ethnic population and median income for Buffer-5 miles better explain Annual

Sales for McDonald’s. This could mean that either these restaurants are located at

highway exits or are more accessible to commuters on their way to or from work

or even that the consumers are more willing to travel a longer distance to eat at a

McDonald’s restaurant as compared to a Burger King restaurant.

• White population is the largest clientele for Burger King and McDonald’s

restaurants in Portage and Summit County.

• The other minority races (besides the Black population) in these predominantly

white counties are too less in numbers to reflect any significant effect on the sales

revenues of the fast food restaurants.

• The ethnic minorities have other food preferences and prefer to eat out at other

specialty restaurants or even cook at home.

• The ethnic minorities are either economically weaker or more health conscious

and prefer to not eat at the fast food restaurants very often.

Analyzing the Probability (F) values for these models show that the coefficients

for Burger King (Summit County) Buffer-1 and -2 miles and Thiessen, Burger King

(Portage and Summit Counties) Buffer-1 and -2 miles and Thiessen, are statistically
 78

significant using alpha value of 0.05 as their p-values are less than 0.05. The statistically

significant models are highlighted by gray in the Table no. 2, 3, 4 & 5.

5.2 Regression Analysis: Stepwise Method

Stepwise is one of the most frequently used regression analysis methods. It includes

regression models in which the choice of predictive variables is carried out by how much

variable contributes to increasing the explained variation in the dependent variable. The

statistical software without manual intervention normally does this automatically. The

statistical software program determines which variables among the specified set of

independent variables will actually be used for the regression, and in which order they

will be introduced, beginning with the forced variables and continuing with the other

variables, one by one.

After each step the algorithm selects from the remaining predictor variables the

variable, which yields the largest reduction in the residual (unexplained) variance of the

dependent variable, unless its contribution to the total F-ratio for the regression remains

below a specified threshold. Similarly, the program evaluates after each step whether the

contribution of any variable already included falls below a specified threshold, in which

case it is dropped from the regression.

Stepwise regression was also run for each of these models. However, results were

obtained for only few of them. Statistical results were not yielded for McDonald’s

restaurants unless there were taken with Burger King. These tables show that a decrease
 79

in which ethnic population group has yielded the largest reduction in annual sales of fast

food restaurant.
 80

Table 6: Spatial Configuration (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Buffer 1 mile
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 1 mile 14 PerBlack 0.273 .. ..
BS 1 mile 19 PerWhite 0.476 .. ..
BPS 1 mile 24 PerBlack 0.428 .. ..
Buffer 2 miles
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BS 2 miles 19 PerTwoPl 0.474 PerPacific 0.565
BPS 2 miles 24 PerWhite 0.368 PerPacific 0.487
Buffer 5 miles
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 5 miles 14 PerNative 0.238 .. ..
BP 5 miles 5 AvMedHH 0.799 .. ..
BS 5 miles 19 PerOthers 0.246 .. ..
BPS 5 miles 24 PerTwoPl 0.271 .. ..
Thiessen
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP Thiessen 14 PerNative 0.468 PerPacific 0.634
BS Thiessen 19 PerWhite 0.528 .. ..
BPS Thiessen 24 PerBlack 0.521 PerPacific 0.595
BMPS Thiessen 64 PerAsian 0.066 .. ..

According to the data summarized in Table 6, the percentage of Pacific Islanders

yielded the largest effect on sales revenues of Burger King Restaurants within Buffer-2

miles and Thiessen. Black population shows the largest effect on annual sales of Burger

King restaurants within Buffer-1 mile. Percentage of Whites shows the largest effect on

sales revenues of Burger King Restaurants in Summit County for Buffer-1, 2 miles and

Thiessen. This could be due to that most of these stores are situated away from the

residential areas and are near office areas, commercial establishments or along highways.

Native Americans reflect the largest effect on sales revenues in Portage County for both

chains in Buffer 5 miles and Thiessen.

 81

Table 7: Restaurant wise (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Burger King
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BPS 1 mile 24 PerBlack 0.428
BS 1 mile 19 PerWhite 0.476
BPS 2 miles 24 PerWhite 0.368 PerPacific 0.487
BS 2 miles 19 PerTwoPl 0.474 PerPacific 0.565
BP 5 miles 5 AvMedHH 0.799
BPS 5 miles 24 PerTwoPl 0.271
BS 5 miles 19 PerOthers 0.246
BPS Thiessen 24 PerBlack 0.521 PerPacific 0.595
BS Thiessen 19 PerWhite 0.528
Burger King and McDonald's
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 1 mile 14 PerBlack 0.273
BMP 5 miles 14 PerNative 0.238
BMP Thiessen 14 PerNative 0.468 PerPacific 0.634
BMPS Thiessen 64 PerAsian 0.066

Table 7 shows that, Pacific Islanders is the most common ethnic minority group

that has shown the largest effect on sales revenues of Burger King due to a decrease in

their numbers. The White populations and the Blacks follow them. For both Burger

Kings and McDonalds’ in Portage County, a decrease in Native American population

reflects the largest effect of increase in sales revenues.

 82

Table 8: County wise (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

County wise
Portage County
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 1 mile 14 PerBlack 0.273 .. ..
BMP 5 miles 14 PerNative 0.238 .. ..
BMP Thiessen 14 PerNative 0.468 PerPacific 0.634
BP 5 miles 5 AvMedHH 0.799 .. ..
Summit County
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BS 1 mile 19 PerWhite 0.476 .. ..
BS 2 miles 19 PerTwoPl 0.474 PerPacific 0.565
BS 5 miles 19 PerOthers 0.246 .. ..
BS Thiessen 19 PerWhite 0.528 .. ..
Portage and Summit County
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BPS 1 mile 24 PerBlack 0.428 .. ..
BPS 2 miles 24 PerWhite 0.368 PerPacific 0.487
BPS 5 miles 24 PerTwoPl 0.271 .. ..
BPS Thiessen 24 PerBlack 0.521 PerPacific 0.595
BMPS Thiessen 64 PerAsian 0.066 .. ..

It is apparent from the data in Table 8 that, a decrease in percentage of Native

American population in Portage County yields the largest effect on sales revenues of

Burger King and McDonald’s. For Summit County a decrease in percentage of White

population has yielded the largest effect on sales revenues of Burger King and

McDonald’s. For both Portage and Summit Counties, a decrease in percentage of Pacific

Islanders and Blacks reflects an increase in annual sales of Burger King Restaurants.
 83

Table 9: Restaurant wise and County wise (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Restaurant wise and County wise

Burger King
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BP 5 miles 5 AvMedHH 0.799 .. ..

Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)

BPS 1 mile 24 PerBlack 0.428 .. ..
BPS 2 miles 24 PerWhite 0.368 PerPacific 0.487
BPS 5 miles 24 PerTwoPl 0.271 .. ..
BPS Thiessen 24 PerBlack 0.521 PerPacific 0.595

Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)

BS 1 mile 19 PerWhite 0.476 .. ..
BS 2 miles 19 PerTwoPl 0.474 PerPacific 0.565
BS 5 miles 19 PerOthers 0.246 .. ..
BS Thiessen 19 PerWhite 0.528 .. ..
Burger King and McDonald's
Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 1 mile 14 PerBlack 0.273 .. ..
BMP 5 miles 14 PerNative 0.238 .. ..
BMP Thiessen 14 PerNative 0.468 PerPacific 0.634

Model Buffer N Var(1) Adj R2(1) Var(2) Adj R2(2)

BMPS Thiessen 64 PerAsian 0.066 .. ..

Data tabulated in Table 9 shows that, Burger Kings in Summit County shows an

increase in sales as a result of decrease in percentage of White population. Burger Kings

in Portage and Summit County experience an increase in annual sales as a result of

decrease in percentage of Pacific Islanders and Blacks. Native Americans yield the

largest effect on sales for Burger Kings and McDonalds’ in Portage, and Asian

population shows an increase in sales revenues of both Burger Kings and McDonalds’ for

Portage and Summit Counties for Thiessen buffers.

 84

Another set of Stepwise Regression analysis is run for the models. This regression

is run excluding the percentage of Pacific Islanders. The other ethnic groups show the

same pattern of exerting effect on sales revenues of restaurants when the regression

analysis was run including the percentage of Pacific Islanders.

Table 10: Spatial Configuration (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Stepwise Regression (minus % of Pacific Population)

Buffer 1 mile
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 14 1 mile PerBlack 0.273 .. ..
BS 19 1 mile PerWhite 0.476 .. ..
BPS 24 1 mile PerBlack 0.428 .. ..
Buffer 2 miles
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BS 19 2 miles PerTwoPl 0.474 .. ..
BPS 24 2 miles PerWhite 0.368 .. ..
Buffer 5 miles
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 14 5 miles PerNative 0.238 .. ..
BP 5 5 miles AvMedHHI 0.799 .. ..
BS 19 5 miles PerOthers 0.246 .. ..
BPS 24 5 miles PerTwoPl 0.271 .. ..
Thiessen
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 14 Thiessen PerNative 0.468 .. ..
BS 19 Thiessen PerWhite 0.528 .. ..
BPS 24 Thiessen PerBlack 0.521 .. ..

Table 10 shows that a variation in Black population yields the largest changes in

sales revenues of Burger Kings and McDonalds’ in Portage County and Burger Kings in

Portage and Summit County for Buffer-1 mile. For Buffer-5 miles and Thiessen, no

single ethnic group shows a consistency in terms of yielding changes in sales in sales
 85

revenues of the Burger Kings Restaurants. The White population exerts changes in sales

revenues for Burger Kings Restaurants in Summit County.

Table 11: Restaurant wise (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Stepwise Regression (minus % of Pacific Population)

Burger King
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BS 19 1 mile PerWhite 0.476 .. ..
BPS 24 1 mile PerBlack 0.428 .. ..
BS 19 2 miles PerTwoPl 0.474 .. ..
BPS 24 2 miles PerWhite 0.368 .. ..
BP 5 5 miles AvMedHHI 0.799 .. ..
BS 19 5 miles PerOthers 0.246 .. ..
BPS 24 5 miles PerTwoPl 0.271 .. ..
BS 19 Thiessen PerWhite 0.528 .. ..
BPS 24 Thiessen PerBlack 0.521 .. ..
Burger King and McDonald's
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 14 1 mile PerBlack 0.273 .. ..
BMP 14 5 miles PerNative 0.238 .. ..
BMP 14 Thiessen PerNative 0.468 .. ..

Data in Table 11 shows that sales revenues of Burger King Restaurants in Summit

County are affected by the changes in percentage of White population in the Buffer

zones. Burger King Restaurants combined for Portage and Summit Counties experience

effects in their annual sales with a change in Black population.

 86

Table 12: County wise (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Stepwise Regression (minus % of Pacific Population)

Portage County
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 14 1 mile PerBlack 0.273 .. ..
BMP 14 5 miles PerNative 0.238 .. ..
BP 5 5 miles AvMedHHI 0.799 .. ..
BMP 14 Thiessen PerNative 0.468 .. ..
Summit County
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BS 19 1 mile PerWhite 0.476 .. ..
BS 19 2 miles PerTwoPl 0.474 .. ..
BS 19 5 miles PerOthers 0.246 .. ..
BS 19 Thiessen PerWhite 0.528 .. ..
Portage County and Summit County
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BPS 24 1 mile PerBlack 0.428 .. ..
BPS 24 2 miles PerWhite 0.368 .. ..
BPS 24 5 miles PerTwoPl 0.271 .. ..
BPS 24 Thiessen PerBlack 0.521 .. ..

In Portage County, a change in percentage of Native Americans exerts effects

on sales revenues of Burger King and McDonald’s Restaurants for Buffer-5 miles and

Thiessen. A decrease in Black population shows an increase in annual sales for within 1-

mile buffer of the restaurants in table 12.

In Summit County, it’s the changes in percentage of white population that

shows effects on sales revenues of Burger King restaurants.

For Portage and Summit Counties together, it is the percentage of Black

population that causes an increase in sales revenues of Burger King within 1-mile from

the restaurants and White population within 2-miles. For a much larger service area of 5-
 87

miles, it’s the percentage of population comprised of two or more races that have an

effect on sales revenues of Burger King restaurants.

Table 13: Restaurant wise and County wise (Stepwise Regression)

(S-Summit, P-Portage, M-McDonald’s, B-Burger King)

Stepwise Regression (minus % of Pacific Population)

Burger King
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BP 5 5 miles AvMedHHI 0.799 .. ..

Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)

BS 19 1 mile PerWhite 0.476 .. ..
BS 19 2 miles PerTwoPl 0.474 .. ..
BS 19 5 miles PerOthers 0.246 .. ..
BS 19 Thiessen PerWhite 0.528 .. ..

Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)

BPS 24 1 mile PerBlack 0.428 .. ..
BPS 24 2 miles PerWhite 0.368 .. ..
BPS 24 5 miles PerTwoPl 0.271 .. ..
BPS 24 Thiessen PerBlack 0.521 .. ..
Burger King and McDonald's
Model N Buffer Var(1) Adj R2(1) Var(2) Adj R2(2)
BMP 14 Thiessen PerNative 0.468 .. ..
BMP 14 1 mile PerBlack 0.273 .. ..
BMP 14 5 miles PerNative 0.238 .. ..

Burger King restaurants in Summit County mainly experience an increase in their

sales as a result of change in percentage of White population, followed by Two Plus

Races and Other Races (including Hispanics). Burger King restaurants in both Portage

and Summit Counties considered together (Table 13), show a variation in sales revenues

as a result of change in percentage of mainly Black population, followed by the Whites.

 88

Sales revenues for both Burger King and McDonald’s in Portage and Summit Counties

are affected by decrease in population of Native Americans, followed by the Blacks.

5.3 Summary:

Regression analysis is a statistical tool that is used to investigate relationship between

variables. In this study, the fast-food annual sales revenue has been taken as the

dependent variable and the demographic data and median household income as the

independent variables. The result of a regression analysis includes the coefficient of

2
multiple determinations, R , and considers whether the overall regression equation was

statistically significant via a F statistic. The R 2 indicates the extent to which the

variations as shown by values in the dependent variable being explained by the variations

among the values of the independent variables.

There are two methods for entering variables into the regression equation that are

used in this study. First the Enter method is used, which is the option of using forced

entry of the independent variables into the model. The SPSS software enters at one time

all specified variables regardless of significance levels. The second method is the

stepwise method that enters or removes one variable at a time based on a preset

significance value. The process comes to an end when there are no additional variables

for addition or removal. In this method the variance explained by certain variables will

change when new variables enter the equation.

The analysis models show that the White population is the largest clientele for the

QS Restaurants in Portage and Summit Counties. The other ethnic minorities seem to
 89

have other food preferences rather than eating at the fast food restaurants. These

restaurants show an increase in Annual Sales figures as a result of decrease in percentage

of Black population in Portage County for Thiessen and buffer polygons of varying

distance around the restaurants. This also indicates an economically weaker condition of

the Black population, which makes them eat out less and a preference to cook at home.

Ethnic population and median household income better explain annual sales variations

for Burger King within buffers of 1 and 2 miles around the restaurants.

For McDonald’s, ethnic population and median household income for 5-mile

buffers better explain the annual sales variations. This suggests that the consumers are

more willing to travel a longer distance to eat at McDonald’s as compared to Burger

King. McDonald’s Restaurants are more in number and are located at highway exits and

accessible routes. A comparative analysis of both types of proximity analysis methods

show that the buffer polygons of 1, 2 and 5 miles are able to better explain the variations

in annual revenues of QSR as a result of changes in ethnic population and median

household income as compared to the Thiessen polygons.

 Chapter 6

Conclusion

Geographic Information System is rapidly becoming a popular tool in retail industry for

the site location analysis. The main advantage of using GIS in solving location problem is

that the spatial as well as attribute data can be integrated and applied to a variety of fields.

There are many companies in the United States that have invested in GIS for site location

analysis and achieved considerable retail benefits and cost savings. For the retailers that

make use of retail location methods, the most frequently applied techniques are analog

methods, statistical modeling and gravity/spatial interaction models.

Retail location analysis has also been used in determining the size and shape of

the retail trade area. The trade area of a retail store is the geographical area from which it

draws most of its clientele and has the highest market penetration. GIS helps in not only

processing the data and retail information, but also provides the presentation of processed

data. A retail location analysis of relationship between the annual sales performance of

McDonald’s and Burger King QRS and various spatial and socio-economic factors of

their trade areas has been done. The results of this study show how location factors

influence the performance of the stores as well as how the socio-economic attributes of

the trade areas affect the annual sales revenues. This retail location analysis has been

done by partitioning the study area into a set of Thiessen polygons and buffers of

different spatial configurations surrounding the QSR outlets.

90
 91

The distance customers are willing to travel to make retail purchases depends

upon the type of goods or services to be purchased. The type of fast food offered by

Burger King and McDonald’s are inexpensive and easily substituted, hence creating a

smaller retail trade zones as compared to other specialty restaurants. Concentric buffers

of varying distances generated around these restaurants show distinct market penetration

patterns. The innermost buffer of 1-mile width accounts for the largest proportion of

customers who can walk or drive to the restaurants. Ethnic composition of the population

and the median household income within these buffer polygons also give an idea

regarding how much time and distance consumers are willing to travel to patronize these

restaurants.

Retail businesses are constantly working towards growing and expanding their

sales revenues and market shares. Market potential of the retail stores is often determined

by the expenditure pattern of the population in the respective trade zones. The market

potential is however not static. It may change due to variations in demographics, ethnic

composition and socio-economic indicators over time. A retail chain’s store location

strategy can be developed not only by keeping in view the future retail setting of the

market, but also getting fully acquainted with the regional shopping areas and consumer

orientation.

A small number of very large retail organizations and a large number of small-

scale outlets characterize the retail market. This retail environment makes it necessary

that the QRS outlets do not ignore their competitors when selecting sites for the store

outlets. Their retail marketing strategy should be greatly dependent upon the store’s value
 92

platform in order to distinguish itself from its competitors. In addition to establishing a

value platform, QRS retailers also follow functional strategies like customer service,

merchandising, advertising, store accessibility and parking, etc. The retailers can select

the segment of its consumers and can orient its services to meet the requirements of that

group. The competitive edge of the fast-food restaurants can be maintained by prices of

goods offered and the location of the outlet. Consumers are likely to patronize a

conveniently located store when choosing between similar retail alternatives.

Besides using GIS tools for retail location analysis, statistical and mathematical

models of analysis have also been used widely by researchers to forecast a retail store’s

future sales and site selection. GIS tools can help in calculating the size and potential of

the market based on the socio-economic profile of the trade area. With GIS software, it is

possible to handle large volumes of data and generate results with higher levels of

precision and details in a short time. In this study, GIS and statistical analysis techniques

have been applied to examine the spatial patterns and dynamics of the QSR chains of

Burger King and McDonald’s in Portage and Summit Counties in Northeastern Ohio.

The research methodology includes the GIS techniques and statistical analysis of

Geocoding, geoprocessing and regression analysis. Each of the studied Burger King and

McDonald’s restaurants has been converted to specific point location on the street

network using the technique of Geocoding. A catchment area analysis has been done for

these restaurants by using Thiessen polygons, buffer polygons and overlay analysis.

Thiessen polygons have been constructed around the geocoded restaurants to represent
 93

areas of influence. Thiessen polygons have been used to evaluate the restaurants in terms

of socio-economic and demographic attributes of each outlet’s trade area.

Geoprocessing has been made use of in this study to apply geographic analysis

and modeling to data to generate new information. Geoprocessing tools of Proximity

analysis (Buffer tool) and Overlay analysis (Union tool) have been used for this study.

Buffer analysis tool has been used to create a new feature class of buffer polygons of

varying widths around the geocoded QSR. These polygons allow us to analyse how they

are related to the performance of the fast food restaurants. The Buffer polygons of width

1-mile, 2-miles and 5-miles help to estimate customer willingness to travel to a store.

Overlay analysis tool has been used to apply weights to several inputs and

combining them into a single output in order to create an integrated analysis. Thiessen

polygons and Buffer polygons of varying widths have been overlaid on the data layer of

Census Block groups associated with ethnic information and median household income.

It has been assumed that both Thiessen polygons and Buffer polygons reasonably

represent trade areas in this study.

Statistical techniques of Regression analysis has been used to quantify the level of

change in the outcome of dependent variable based upon given level of change in one or

more independent variables. Multi-variate Regression analysis regression analysis models

the degree to which the variation within values of the dependent variable (Annual Sales

of studied QSR) as explained by predictors/independent variables (percentage of ethnic

populations and median household incomes). The regression analysis gives the degree to

which variation among values in the dependent variable is explained by the combined
 94

variation of the predictor variables. Regression models are tested with a hypothesis of no

statistical significance, which means that the variation among the values of the dependent

variables do not have a statistically significant relationship with the variation among

values in the predictor variables. There are two methods of entering variables into

Regression analysis used in this study. The Enter method is the forced entry option and

the Stepwise method enters or removes one variable at a time based on a preset

significance value.

2
A comparative analysis of the coefficient of determination ( R ) of the Thiessen

polygons and Buffer polygons with varying buffering limits of a total of 68 fast food

restaurants show how well or poorly either of these two approaches capture the variations

2
in the annual sales figures. A comparative analysis of R of three Buffer limits around

these restaurants determine a near ideal buffering limit for each of the fast food retail

stores.

The analysis of regression models show that Burger King’s annual sales revenues

are better explained by the ethnic population and median house-hold incomes for 1-mile

& 2-miles buffers than those of McDonald’s by the same set of variables. Annual sales

are better explained by the variables for McDonald’s for a 5-miles buffer in Portage and

Summit Counties. This suggests that consumers are more willing to travel a longer

distance to eat at/from McDonald’s as compared to Burger King. The ethnic population

and median household income better explain annual sales figures for McDonald’s than

those for Burger King, within the Thiessen polygons generated around each restaurant.

Further analysis of ethnic data suggests that white population is the largest clientele for
 95

both Burger King & McDonald’s and other minorities (except the Black population) are

too few in numbers to reflect any significant effect on annual sales figures. This could

also suggest that the ethnic minorities have other preferences and eat at other specialty

restaurants or maybe even cook more at home.

McDonald’s and Burger King are generally assumed to be maintaining a

consistency in the quality of their food served at other outlets nation-wide. The socio-

economic and demographic attributes found in the study area are consistent with those of

Ohio state as well as national averages. The study area also represents a good mix of

urban, rural and suburban lifestyles. A comparative analysis of socio-economic and

demographic factors can help us to understand the performance of markets with different

location factors. The results from this study can be applied to other fast-food chains as

well as other types of businesses.

This study can be further developed to improve upon site selection location

research tools for successful site selection of retail outlets. In order to make informed site

location and retail location decisions, a more efficient approach of on-site assessment and

evaluation of customers and markets can be undertaken. This would help in a more

accurate analysis of the market, sites and trade areas and also help in uncovering

potential, in-fill opportunities and eliminating overlaps in existing markets. A more

detailed business data, market surveys, demographic data, and household consumer data

can be used to uncover new marketing opportunities and identify profitable consumers.

By analyzing and combining the data layers formed in this study and benchmarking the
 96

current sales performance of the stores, we can also find an optimal store location and

manage the retail network more efficiently

 Appendix

(Maps and Figures)

 98

Fig: 6 Burger King and McDonald’s in Portage County (Thiessen Polygons)

 99

Fig: 7 Burger King and McDonald’s in Summit County (Thiessen Polygon)

 100

Fig: 8 Burger Kings in Portage County (Thiessen Polygons)

 101

Fig: 9 Burger Kings in Summit County (Thiessen Polygons)

 102

Fig: 10 Burger King in Portage and Summit Counties (Thiessen Polygons)

 103

Fig 11: McDonald’s in Portage County (Thiessen Polygons)

 104

Fig 12: McDonald’s in Summit County (Thiessen Polygons)

 105

Fig 13: McDonald’s in Portage and Summit Counties (Thiessen Polygons)

 106

Fig 17: Burger King & McDonald’s in Portage County (Buffer 1 mile)
 107

Fig 18: Burger King & McDonald’s in Portage County (Buffer 2 miles)
 108

Fig 19: Burger King & McDonald’s in Portage County (Buffer 5 miles)
 109

Fig 20: Burger King & McDonald’s in Summit County (Buffer 1 mile)
 110

Fig 21: Burger King & McDonald’s in Summit County (Buffer 2 miles)
 111

Fig 22: Burger King & McDonald’s in Summit County (Buffer 5 miles)
 112

Fig 23: Burger King in Portage County (Buffer 1 mile)

 113

Fig 24: Burger Kings in Portage County (Buffer 2 miles)

 114

Fig 25: Burger Kings in Portage County (Buffer 5 miles)

 115

Fig 26: Burger Kings in Summit County (Buffer 1 mile)

 116

Fig 27: Burger Kings in Summit County (Buffer 2 miles)

 117

Fig 28: Burger Kings in Summit County (Buffer 5 miles)

 118

Fig 29: Burger Kings in Portage & Summit Counties (Buffer 1 mile)
 119

Fig 30: Burger Kings in Portage & Summit Counties (Buffer 2 miles)
 120

Fig 31: Burger Kings in Portage & Summit Counties (Buffer 5 miles)
 121

Fig 32: McDonalds’ in Portage County (Buffer 1 mile)

 122

Fig 33: McDonalds’ in Portage County (Buffer 2 miles)

 123

Fig 34: McDonalds’ in Portage County (Buffer 5 miles)

 124

Fig 35: McDonalds’ in Summit County (Buffer 1 mile)

 125

Fig 36: McDonalds’ in Summit County (Buffer 2 miles)

 126

Fig 37: McDonalds’ in Summit County (Buffer 5 miles)

 127

Fig 38: McDonalds’ in Portage & Summit Counties (Buffer 1 mile)

 128

Fig 39: McDonalds’ in Portage & Summit Counties (Buffer 2 miles)

 129

Fig 40: McDonalds’ in Portage & Summit Counties (Buffer 5 miles)

 References

“An Introduction to Marketing Plans”, Malcolm McDonald, Elsevier Butterworth

Heinemann DVD, 190 min, Cambridge Marketing Colleges.

Anderson, Carol, H., (1993) Retailing- Concepts, Strategy and Information, West
Publishing Company, Minneapolis, USA.

Applebaum, William, (1968) The Analog Method for Estimating Potential Store Sales in
Guide to Store Location Research, (Ed.) Curt Kornblau, Addison-Wesley
Publishing Company, Massachusetts, USA.

ArcGIS Desktop Help 9.2, ESRI, California, USA.

http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=An_overview_
of_the_Overlay_toolset.

Baker, Robert G.V.,(2002) “On modeling internet transactions as a time-dependent

random walk: An application of the Retail Aggregate Space-Time Trip (RASTT)
Model”, in Barry Boots, et al (Eds.), Modelling Geographical Systems- Statistical
and Computational Applications (Kluwer Academic Publications, the
Netherlands).

Beaumont, J.R., (1991) An Introduction to Market Analysis, CATMOG, 53,

GeoAbstracts, Norwich.

Brabyn, Lars and Skelly, Chris, (2002) Modeling Population access to New Zealand
Public Hospitals, International Journal of Health Geographics, 1:3.

Brabyn, Lars, Population access to Hospital emergency departments and their impacts of
health reform in New Zealand, Health Informatics Journal, vol 12, No.3, p 227-
37, 2006.

130
 131

Bateman, I.J., Brainard J.S., Lovett A.A. and Garrod G.D., (1999) “The impact of
measurement assumptions upon individual travel cost estimates of consumer
surplus: a GIS analysis”, in Regional Environmental Change
1(1), November, pp- 24-30.

Benoit, D. and Clarke G.P., (1997) “Assessing GIS for retail location planning”, in
Journal of Retailing and Consumer Services, Vol. 4, No. 4, , pp- 239-258.

Birkin, Mark, “Retail location modeling in GIS”, in Longley Paul and Batty Michael
(eds.) Spatial Analysis: Modeling in a GIS Environment pp-207-226.

Birkin, Mark, Clarke G. and Clarke M., (2002) Retail Geography & Intelligent Network
Planning (John Wiley & Sons, Ltd, England).

Birkin, M., Clarke G., Clarke M. And Wilson Alan., (1996) Intelligent GIS: Location
Decisions and Strategic Planning, (GeoInformation International, Bell and Bain,
Glasgow, UK).

Birks, David F., Nasirin Syed and Zailani S.H. M., (2003), “Factors influencing GIS
project implementation failure in the UK retail industry”, in International Journal
of Information Management, 23 pp-73-82.

Boots, Bary and South, Robert, (1997) Modeling Retail Trade Areas Using Higher-Order,
Multiplicatively Weighted Voronoi Diagrams, Journal of Retailing, Vol. 73(4), p
519-536.

Boots, B.N., (1980) “Weighted Thiessen Polygons”, Economic Geography, 56, p 248-
259.

Bremner, Brian and Gail DeGeorge, (1989) “McDonald’s Stoops to Conquer”, Business
Week, October 30, p 120-124.

Brown, Stephen, (1992) Retail Location: A Micro-Scale Perspective, Ashgate Publishing

Ltd., Avebury, UK.
 132

Burt, S. and Dawson, J. A., (1990) “From small shop to hypermarket: the dynamics of
retailing”, in Pinder, D. (Ed.), Western Europe: Challenge and Change, Belhaven,
London, p 142-161.

Byrom, J.W., (2005) “The use of data in outlet locational planning: A preliminary
examination across retail and service sectors”, in Management Research News,
Vol. 28, No. 5, pp-63-74.

Church, Richard L., (2002) “Geographical information systems and location science”, in
Computers & Operations Research, 29pp-541-62.

Church, Richard and Sorensen Paul, “Integrating Normative location models into GIS:
problems and prospects with the p-median model”, in Longley Paul and Batty
Michael (eds.) Spatial Analysis: Modeling in a GIS Environment pp-167-184.

Clarke Graham, (1998), “Changing methods of location planning for retail companies”,
in GeoJournal 45 p 289-298.

Clawson, J., (1974) “Fitting branch locations, performance standards & marketing
strategies to local conditions”, Journal of Marketing, 38, p 8-14.

Converse, P.D., (1943) “A Study of Retail Trade Areas in East Central Illinois”, Urbana,
Business Studies, No. 2, University of Illinois.

Converse, P.D., (1949) “New Laws of Retail Gravitation, Journal of Marketing, Vol. 14,
October, p 379-384.

Cottrell, J., (1973) “An environmental model of performance measurement in a chain of

supermarkets”, Journal of Retailing, 49(43), p 51-63.

Crosier, Scott, (2004) ArcGIS 9: Geocoding in ArcGIS, ESRI, California, USA.

Darien, Jean and Judy Cohen, (1995) Segmenting by consumer time shortage, Journal of
Consumer Marketing, Vol. 12, No.1, p 32-44.
 133

Dartiguenave, Christine and Maidment, David R., Computing the Mean Areal
Precipitation, CE 394K Surface Water Hydrology, University of Texas at Austin.
(http://www.ce.utexas.edu/prof/maidment/ce394k/rainfall/rainfall.htm#procedure)

Davies, R. L. and D. S. Rogers, (1984) Store Location and Store Assessment Research,
John Wiley & Sons Ltd., New York, USA.

Davies, Martin and Clarke Ian, (1994) “A framework for network planning”, in
International Journal of Retail and Distribution Management, Vol.22, No.6, pp-
6-10.

Dawson, J. A., (1991) “Market Services in the United Kingdom”, in Johnston, R. J. and
Gardiner, V. (Eds.) The Changing Geography of the United Kingdom, Routledge,
London.

Delone, B.N., (1961) Proof of the fundamental theorem in the theory of stereohedra,
Soviet Mathematics, 2, p 812-815.

Densham, Paul, “Visual interactive locational analysis”, in Longley Paul and Batty
Michael (eds.) Spatial Analysis: Modeling in a GIS Environment pp-185-206.

Digital Geography Research Corporation (http://dgrc.ca/services/retail/index.html).

Dirichlet, G.L., (1850) Über die Reduction der positeven quadratischen Formen mit drei
unbestimmten ganzen Zahler, Journal für die reine und Angewandte Mathematik,
40, p 209-277.

Dramowicz, Ela, Retail Trade Area Analysis Using the Huff Model, Jul 02, 2005,
Directions Magazine.
(http://www.directionsmag.com/article.php?article_id=896&trv=1).

Elliot, C., (1991) Store planning using GIS, in Geographic Information 1991, Yearbook
of the Association of Geographic Information, ed J Cadoux-Hudson and D. I.
Heywood, Taylor and Francis, London, p 169-172.
 134

Erickson, G. F. and Finkler, S. A., (1985) “Determinants of market share for a hospital’s
services”, Medical Care, 23, p 1003-1014.

Fairbairn, F., (1984) The Urban pattern of retailing within North America, Store Location
and Store Asessment Research, (Ed. Davies, R.L. and D. S. Rogers), John Wiley
and Sons, Ltd., UK.

Fiedler, Fritz, (2003), Simple, Practical Method for Determining Station Weights Using
Thiessen Polygons and Isohyetal Maps, Journal of Hydrological Engineering,
July-August.

Fik, Timothy, (1988) “Spatial competition and price reporting in retail food markets”, in
Economic Geography, Vol. 64, January, pp-29-44.

Garreau, Joel, (1981) The Nine Nations of North America, Avon Books, New York,
USA.

Fine, Isadore, V., (1954) Retail Trade Area Analysis, Wisconsin Commerce Papers, Vol.
I, No. 6 (January), University of Wisconsin, Madison, USA.

Gething, Peter, Abdisalan M. Noor, Dejan Zurovac, Peter M. Atkinson, Simon I. Hay,
Mark S. Nixon and Robert W. Snow (2004) Empirical Modelling of government
health service use by children with fevers in Kenya, Acta Tropica 91, p 227-237.

Gibson, Richard, (1990) “Discounted Menu is Coming to McDonald’s as Chain tries to

Win Back Customers”, The Wall Street Journal, November 30, p B1-B6.

Gombosi, Matej and Borut Zalik, (2005) Point-in-polygon tests for geometric buffers, in
Computers & Geosciences 31, p 1201-1212.

Green, Howard, (1986) “The Art and Science of Retail Site Selection,” National
Association of Corporate Real Estate Executives Annual Symposium (April 14),
San Antonio, Texas, USA.
 135

Hargrove, William, Winterfield, Richard F. and Levine, Danile, (1995), Dynamic

Segmentation and Thiessen Polygons: A solution to the River Mile Problems,
Proceedings ESRI Arc/Info Users Conference, Palm Spring, CA.
(http://research.esd.ornl.gov/CRERP/DOCS/RIVERMI/P114.HTM).

Hernandez, Tony, Bennison David and Cornelius Sarah, (1998) “The organizational
context of retail locational planning”, in GeoJournal 45, p-299-308.

Hess, Ronald L., Rubin Ronald S. and Lawrence A. West Jr., (2004) “Geographic
information systems as a marketing information system technology”, in Decision
Support Systems, 38pp-197-212.

Hise, R. T., Kelly, J. P., Gable, M. and McDonald, J. B., (1983) “Factors affecting the
performance of individual chain store units: an empirical analysis”, Journal of
Retailing, 59, p 1-18.

Horton, R.E., (1917) Rational study of rainfall data makes possible better estimates of
water yield, Engineering News Record, 79, p 211-213.

Howe, A., (1991) Assessing potential of bank outlets using GIS, in Geographic
Information 1991, Yearbook of the Association of Geographic Information, ed J
Cadoux-Hudson and D.I. Heywood, Taylor and Francis, London, p 173-175.

Huff, D.C., (1963) A Probabilistic Analysis of Consumer Spatial Behaviour, Real Estate
Research Program Reprint No. 18, Graduate School of Business Administration,
University of California, Los Angeles, USA.

Huff, David, L. and Ronald T. Rust, (1984) “Measuring the Congruence of Market
Areas”, in Journal of Marketing, 48 (Winter), p 68-74.

JC Penny Inc., (1987) Annual Report.

Jensen, Prof. Arthur, N., (2006) College of Business Administration, California State
University, Sacramento, CA, August 25.
 136

Jeyanandhini V, Uma Mu, Health Impact Assessment on Beach Water Sampling- A Case
Study, Anna University, Chennai, India.

Jones, K. G. and Mock, D. R., (1984) “Evaluating retail trading performance”, (Ed.
Davies, R. L. and Rogers, D. S., Store Location and Store Assessment Research,
John Wiley & Sons, New York, USA).

Jones, K and Simmons, J., (1993) Location, Location, Location: Analyzing the Retail
Environment, 2nd Edition, Scarborough, Nelson, Ontario, Canada.

Joseph, Lawrence, G., (2005) Unpublished M.A. Thesis, Kent State University,
Department of Geography, Kent, Ohio.

Journel, A.G., (1992) A letter to the Editor of the Journal for Mathematical Geology,
October 15, 1992.

Kohsaka, Hiroyuki, (1997) “Monitoring and analysis of a retail trading area by a card
information/GIS approach”, in Journal of Retailing and Consumer Services,
Vol.4, No. 2, pp-109-115.

Kopec, R.J., (1963) “An Alternative Method For The Construction Of Thiessen
Polygons”, in The Professional Geographer, 15, p 24-26.

Ladak, Alnoor and Martinez, Roberto B. (1996), Automated Derivation of High

Accuracy Road Centerlines Thiessen Polygon Technique, in ESRI User
Conference Proceedings, London, England.
(http://gis.esri.com/library/userconf/proc96/TO400/PAP370/P370.HTM).

Lakshmanan, T.R. and Hansen, W.G., (1965) A Retail Market Potential Model, Journal
of the American Institute of Planners, Vol. 31, p 134-143.

Lao, Yong, (1993) “ Solving large-scale location spatial interaction models for retail
analysis: A GIS supported heuristic approach”, PhD dissertation, Ohio State
University.
 137

Longley, P., and Clarke G., (eds.), (1995) GIS for Business and Service Planning,
GeoInformation International (Cambridge).

Lord, Dennis, J. and Charles D. Lynds, (1981) “The Use of Regression Analysis in Store
Location Research: A Review and Case Study,” Akron Business and Economic
Review (Summer), p 13-19.

Lynge, M. and Shin, T., (1981) “Factors affecting rural bank market share”, Akron
Business and Economic Review, 10, p 35-39.

Mahoney, M.S., (1979) Rene Descartes. Le Monde, ou Traite de la lumiere (Translation

and Introduction) Abaris Books, New York, USA.

Malczewski, Jacek, (2004), “GIS- based land use suitability analysis: a critical
overview”, in Progress in Planning, 62, pp-3-65.

Martin, P., (1967) “Savings and loans in new submarkets: Search for predictive factors”,
Journal of Marketing Research, 4, p 163-166.

Matthews, Stephen, A., (2002) Geographically Weighted Regression, GIS Resource

Document 02-07(GIS-RD-02-07). (http://www.pop.psu.edu/gia-
core/pdfs/gis_rd_02-07.pdf).

Mercurio, J., (1984) Store Location Strategies, Store Location and Store Assessment
Research, (Ed. Davies, R. L. and D. S. Rogers, (1984) Store Location and Store
Assessment Research, John Willey & Sons, Ltd., New York.).

Nakaya, Tomoki, (2002) “Local Spatial Interaction Modeling based on the

Geographically Weighted Regression Approach”, in Barry Boots, Atsuyuki and
Richard Thomas (eds.), Modeling Geographical Systems- statistical and
Computational Applications (Kluwer Academic Publishers, the Netherland).

Nasirin, Syed and David F. Birks, (2003) “DSS implementation in the UK retail
organizations: a GIS perspective”, in Information and Management 40 p 325-336.
 138

Nowacki, W., (1933) Der Begriff ‘ Voronoischer Bereich’, Zeitschrift für

Kristallographie, 85, p 331-332.

Nowacki, W., (1976) Über allgemeine Eigenschaften von wirkungsbereichen für

Kristallographie, 143, p 360-385.

Okabe, A. and A. Suzuki, (1997) “Locational Optimization Problems Solved Through

Voronoi Diagrams”, in European Journal of Operational Research, 98, p 445-
456.

Okabe, Atsuyuki, Boots, Barry, Sugihara, Kokichi and Chiu, SungNok, (2000) Spatial
Tessellations: Concepts and Applications of Voronoi Diagrams, (2nd Edition),
John Wiley & Sons, Ltd, West Sussex, England.

Okabe, Atsuyuki and Okunuki Kei-ichi, (2001) “A computational method for estimating
the demand of retail stores on a street network and its implementation in GIS”, in
Transactions in GIS, 5(3), p- 209-220.

Olsen, L. M. and Lord, J. D., (1979) “Market area characteristics and branch bank
performance”, Journal of Bank Research (summer) 10, p 102-110.

O’Malley, Lisa, Patternson Maurice and Evans Martin, (1997) “Retailer use of
geodemographic and other data sources: an empirical investigation”, in
International Journal of Retail and Distribution Management, Vol.2, No.6, , pp-
188-96.

Pettit, C., and Pullar, D., (1999) “An integrated planning tool based upon multiple criteria
evaluation of spatial information”, in Computers, Environment and Urban
Systems, 23pp-339-57.

Price, Maribeth, (2004) Mastering ArcGIS, McGraw Hill, New York, USA.

QSR Magazine, “The Revolution”, November 2002, p 44.

 139

Reid, H.G., (1993) Retail Trade, in Profiting from a Geographic Information System, ed
G. Castle, GIS World Inc, Fort Collins, CO, p 131-152.

Reilly, William, J., (1929) Methods for the Study of Retail Relationships, Bureau of
Business Research, Monograph No. 4, University of Texas, Austin, Texas, USA.

Reilly, William, J., (1931) The Law of Retail Gravitation, W.J. Reilly Co., New York,
USA.

Rob, Mohammad A., (2003) “Some challenges of integrating spatial and non-spatial
datasets using a geographical information system”, in Information Technology for
Development, 10 pp- 171-178.

Rogers, David and Green, Howard, (1979) “A New Perspective on Forecasting Store
Sales: Applying Statistical Models and Techniques in the Analog Approach”,
Geographical Review, Vol. 69, No. 4, October, p 449-458.

Sadahiro, Yukio, (2001) “ A PDF-based analysis of the spatial structure of retailing”, in

GeoJournal 52, p 237-252.

Sadahiro, Yukio, (2001) Number of polygons generated by map overlay: The case of
convex polygons, in Transactions in GIS, 5(4), p 345-353.

Shamos, M.I. and D. Hoey, (1975) Closest-point problems, Proceedings of the 16th
Annual IEEE Symposium on Foundations of Computer Science, p 151-162.

Shaw, S.L., (1991) Urban Transit Accessibility Analysis Using A GIS: A Case Study of
Florida’s Tri-Rail System, in Southeastern Geographer, 31, p 15-30.

Sierra, R., Rodriguez, F. and Losos, E., (1999) Forest resource use change during early
market integration in tropical rainforests: the Huaorani of Upper Amazonia,
Ecological Economics, 30, p 107-119.
 140

Skrepnek, Grant, (2005) Regression Methods in the Empiric Analysis of Health Care
Data, in Journal of Managed Care Pharmacy, April, Vol. 11, No. 3, p 240-251.

Taylor P.J., and Carmichael C.L.,(1980). Health and the application of geographical
methodology, Community Dentistry and Oral Epidemiology, Vol. 8, Issue 3, p
117-122.

Thiessen, A.H., (1911) Precipitation averages for large areas, Monthly Weather Review,
39, p 1082-1084.

“The Merchants of Cool”, (2004) PBS Video, 60 min, FROL6909, WGBH Educational
Foundation.

Thompson, John, S., (1982) Site Selection, Chain Store Publishing Corp., New York,
USA.

US Census Bureau, Census 2000 Summary File 3Technical Documentation, 2002.

USA Today, “Fast Food World Says Drive-thru is the Way to Go”, April 3,2002, p 2A.

Vlachopoulou, Maro, Silleos George and Manthou Vassiliki, “Geographic information

systems in warehouse site selection decisions”, in International Journal of
Production Economics, 71 (2001) pp-205-212.

Van Wee, B., Hagoort, M., and Annema, J. A., (2001) “Accessibility Measures with
Competition”, Journal of Transport Geography, 9, p 199-208.

Voronoi, G.C., (1907) Nouvelles applications des parameters continues a la theorie des
formes quadratiques. Premier Memoire: Sur quelques proprietès des formes
quadratiques positives parfaits, Journal für die Reine und Angewandte
Mathematik, 133, p 97-178.

Wang, Fahui, (2006) Quantitative Methods and Applications in GIS, CRC Press, Taylor
and Francis Group, Florida, USA.
 141

Whitney, E.N., (1929) Areal rainfall estimates, Monthly Weather Review, 57, p 462-463.

Wikipedia information about ‘spatial dependence’ on Answers.com. Wikipedia

Copyright © 2005 by Wikipedia. Published by Wikipedia.

Wilkin, Teddy, A., Dany Garant, Andrew G. Gosler and Ben C. Sheldon, (2006)
“Density effects on life-history traits in a wild population of the great tit Parus
major: Analysis of long-term data with GIS techniques”, Journal of Animal
Ecology, 75, p 604-615.

Witlox, Frank, “MATISSE: a relational expert system for industrial site selection”, in
Expert Systems with Applications 24(2003), pp- 133-144.

Wood, R. Victor, Jr. (Senior Vice President/ Director of Real Estate and Construction,
Volume Shoe Corporation), (1986) “Restaurant/Retail Site Selection – Is it an art
or a science?” Panel Presentation (April 14), National Association of Corporate
Real Estate Executives Annual Symposium, San Antonio, Texas, USA.

(http://www.fundinguniverse.com/company-histories/McDonalds-Corporation-Company-
History.html).

(http://www.mcdonalds.ca/pdfs/history_final.pdf).

Zhen, Biao, “ A GIS-based prototype for retail trade”, Masters of Arts Thesis, University
of Cincinnati, June 1996.