EXAMINING THE BEHAVIORAL INTERACTIONS BETWEEN URBAN RESIDENTS
AND THEIR FOOD ENVIRONMENT:
A CASE STUDY OF GREATER LANSING, MICHIGAN
By
Meghan Reed

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Geography
2011

ABSTRACT
EXAMINING THE BEHAVIORAL INTERACTIONS BETWEEN URBAN RESIDENTS
AND THEIR FOOD ENVIRONMENT:
A CASE STUDY OF GREATER LANSING, MICHIGAN
By
Meghan Reed

Access to nutritious food is essential to human survival and the consumption of fresh produce is
an important part of a healthy diet. Previous research has found that residents of urban
neighborhoods who lack access to fresh produce have greater difficulty maintaining a healthy
diet. This thesis considers the interaction between residents of Greater Lansing, Michigan and
their food environment. An in-person survey of 185 Greater Lansing residents was conducted
during the summer of 2010 at twelve retail stores that sold fresh produce items. Participants
were asked questions about their access to transportation, shopping habits, diet, and primary
location of residence. Statistical analyses were conducted to examine the relationships between
the collected data and calculated models of spatial accessibility. Findings suggest that (1) the
perception of Greater Lansing residents is that food shopping is not done in close proximity to
the primary residence; (2) there is not a straightforward association between calculated access to
and the consumption of fresh produce; (3) supermarkets are the dominant food source for
residents; and (4) individuals who receive nutritional assistance are significantly more likely to
shop at neighborhood stores and consume a less healthy diet. Understanding how Greater
Lansing residents interact with their food environment is essential in finding solutions to the
growing obesity and overweight epidemic.

Copyright by
Meghan Reed
2011

ACKNOWLEDGMENTS

First and most importantly I would like to thank my advisor Dr. Kirk Goldsberry who
kept me on task and supported me throughout the completion of my thesis. His assistance with
developing my ideas has been invaluable and I have truly appreciated his guidance and insight.
Dr. Antoinette WinklerPrins and Dr. Ashton Shortridge, my committee members, have also been
incredibly helpful in assisting me in writing my thesis. Without the support of my entire
committee I would have been unable to complete this project.
I would also like to express my sincere gratitude to my fellow graduate students for their
support and guidance throughout my entire experience at MSU. Countless graduate students
were available for me to bounce ideas off of and assist me with developing this thesis including,
Steven Fuller, Courtney Gallaher, Jia Feng, Ritaumaria Pereira, Mark DeVisser, Josh Stevens,
and Carolyn Fish. I am not sure that I would have been able to complete this program without
these students and I am forever grateful for all of their assistance.
The completion of this project would have been entirely impossible without the support
of local store managers and retail establishments who allowed me to perform my survey at their
stores. Before I started this project I had a lot of doubts that I would be able to perform my
survey at retail establishments and I was pleasantly surprised at how warmly I was received and
the willingness of local stores to help increase the knowledge of the local food environment.
Finally I would like to acknowledge my parents who have instilled in me a love of
knowledge and have supported me throughout my time in East Lansing. Their love and humor
have made my time at MSU incredibly enjoyable, especially the constant reminders from my
v

mother of the family motto ―work hard, play hard‖. I certainly could not have completed this
program without the help of my family.

vi

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................................... v
LIST OF FIGURES ....................................................................................................................... vi
LIST OF EQUATIONS ................................................................................................................ vii
CHAPTER 1: INTRODUCTION ................................................................................................... 1
1.1
Background ...................................................................................................................... 1
1.2
Statement of problem ....................................................................................................... 2
1.3
Research objectives .......................................................................................................... 3
1.4
Research questions ........................................................................................................... 4
1.5
Hypotheses ....................................................................................................................... 5
CHAPTER 2: LITERATURE REVIEW ........................................................................................ 7
2.1 Obesity and overweight in the United States ........................................................................ 7
2.2 The consumption of fruits and vegetables by Americans ..................................................... 8
2.3 Nutritional assistance programs ............................................................................................ 9
2.4 Food environments .............................................................................................................. 10
2.5 Accessibility ........................................................................................................................ 12
2.5.1 The automobile and access ........................................................................................... 12
2.5.2 The food environment and access ................................................................................ 14
2.6 Research on spatial variation within food environments .................................................... 16
2.6.1 Socio-economic status .................................................................................................. 17
2.6.2 Ethnicity........................................................................................................................ 18
2.7 Research on the impact of spatial variation in food access on dietary behaviors ............... 19
2.7.1 Food environments where healthy food is not available .............................................. 19
2.7.2 Importance of supermarkets ......................................................................................... 20
2.7.3 Socioeconomic status ................................................................................................... 20
2.7.4 Mobility ........................................................................................................................ 21
2.7.5 Level of impact of spatial variation in food access on dietary behaviors..................... 21
2.7.6 Food environment does not influence dietary behavior ............................................... 22
2.8 Behavioral geography ......................................................................................................... 23
2.9 Previous food environment research in Greater Lansing, MI ............................................. 24
CHAPTER 3: METHODS ............................................................................................................ 29
3.1
Overview ........................................................................................................................ 29
3.2 Study area ............................................................................................................................ 30
3.3 Questionnaire development ................................................................................................. 31
3.3.1 Interviewer notes .......................................................................................................... 32
3.3.2 Transportation ............................................................................................................... 32
vii

3.3.3 Consumer behavior ....................................................................................................... 33
3.3.4 Diet and use of produce ................................................................................................ 33
3.3.5 Location of residence.................................................................................................... 33
3.3.6 Demographic information............................................................................................. 33
3.3.7 Pretesting and IRB approval ......................................................................................... 34
3.4 Selection of study sites ........................................................................................................ 34
3.5 Sampling method................................................................................................................. 37
3.6 Statistical analysis ............................................................................................................... 38
3.6.1 Calculation of access scores ......................................................................................... 39
3.6.2 Correlation .................................................................................................................... 40
3.6.3 Backwards stepwise regression .................................................................................... 41
3.6.4 Geographically weighted regression ............................................................................ 43
3.6.5 Two-sample t-test ......................................................................................................... 44
CHAPTER 4: RESULTS AND ANALYSIS ............................................................................... 46
4.1 Introduction ......................................................................................................................... 46
4.2 General survey results ......................................................................................................... 46
4.2.1 Travel behavior ............................................................................................................. 48
4.2.2 Shopping behavior ........................................................................................................ 50
4.1.3 Dietary behavior ........................................................................................................... 52
4.3 Correlation analysis ............................................................................................................. 53
4.4 General population backwards stepwise regression models ............................................... 54
4.4.1 Model #1: Number of different produce items (listed) purchased each week .............. 55
4.4.2 Model #2: Number of different produce items (estimated) purchased each week ....... 58
4.4.3 Model #3: Number of times meals are prepared with fresh produce items in a week . 60
4.4.4 Overall analysis of general population backwards stepwise regression models .......... 63
4.5 Geographically weighted regression ................................................................................... 64
4.5.1 Model #1: Number of different produce items (listed) purchased each week.............. 65
4.5.2 Model #2: Number of different produce items (estimated) purchased each week ....... 66
4.5.3 Model #3: Number of times meals are prepared with fresh produce items in a week . 68
4.5.4 GWR analysis ............................................................................................................... 70
4.6 Analysis of population who receive nutritional assistance ................................................. 71
4.6.1 Correlation analysis ...................................................................................................... 73
4.6.2 Difference in travel time ............................................................................................... 74
4.6.3 Difference in food retail store frequency ...................................................................... 76
4.6.4 Difference in the consumption of fresh produce .......................................................... 77
CHAPTER 5: DISCUSSION ........................................................................................................ 80
5.1 Research question #1: Association between calculated access to and the use of fruits and
vegetables in the diets of Greater Lansing residents ................................................................. 80
5.2 Research question #2: Variables that influence dietary behavior ....................................... 81
5.3 Research question #3: How Greater Lansing residents view their food environment ........ 83
5.4 Research question #4: Difference between Lansing residents who receive nutritional
assistance and those who do not ................................................................................................ 83
viii

5.5 Importance of examining the food environment as a whole ............................................... 86
5.6 Limitations of research ........................................................................................................ 87
CHAPTER 6: CONCLUSION ..................................................................................................... 91
6.1 Modeling the Lansing food environment ............................................................................ 91
6.2 Policy implications .............................................................................................................. 94
6.3 Future research .................................................................................................................... 96
APPENDICES .............................................................................................................................. 98
APPENDIX A: SURVEY INSTRUMENT .................................................................................. 99
APPENDIX B: LIST OF VARIABLES INCLUDED IN BACKWARDS STEPWISE
REGRESSION MODEL ............................................................................................................. 107
REFERENCES ........................................................................................................................... 110

ix

LIST OF TABLES

Table 4.1: Demographic results .................................................................................................... 47
Table 4.2: Travel time based on retail store type .......................................................................... 49
Table 4.3: Consumer behavior based on retail store category ...................................................... 51
Table 4.4: Participant diet perceptions.......................................................................................... 53
Table 4.5: Pearson correlation values between modeled access to fresh produce and dietary
behavior......................................................................................................................................... 54
Table 4.6: Number of different fresh produce items (listed) purchased each week: Backward
stepwise regression results ............................................................................................................ 56
Table 4.7: Number of different fresh produce items (estimated) purchased each week: Backward
stepwise regression results ............................................................................................................ 58
Table 4.8: Number of times meals are prepared with fresh produce items each week: Backward
stepwise regression results ............................................................................................................ 61
Table 4.9: Dietary perceptions of those who receive nutritional assistance in comparison to those
who do not..................................................................................................................................... 73
Table 4.10: Pearson correlation vales between calculated access and dietary behavior for those
who receive nutritional assistance and those who do not ............................................................. 74
Table 4.11: Mean and standard deviation for perceived travel time from home to retail location75
Table 4.12: Food retail store frequency two-sample t-test results ................................................ 77
Table 4.13: Consumption of fresh produce two-sample t-test results .......................................... 79

v

LIST OF FIGURES

Figure 2.1: Container method accessibility model of Lansing food environment (pedestrian)
(Goldsberry et al. 2010) ................................................................................................................ 25
Figure 2.2: Weighted method accessibility model of Lansing food environment (pedestrian)
(Goldsberry et al. 2010) ................................................................................................................ 26
Figure 2.3: Container method accessibility model of Lansing food environment (Goldsberry et al.
2010) ............................................................................................................................................. 27
Figure 3.1: Location of study sites ................................................................................................ 36
Figure 4.1: Primary residence of participants ............................................................................... 48
Figure 4.2: Number of different fresh produce items (listed) purchased each week: Model
residuals ........................................................................................................................................ 57
Figure 4.3: Number of different fresh produce items (estimated) purchased each week: Model
residuals ........................................................................................................................................ 60
Figure 4.4: Number of times meals are prepared with fresh produce items each week: Model
residuals ........................................................................................................................................ 62
Figure 4.5: Number of produce items purchased each week (listed) GWR R2 values (predictive
power of number of produce items purchased)............................................................................. 66
Figure 4.6: Number of produce items purchased each week (estimated) GWR R2 values
(predictive power of number of produce items purchased) .......................................................... 68
Figure 4.7: Number of times meals are prepared with fresh produce items in a week GWR R2
values (predictive power of number of times meals are prepared with fresh produce items in a
week) ............................................................................................................................................. 69

vi

LIST OF EQUATIONS

Equation 3.1: OLS regression ....................................................................................................... 42
Equation 3.2: Traditional regression ............................................................................................. 43
Equation 3.3: Geographically Weighted Regression .................................................................... 44

vii

CHAPTER 1: INTRODUCTION

1.1 Background
The United States is currently faced with a rapidly expanding epidemic of obesity and
1

overweight , which has been associated with a number of different health problems including
type 2 diabetes, heart disease, stroke, hypertension, and certain types of cancer (Sherry et al.
2010). In 2008 more than 72 million adults in the United States were classified as obese and
research has estimated that more than 65% of the United States population are either obese or
overweight (Hedley et al. 2004; Flegal et al. 2010; Sherry et al. 2010). As a result, the health
risks that have been associated with obesity and overweight have had a tremendous impact on
the cost of medical care in the United States. Obesity related illnesses alone accounted for more
than one-quarter of the increase in United States healthcare costs between 1987 and 2001
(Thorpe et al. 2004). Obesity and overweight related health issues and healthcare costs continue
to be a major challenge facing the United States.
An integral part of helping to decrease obesity and overweight in the United States will
be ensuring that Americans consume a healthy and balanced diet. The United States
Government established the Dietary Guidelines for Americans in 1980 to provide advice and
recommendations on achieving a healthy diet in order to reduce incidents of major chronic
diseases and ensure that Americans receive the proper nutrition they require (U.S. Department of

1

According to the Centers for Disease Control and Prevention an adult who has a BMI between
25 and 29.9 is considered overweight and adult who has a BMI of 30 or higher is considered
obese (CDC 2010).
1

Health and Human Services and U.S. Department of Agriculture 2010). Americans are
recommended to consume a diverse diet, including a variety of different fruits and vegetables, in
order to obtain their required daily nutrients. Unfortunately, research has indicated that
Americans are not consuming the amount of fruit and vegetables each day necessary to achieve a
healthy and balanced diet (Kantor 1998; Casagrande et al. 2007; Usual Dietary Intakes 2010).
The inability of many Americans to consume a healthy and balanced diet, rich in fruits and
vegetables and low in calorie-dense foods, has contributed greatly to the obesity epidemic and
the health risks that are associated with obesity (Usual Dietary Intakes 2010). Improving the
nutritional intake of Americans is an important step towards combating obesity and related
illnesses in the United States. In order to improve the dietary behavior of Americans an
understanding of why Americans are not consuming a healthy and balanced diet must be
achieved.
1.2 Statement of problem
Previous research has revealed that adequate access to healthy and nutritional food items
is a barrier to consuming a balanced diet (Clarke 2002; Campbell 2004; Pothukuchi 2004).
Unfortunately, drastic changes that have taken place in the food-retail environment in the United
States since 1970 have influenced and changed urban areas substantially. Major supermarket
chains have moved out of the urban core and have located in the periphery of urban areas
(Morland et al. 2002b). Small scale neighborhood grocery stores have been replaced by
significantly larger scale supermarkets that are meant to serve substantial geographic areas
(Lavin 2005). As a result, place and space have been recognized in recent years as being
increasingly important in influencing health and health related behaviors (Ball et al. 2006). The
term ‗food desert‘ was first used by the Nutrition Task Force of the Department of Public Health
2

in 1995 in recognition of this developing problem and to help describe urban areas that did not
have access to cheap, nutritious food items (Lavin 2005). The suburbanization of supermarkets
has disproportionately affected urban residents with low incomes and decreased mobility, who
reside in areas where cheap and nutritious food is unavailable and are disproportionately
impacted by the limited food choice available in their neighborhood (Piachaud and Webb 1996;
Whelan et al. 2002). Recent research has begun to suggest that the level of access an individual
has to supermarkets and large scale grocery stores does in fact have an impact on their dietary
behavior (Rose and Richards 2004; Henderickson et al. 2006). However, very few research
studies have focused on how consumers actually interact with their food retail environment and
have made assumptions based on large aggregated datasets (Morland et al. 2002b). As a result, it
remains unclear about how the food environment actually impacts the behavior of local residents.
1.3 Research objectives
The overall objective of this thesis is to provide understanding of the interaction that
Greater Lansing residents have with their food environment. In order to establish better models
of the food environment and its impact on dietary behavior more information is needed on the
behavioral patterns of Greater Lansing residents. This thesis will build off previous research
conducted on the food environment in Greater Lansing, Michigan (Duvall et al. 2010;
Goldsberry et al. 2010) in order to come to a greater understanding of how local residents
interact with their food environment. Specifically, this thesis seeks associations between
previous models of the Greater Lansing food environment (Goldsberry et al. 2010) and the level
of consumption of fresh fruit and vegetable items by local residents. This information is
essential in order to help understand complex interactions between urban food environments and
dietary choices of Greater Lansing residents. Additional information on how Greater Lansing
3

residents travel, shop, and behave within the food environment was also obtained in order to
further understand the interaction between Greater Lansing residents and their food environment.
Residents who receive nutritional assistance were examined in depth to understand if they
interact with the food environment differently than residents who do not receive nutritional
assistance. The results of this research will help to achieve a greater understanding of how
Greater Lansing residents interact with their food environment and help gain new insight into
how the built environment influences dietary behavior.
1.4 Research questions
The primary focus of this thesis is to examine the associations between the food
environment and the behavior of Greater Lansing residents; specifically this thesis attempts to
answer four research questions:
1) Is there an association between calculated access to fresh produce products and
the use of fruits and vegetables in the diets of Greater Lansing residents?;
2) Are there variables other than the retail food environment that influence the
dietary behavior of Greater Lansing residents?;
3) Do Greater Lansing residents believe that the food retail environment in the
Greater Lansing area is adequate?;
4) Do Greater Lansing residents who receive nutritional assistance interact with the
food environment differently than Greater Lansing residents who do not receive
nutritional assistance?
In order to answer the four primary research questions, interviews with Greater Lansing residents
were conducted to understand their behavioral decisions and how they interact with their local
food environment. The information obtained from this research will be compared to previous
4

models of the food environment that were created by Goldsberry et al. (2010) to understand how
well models of the food environment predict dietary behavior. This thesis will provide new
insight into the actual interaction between Greater Lansing residents and their food environment
that can be used to inform future representations of the Greater Lansing food environment.
1.5 Hypotheses
The expected results of this research are that: 1) calculated access to fresh produce does
impact the dietary choices of Greater Lansing residents and that previously calculated models of
access to fresh fruit and vegetable items (Goldsberry et al. 2010) will be able to predict the
quantity of fresh produce products an individual consumes (Rose and Richards 2004;
Henderickson et al. 2006); 2) variables other then the retail food environment will influence the
dietary behavior of Greater Lansing residents and the main variables that will predict the amount
of fresh produce products an individual consumes will be location of primary residence, income,
ethnicity and whether or not an individual receives nutritional assistance (Morland et al. 2002b;
Rose and Richards 2004; Lavin 2005); 3) Greater Lansing residents will not believe that their
food environment is adequate and recognize that not everyone has equal access to fresh produce
products and will feel that improvements need to be made to achieve adequate shopping near
their primary place of residence; 4) one of the main contributors to why Greater Lansing
residents will not believe their food environment is adequate is that most Greater Lansing
residents will have to travel farther than a 10 minute drive or walk to purchase food items and
will have to travel by vehicle to get there; 5) Greater Lansing residents who live in areas that
have low calculated levels of access to fresh produce products will shop more frequently at
neighborhood type convenience and liquor stores; 6) Greater Lansing residents who receive
nutritional assistance will consume less fresh produce items and be more impacted by their food
5

environment then Greater Lansing residents who do not receive nutritional assistance. The
results of this analysis are meant to inform public policy and enable planners to better understand
the actual interaction between people and their food environment. Hopefully this research will
assist in the development of built environments that will enable Americans to achieve a healthy
and balanced diet.

6

CHAPTER 2: LITERATURE REVIEW

2.1 Obesity and overweight in the United States
The United States is faced with an epidemic of obesity, which has had major
consequences for the overall health of the United States population. More than 30% of
Americans in most sex and age groups are currently classified as obese and 65% of Americans
are either obese or overweight (Hedley et al. 2004; Flegal et al. 2010). While there have been no
statistically significant changes in the United States obesity rate since 1999 there is also no
evidence that the obesity rate has been decreasing (Hedley et al. 2004; Felgal et al. 2010).
Obesity is associated with a number of different health problems including heart disease, stroke,
type II diabetes, and certain types of cancer (Sherry et al. 2010). As a result of the
extraordinarily large rate of obese and overweight Americans the average yearly healthcare costs
associated with obesity have been estimated to be $147 billion dollars, with an obese individual
being estimated to pay $1,429 more yearly on medical costs than a non-obese person (Sherry et
al. 2010). Obesity and overweight will continue to be major obstacles to the overall health of the
st

American population as the United States moves forward in the 21 century.
Research has indicated that a number of different factors contribute to obesity including
diet, physical activity, and the environment (Papas et al. 2007). Specifically, a diet that includes
a high intake of calorie-dense foods and low intake of calorie-poor items has been positively
associated with obesity (Liese et al. 2009). An integral part of helping to decrease the rates of
obesity and overweight in the United States will be ensuring that Americans are consuming a
healthy and balanced diet. In response to the expanding rate of obesity and overweight the
United States Government created the Dietary Guidelines for Americans in 1980 to provide
7

advice and recommendations on achieving a healthy diet with the purpose of helping to reduce
incidents of major chronic diseases, which were beginning to strain the country‘s healthcare
system (U.S. Department of Health and Human Services and U.S. Department of Agriculture
2010). The primary purpose of the Dietary Guidelines for Americans is to provide evidence
based nutrition information for any American over the age of two, as well as to provide
education on how to achieve a healthy and balanced diet (U.S. Department of Health and Human
Services and U.S. Department of Agriculture 2010). Unfortunately even with national dietary
guidelines Americans have been unable to meet the recommendations for achieving a healthy
diet for many years, which has caused concern for public health officials (Morland et al. 2002b).
2.2 The consumption of fruits and vegetables by Americans
The 2010 Dietary Guidelines for Americans stress the importance of consuming fruits
and vegetables in a healthy and balanced diet. Research has indicated that the consumption of
fruits and vegetables helps to protect against cardiovascular disease, coronary heart disease, and
ischemic stroke (Joshipura et al. 1999; Joshipura et al. 2001; John et al. 2002; Adebawo et al.
2006) and provide nutrients that are essential to a healthy diet (U.S. Department of Health and
Human Services and U.S. Department of Agriculture 2010). However, fruit and vegetable intake
by Americans is still well below the recommended dietary guidelines (Kantor 1998; Casagrande
et al. 2007; Usual Dietary Intakes 2010). While research has found that the overall consumption
of fruits and vegetables in the United States has increased since 1970 much of this has been a
result of the growth in potato consumption, which accounted for 17.5% of all vegetable servings
consumed by Americans in 1996 (Kantor 1998). Americans are simply not consuming the
amount and diversity of fruits and vegetables required to achieve a healthy and balanced diet.

8

Research has been divided about the current availability of fruits and vegetables in retail
stores in the United States. Putnam and Allhouse (1997) found that there has been an increase of
19% in the amount of fruits and vegetables stocked in retail stores since the 1970s. However,
Krebs-Smith et al. (2010) found that from 1970 to 2007 the amount of fresh fruits and vegetables
supplied by retailers in the United States has in fact changed very little. More importantly, the
supply of fresh fruits and vegetables in the United States continues to remain approximately half
of what it would need to be to adequately satisfy the fruit and vegetable requirements in the
Dietary Guidelines for Americans. The supply of vegetables in particular would have to increase
by 70% in order to adequately satisfy the established guidelines. Currently the available supply
of fruits and vegetables in retail stores is insufficient to enable Americans to achieve a healthy
and balanced diet (Krebs-Smith et al. 2010).
2.3 Nutritional assistance programs
An additional United States Government response has had to the nutritional problems
plaguing Americans was the creation of the Food Stamp Program (now Nutrition Assistance
Program) in 1970 in order to ensure that low-income households were able to maintain
nutritionally adequate diets (Butler and Raymond 1996). The federal government originally
created nutritional assistance programs because research indicated low socioeconomic status
households were more likely to have nutritional deficiencies than the general population (Rose
and Richards 2004). According to the United States Department of Agriculture Food and
Nutrition Service, nutritional assistance programs were established with the purpose of helping
to provide children and individuals with a low socioeconomic status access to food, a healthy
diet, and nutritional education. While Michigan and other states also offer state programs to aid
nutrition assistance, by far the largest provider of nutrition assistance in the United States is the
9

federal government (Gunderson and Oliveira 2001; Rose and Richards 2004). Even today,
individuals receiving nutritional assistance have higher food insufficiency rates than individuals
not receiving nutritional assistance (Gunderson and Oliveira 2001). Understanding why the
subpopulation who receives nutritional assistance continues to be at risk for nutritional
deficiencies is essential to helping decrease the rates of obesity and overweight in the United
States.
2.4 Food environments
While there is not one individual factor that influences food choice (Sobal and Bisogni
2009) research has begun to suggest that the built food environment may play a role in an
individual‘s dietary decision making process (Baker et al. 2006b; Cummins and Macintyre 2006;
Kamphuis et al. 2006; Morland et al. 2006; Jago et al. 2007). Research has documented that the
size of supermarkets and the market area that they are meant to serve has increased in recent
years, which has led to a decline in the total number of food stores available in urban areas
(Whelan et al. 2002; Clifton 2004). Analyzing the impact that this change has had on the overall
food environment is important because of the potential for long lasting behavioral changes in
dietary habits (Swinburn et al. 1999). Therefore, it is important to examine how neighborhoods
may have been impacted by changes in the built environment as a result of the movement of
large scale supermarkets to the urban periphery.
Food environments are complicated and multi-dimensional and have therefore proven
difficult to define. Although much research has been done in recent years on food environments
and their influences on dietary behavior there still is not a strong conceptual model to help
establish a firm definition (Laraia et al. 2004; Ball et al. 2006). Three primary terms have been
used to describe the spatial distribution of food retailers and the corresponding access to food
10

items within a specified geographic area; food desert, nutrition environment, and food
environment (Leviton 2010). Initially much geographic research utilized the term food desert as
a way to define areas where inexpensive, nutritious food items are unavailable (Lavin 2005;
Beaulac et al. 2009). However, research has begun to deviate from this term as the concept of a
food desert tends to focus only on the places that have inadequate access to food items and does
not really describe all of the different interactions taking place in the environment (Glanz et al.
2005).
Recently research has begun to use the terms nutrition environment and food
environment to define the access to food items in a defined area. Glanz et al. (2005) used the
term nutrition environment to define the location of food retailers and their accessibility, the
availability and cost of healthy options, how healthy food items are displayed and promoted, and
the different geographical locations an individual operates in a given day. While all of these
components are certainly essential elements to the food environment Ohri-Vachaspati and
Leviton (2010) dispute the use of the term nutritional environment because of the fact that the
word nutrition is more commonly thought of as the interaction of food items with a person‘s
body and not the physical environment. Instead, researchers have begun to gravitate towards the
term food environment, which can be broadly defined as the physical availability and
accessibility of food retailers within a distinct physical boundary (Morland et al. 2002b; OhriVachaspati and Leviton 2010).
The definition of the food environment that was utilized by this research was adopted
from previous research conducted by Duvall et al. (2010) and Goldsberry et al (2010). In terms
of this thesis the Lansing food environment is defined as the physical availability of fresh
produce items and the accessibility of food retailers who sell fresh produce items within the
11

Greater Lansing area. Fresh produce was used as an indicator of the overall food environment
based on previous research that has used the availability of fruit and vegetable items as an
indicator of the food environment because of the nutritional diversity that these items bring to a
healthy diet (Morland et al. 2002a; Bertrand et al. 2008).
2.5 Accessibility
The word ―access‖ is often used in the vocabulary of researchers who are studying food
environments in reference to the ability of area residents to obtain access to food items within the
food environment. However, coming up with a satisfactory understanding of what the ability to
access the food environment means has proven to be extraordinarily difficult. Peter Gould once
wrote that, ―accessibility…is a slippery notion…one of those common terms that everyone uses
until faced with the problem of defining and measuring it‖ (Gould 1969, 64). While defining the
idea of access is extraordinarily difficult the concept is very important when considering how
residents interact with their neighborhoods (Church and Marston 2003). Previous research has
indicated that there are three essential components that characterize the idea of accessibility: (1)
locations of destinations, (2) the characteristics of the potential destinations, and (3) the ability to
reach potential destinations (Burns and Golob 1976; Handy and Niemeier 1997). Recent
research on food environments has begun to examine how those three components interact with
residents‘ ability to access their local food environment.
2.5.1 The automobile and access
In 1970 Hagerstrand developed the space-time framework, which was incredibly
influential in the measurement of access. The space-time framework states that there are both
spatial and temporal dimensions to how an individual takes part in an activity and that the
transportation system can be a limiting factor in terms of the time it takes an individual to reach a
12

final destination. In the case of access to the food environment the space-time framework is
certainly influential in understanding people‘s ability to access food items. Today in many
places in the United States one of the most important components necessary to access the food
environment is the automobile. Since the 1970s automobile use in the United States has
increased dramatically for commuting to work and small trips (French et al. 2001). As
supermarkets have moved to the urban periphery adequate access to an automobile has become
increasingly important in order to obtain access to food retail stores (Pearce et al. 2006). The
automobile has therefore become an essential part of an individual‘s ability to access food items
within their food environment.
In the US and North America research has found that healthy food items are often more
available to individuals with access to automobiles (Bertrand et al. 2008). Unfortunately, there
are often disparities in who has access to an automobile and as a result who has the ability to
obtain healthy food items. Individuals who have a high socioeconomic status are more likely to
own an automobile and are therefore not as likely to depend solely on the stores located in their
neighborhood (Wang et al. 2007). On the other hand, African Americans and those with low
socioeconomic status have been found to have less access to private transportation (Morland et
al. 2002b). Research has indicated that urban residents that have a low socioeconomic status
have different shopping habits as a result of the fact that they do not have reliable access to
transportation (Hendrickson et al. 2006). Specifically, research has suggested that for the
subpopulation who lacks regular access to an automobile has limited access to healthy food items
as a result of their inability to obtain access to food stores that provide these products.

13

2.5.2 The food environment and access
In the context of food environments access is often defined as the ability to obtain healthy
food items (Donkin et al. 2000; Morland et al. 2002b). However, surprisingly few different
approaches have been utilized by researchers in order to model the accessibility of food
environments. Many studies obtain information on food store locations from previously
established databases such as local departments of health and agriculture and InfoUSA (Morland
et al. 2002b; Algert et al. 2006; Bertrand et al. 2008; Franco et al. 2008). Some researchers did
conduct on-site inspections on a limited number of stores and validated other locations using
phone records, telephone conversations, store descriptions, and food license records (Algert et al.
2006; Franco et al. 2008). Few research studies actually checked the list of food stores that they
obtained from databases, and much research on food environments is conducted without
ensuring if the retail food stores identified are still in business.
There are two primary methodologies that are utilized by researchers to measure the food
environment, density based approaches, and proximity based approaches. The density approach
measures the number of available retail outlets in defined area using a buffering method, kernel
density estimation or spatial clustering (Charreire et al. 2010). Overwhelmingly the dominate
methodology in this research is to use the census tract or block in which an individual lives as a
proxy for their local food environment. In other words, these models make the assumption that
the only accessible food stores to an individual are those located in the census tract or block of
residence (Diez-Roux et al. 1999; Morland et al. 2002a,b; Inagami et al. 2006; Howard and
Fulfrost 2007; Wang et al. 2007; Franco et al. 2008). A major flaw with this methodological
approach is that just because an individual lives in a census tract or block does not mean that
they are limited to only shopping within that same census tract or block (Morland et al. 2002a).
14

Census tracts and blocks are also not necessarily representative of a neighborhood or community.
In a given day people function in a variety of different environments, including (but not limited
to) where an individual shops, lives, and works (Ball et al. 2006). Research has also identified
that there are often barriers that impact whether or not individuals shop within their own
community. Hendrickson et al. (2006) found that cost, quality of food and food choice limited
whether or not study participants shopped locally in their community or chose to shop in a
location outside their community where food items were less expensive.
Even researchers who utilize this type of methodology recognize that in an ideal situation
the smallest possible unit of analysis should be used to examine food environments (Howard and
Fulfrost 2007). A methodology that utilizes zonal objects introduces the modifiable areal unit
problem (MAUP) and edge effect problems to the research as the borders of the regions used are
arbitrary and potentially modifiable introducing bias into the results (Openshaw 1983). A
methodology based solely on zonal objects is also at odds with the space-time framework as it
fails to take into account the temporal dimension of access. Just because a food retail store is
located in a census tract that an individual resides in does not mean that individual can
reasonably access that store through the transportation network available. There may be a store
located in a neighboring census tract that is easier to access and this factor is left out when an
individual‘s food environment is limited to the census tract or block in which they reside.
As a result of the flaws associated with a methodology utilizing zonal objects researchers
have introduced several different methods to characterize and model the local food environment.
One of these methodologies uses the distance from available food retail outlets through the
established travel network to measure the local food environment (Charreire et al. 2010). This
type of approach uses network distance as a method of analyzing access to the food environment
15

(Algert et al. 2006; Goldsberry et al. 2010) and measure the actual distance that it would take a
person to travel to a food store along the available transportation network. For example Donkin
et al. (2000) established indices that measured the cost and availability of items within a
geographical area in London. Other research has suggested that the benefit of using multiple
models of the food environment to identify areas that are underserved by food retailers (Clarke et
al. 2002; Charreire et al. 2010). Overall, the concept of modeling the food environment is still in
its preliminary stages and there is much research that must be done in order to establish more
accurate models of urban food environments.
2.6 Research on spatial variation within food environments
There are two main ways that the food environment can impact an individual‘s behavior:
how a person perceives the environment and the actual physical characteristics of the
environment (Ohri-Vachaspati and Leviton 2010). Much previous research on food
environments has focused on the physical characteristics of the environment and more
specifically the presence (or lack thereof) of different retail store types within a food
environment. The implication of this research is that the spatial variation present within the food
environment impacts how individuals behave. Research has indicated that the presence of
supermarkets in a food environment can lower the price of groceries and residents of
neighborhoods that do not have access to a grocery store have been found to pay more for
groceries (Lavin 2005). The introduction of farmers markets has also been found to increase the
availability of healthy food and decrease the cost of food items within the neighborhood (Larson
and Gilliland 2009). Much research has focused on the disproportionate lack of access to
supermarkets and large scale grocery stores that some sub-populations have. The primary
assumption of this research is that the level of access an individual has to a food item is directly
16

related to how likely they are to purchase and consume that item. However, most of the research
that has been conducted on this particular topic has employed indirect and intermediate methods,
which do not involve an interaction between the researcher and the environment (Booth et al.
2005). As a result this research relied on assumptions that were not based on actual information
obtained from people who interact with the environment.
2.6.1 Socio-economic status
As food retail stores that are meant to serve large geographic areas are built many food
stores located in or near neighborhoods with low socioeconomic status have been closed.
Unfortunately new stores have not located in these areas, which has created a massive deficit of
available food options in neighborhoods with low socioeconomic status (Clifton 2004). More
importantly, neighborhoods with a low socioeconomic status have been found to have less
calculated access to food items (Larson and Gilliland 2008; Beaulac et al. 2009) and less
available healthy food items (Campbell et al. 2004; Andreyeva et al. 2008; Baker et al. 2006a;
Hendrickson et al. 2006; Franco et al. 2008). Neighborhoods with low socioeconomic status are
also more likely to have fresh produce available that is of a poorer microbial quality, which
could potentially impact the appeal of these fresh fruit and vegetable items (Koro et al. 2010).
Research has found that there is a difference in the types of stores available in
neighborhoods with different socioeconomic status. Supermarkets have been found to be more
likely to be located in neighborhoods with a high socioeconomic status (Moore and Diez Roux
2006; Franco et al. 2008). Supermarket chain stores, which have the lowest prices on food items,
tend also not to be located in neighborhoods with low socioeconomic status (Chung and Myers
1999). In contrast, corner stores tend to be much more common in neighborhoods with low
socioeconomic status (Moore and Diez Roux 2006; Bertrand et al. 2008; Larsen et al. 2009).
17

Unfortunately the cheap food items which are available in these types of stores are often high in
fat and sugar (Cummins and Macintyre 2002). There appears to be an overwhelming disparity in
the level of access to healthy food items between high- and low socioeconomic status.
Little research focuses exclusively on the accessible food environment for individuals
who receive nutritional assistance. Algert et al. (2006) found that those who receive nutritional
assistance through food pantries in Los Angeles have less access to food stores that carry a
variety of fresh produce products, with 41% of food pantry clients lacking pedestrian access to
fresh produce products. However, this appears to be an area of research with significant room
for exploration.
2.6.2 Ethnicity
Research has indicated that predominantly minority urban areas have less access to
supermarkets and significantly more access to small scale grocery stores that have fewer fruit
and vegetable options (Moore and Diez Roux 2006; Morland and Filomena 2007; Beaulac et al.
2009; Larsen et al. 2009). In general most research that examines ethnic differences in access to
food retailers has focused on differences for African Americans and white (non-Hispanic)
populations (Howard and Fulfrost 2007). Research has found that African American
neighborhoods in urban areas have less access to healthy foods (Sloane et al. 2003; Baker et al.
2006a; Moore and Diez Roux 2006; Franco et al. 2008). African American communities are also
less likely to have access to fruit and vegetable items (Sloane et al. 2003). Supermarkets are also
more likely to be located in white neighborhoods then in African American neighborhoods
(Morland et al. 2002b; Zenk et al. 2005b; Morland and Filomena 2007; Franco et al. 2008).
Neighborhoods in which the majority of residents are Latino have been found to have less
density of retail food outlets (Howard and Fulfrost 2007). Research has clearly indicated that
18

there is a disparity in the availability of healthy food items between neighborhoods where white
(non-Hispanic) populations reside and neighborhoods where minorities reside.
2.7 Research on the impact of spatial variation in food access on dietary behaviors
Some research has also suggested that the spatial variation present in the local food
environment has an impact on individuals‘ dietary choices (Baker et al. 2006b; Cummins and
Macintyre 2006; Morland et al. 2006; Jago et al. 2007). There may be a relationship between the
local food environment and obesity and overweight rates within that neighborhood (Morland et
al. 2006)and a direct relationship has been established between the availability of fruits and
vegetables and their consumption and changes in availability will result in changes in
consumption practices (Jago et al. 2007). While there has been a plethora of research examining
the association between environmental influences and dietary practices around the world,
particularly in Europe, there only appears to be a relationship between access to food items and
dietary behavior in North America (Cummins and Macintyre 2006). However, associations
between the food environment and dietary behavior in the United States are still uncertain.
2.7.1 Food environments where healthy food is not available
One approach to understand the relationship between the food environment and dietary
behavior is to examine neighborhoods where healthy food items are unavailable. Residents of
neighborhoods that do not have access to healthy food items have been found to be at an
increased risk for obesity and overweight (Wang et al. 2007). The presence of convenience
stores in a neighborhood has also been associated with a higher amount of overweight and obese
individuals (Morland et al. 2006). Finally, research has found that there is a positive correlation
between the availability of healthy food products and the healthfulness of resident‘s diet at both a
community and zip code level (Cheadle et al. 1991). The results of this research indicate that
19

there might be an association between the fact that there are higher rates of obesity in
neighborhoods that have less access to healthy food items.
2.7.2 Importance of supermarkets
Much research on the food environment focuses on the supermarket because of its unique
ability to provide a plethora of healthy food items at a low cost (Lavin 2005). Research has
indicated that individuals who shop at supermarkets consume a significantly healthier diet than
individuals who do not shop at supermarkets. African-American women who shop at
supermarkets have been found to consume more fruits and vegetables than those who shop at
independent grocers (Zenk et al. 2005a). Also, pregnant woman who live more than 4 miles
from a supermarket were found to have a significantly worse diet quality than pregnant women
who live near a supermarket (Laraia et al. 2004).
The presence of a supermarket in a local food environment has been found to be
associated with lower rates of obesity and overweight individuals in the neighborhood (Morland
et al. 2006; Larsen et al. 2009). A plethora of research studies have also found that the addition
of a supermarket to a neighborhood increases the consumption of fresh fruits and vegetables by
local residents (Morland et al. 2002a; Wrigley et al. 2002; Rose and Richards 2004). An analysis
of the Atherosclerosis Risk Study sponsored by the National Heart, Lung, and Blood Institute
(NHLBI) found that the addition of a supermarket to a census tract increased the consumption of
fruits and vegetables for white Americans by 11% and African-Americans by 32% (Morland et
al. 2002a).
2.7.3 Socioeconomic status
The socioeconomic status of a neighborhood has also been found to influence health
related behaviors (Baker et al. 2006b). Research has previously indicated that environmental
20

factors have an influence on the dietary choices of participants in the nutritional assistance
program (Rose and Richards 2004). Individuals with a low socioeconomic status have been
found to have lower food expenditures, lower fruit and vegetable consumption, increased
consumption of non-core foods including chips and sugar sweetened beverages, pay higher
prices for food items, and have overall lower quality diets (Bell and Burlin 1993; Diez-Roux et
al. 1999; Drewnowski and Specter 2004; Vinkeles Melchers et al. 2009). An association has also
been identified between living in a socioeconomically advantaged neighborhood and the
purchase of healthy food items (Turrell et al. 2003). Neighborhoods with a low socioeconomic
status have been found to have an increased risk for obesity and a higher Body Mass Index
(BMI) (Inagami et al. 2006; Wang et al. 2007). The lower quality diets that are found in
neighborhoods of low socioeconomic status are associated with the inadequate access these
neighborhoods have to healthy food items (Morland et al. 2002b; Rose and Richards 2004).
2.7.4 Mobility
Finally, research has indicated that the mobility of an individual impacts their ability to
achieve a healthy diet. An association has been found between individuals who have to travel
farther by automobile to obtain food items and a high BMI (Inagami et al. 2006). Elderly
individuals who have restricted mobility consumed inadequate amounts of fruits and vegetables
(Wylie et al. 1999). The level of mobility that an individual possesses is an important factor to
consider when attempting to understand access to healthy food items.
2.7.5 Level of impact of spatial variation in food access on dietary behaviors
There are a number of reasons why the availability of healthy food items may impact
dietary behavior. Most basically, increased availability of fresh produce items makes it more
convenient to purchase fresh fruits and vegetables (Jago et al. 2007). Visual cues, long term
21

exposure to healthy food items, and culture may also impact an individual‘s purchasing
preferences leading to an increased consumption of fresh fruits and vegetables (Jago et al. 2007).
However, research has indicated that it is often a number of different factors that contribute to an
increased risk of obesity and that consumption of healthy food items is not only influenced by
the food environment (Ford and Dzewaltowski 2008). Therefore it is often a combination of
access to healthy food items and the population residing in the neighborhood that influences
dietary behavior.
2.7.6 Food environment does not influence dietary behavior
Some research, however, indicates there is not enough evidence to suggest that there is an
association between access to food retailers and dietary habits (White 2007). Other research has
concluded that there is no association between the socio-economic status of a neighborhood and
a healthy food environment (Bertrand et al. 2008). Living near to a supermarket is not
necessarily associated with a lower risk of obesity (Wang et al. 2007). Research on the
interaction between the food environment and dietary behavior is incomplete. There is some
indication that there are neighborhoods with limited access to healthy food items and that in
many cases those same neighborhoods have poor dietary practices, however, there has not been
enough evidence to indicate that there is a clear correlation between the two phenomena. Often
the assumption is made that the limited access to healthy food items is the cause of the poor
dietary habits and obesity in those same neighborhoods. Correlation, however, does not imply
causation and few studies have examined the actual impact that the food environment has on
dietary behavior and how residents of neighborhoods actually interact with their local food
environment.

22

2.8 Behavioral geography
Human Geography is a science that draws upon theories from many different disciplines
in the social sciences. In the case of this research many of the concepts and principals of
behavioral research can also be utilized to help understand the interactions between the residents
of Greater Lansing and their food environment. Traditional human geographic analysis is very
cause and effect oriented and fails to take into account the larger reasoning the goes into how
people make decisions in their day to day lives (Norton 1997). One of the major downsides of
the introduction of theory to geographic research by the quantitative revolution of the 1950s and
1960s was that theory is based on rigid constraints that are not entirely in line with how the real
world operates (Golledge 2008). This left researchers seeking to better understand how an
individual‘s decision making eventually influenced their behavioral choices (Golledge 2008).
Behavioral geography is based on inferences that are made from the observed actions of
individuals (Gale 1972). When a person makes their decision of where to shop for food items
the problem is very complex, causing the person to weigh a number of different options, with
individual reasons for each part of their decision. Behavioral geography recognizes this
phenomena and attempts to provide an explanation for how these chains of decisions are made in
order to understand behavioral actions (Gale 1972). As opposed to the methods employed by
those involved in the quantitative revolution, behavioral geography examines individual
disaggregate behavior based on primary data (Golledge 2008). This research draws upon
behavioral geography in an attempt to understand how residents of Greater Lansing come to their
decision of what food items to consume utilizing primary data in order to help develop better
models of how to conceptualize the food environment.

23

2.9 Previous food environment research in Greater Lansing, MI
Previous models of access to fresh produce in Lansing created by Duvall et al. (2010) and
Goldsberry et al. (2010) provided the background measure of the food environment that was
used for this research. In order to identify the retail stores to be used in the models of the
Lansing food environment Duvall et al. (2010) and Goldsberry et al. (2010) established a list of
all potential food retail businesses (not including restaurants) that were located in Lansing. The
list was compiled based on phone book listings, Internet searches, on-the-ground surveys and
commercial data from ESRI based on the InfoUSA database (Duvall et al. 2010). After the list
of potential retail stores that sold fresh produce was established each store was checked through
visits and telephone calls to make sure that they were operating and that they sold fresh produce.
A total of 94 retail stores were identified that sold fresh produce in Lansing. Between February
and April 2008 data were recorded from each of these stores and a list of available produce items
was established for each store (Duvall et al. 2010). From this data Goldsberry et al. (2010)
established three different models of access to fresh produce in Lansing: a container method, a
weighted method, and a cumulative distance method. Each of these different models created
drastically different visualizations of the food environment in Lansing. The container and
weighted method models of access were also established for both pedestrian and automotive
access. A total of five models of access to fresh produce items in the Lansing food environment
were available to be used as measures of accessibility in the food environment for this research.
The container method of accessibility (Figure 2.1) in the Lansing food environment that
was used by Goldsberry et al. (2010) is a measure of the number of opportunities to purchase
fresh produce items within a 10 minute walk (.5 miles) for pedestrian access and a 10 minute
drive for automobile access. This type of accessibility measure was first developed by Wachs
24

and Kumagai (1973). One of the main flaws with this type of model of access is that not all
individuals have the same opportunity to purchase fresh produce items (Church and Marston
2003) and the container method weights an item that is a 10 minute drive away from an
individual the same as an item that is right next door.
Figure 2.1: Container method accessibility model of Lansing food environment (pedestrian)
(Goldsberry et al. 2010)

The weighted method of accessibility (Figure 2.2) established by Goldsberry et al (2010)
follows the same basic model as the container method but instead of using a blanket number of
opportunities this model gives higher accessibility scores to fresh produce items that are located
closer to an individual location then to fresh produce items that are located farther away. This
25

type of accessibility model was first applied to the social sciences by Hodgard (1978) and is a
slightly more complex measurement then the container method as it accounts for the distance to
obtain fresh produce items (Church and Marston 2003).
Figure 2.2: Weighted method accessibility model of Lansing food environment (pedestrian)
(Goldsberry et al. 2010)

Finally, the cumulative distance method (Figure 2.3) establishes an access score based on
the overall distance that a consumer would need to travel to obtain every produce item that was
available in Lansing. This method is another version of a weighted measure with a different set
of criteria. Instead of measuring the amount of items available within a certain distance this
26

model measures the total distance that an individual would have to travel from an initial starting
point to obtain all produce items that are available in the Lansing food environment (Goldsberry
et al. 2010). This model was only established for automobile access.
Figure 2.3: Container method accessibility model of Lansing food environment
(Goldsberry et al. 2010)

All five of these separate models of access are used as measures of the food environment
in Greater Lansing for this thesis. This decision was made because each different model
established a drastically different picture of the Greater Lansing food environment. Previous
research has found that there are many benefits associated with conceptualizing the food
environment through multiple models of access (Clarke et al. 2002; Charreire et al. 2010). At
27

this point in time there is no straightforward methodology for modeling the food environment
and as a result it is important to understand how all of the different models that have been
established interact with data that has been collected on the ground. Comparing how these
models interact with survey data will help to achieve a greater understanding of the Lansing food
environment as a whole and enable better models of access to be established in the future.

28

CHAPTER 3: METHODS

3.1 Overview
The methods that were employed in this thesis were designed to help understand the
behavioral interactions between Greater Lansing residents and their food environment. Although
there has been much qualitative and observational assessment of food environments in urban
areas, there has been very little quantitative analysis attempting to understand how residents
actually interact with their food environment (Morland et al. 2002b). The primary objective of
this thesis is to understand the behavioral interactions between Greater Lansing residents and
their food environment using quantitative methods. Specifically, this thesis seeks to answer the
following research questions:
1) Is there an association between calculated access to fresh produce products and
the use of fruits and vegetables in the diets of Greater Lansing residents?;
2) Are there variables other than the retail food environment that influence the
dietary behavior of Greater Lansing residents?;
3) Do Greater Lansing residents believe that the food retail environment in Lansing
is adequate?;
4) Do Greater Lansing residents who receive nutritional assistance interact with the
food environment differently than Lansing residents who do not receive
nutritional assistance?
In order to achieve the primary objective and answer the research questions proposed by this
thesis, customers of food retail stores in Lansing were surveyed to obtain information on the
shopping and dietary behavior of area residents. Statistical, geographic, and spatial analyses
29

were performed as a means to understand and explore the interaction between Greater Lansing
residents and their food environment.
3.2 Study area
The Greater Lansing, Michigan area is considered to be a mid-sized metropolitan area
with an overall population of around 450,000 (ACS 2009). The specific study area that was used
for this research was approximately 1,200 square miles with an estimated population of 291,000
residents (Goldsberry et al. 2010). The city of Lansing is comprised of three counties Ingham,
Eaton, and Clinton and portions of all three counties were included in the study area for this
research. Based on 2009 American Community Survey data, the majority of Greater Lansing
residents are White (81%), with the rest of the population reporting that they were African
American (8%), Hispanic (5 %), Asian (3%), and Other (3%).
As a result of its situation as a mid-sized city with a heavy reliance on the automobile
Lansing provides an excellent case study to examine how area residents interact with their food
environment. Much of Lansing is difficult to access by walking or biking, which forces residents
to rely on automobiles or public transportation options to move around the city. As a result, the
built environment creates a number of obstacles for many of the residents of Lansing who lack
access to an automobile. In fact, many such residents also receive nutritional assistance and are
at risk for not achieving their required daily nutritional intake. In Ingham County 28% of those
that receive emergency food assistance have no access to an automobile for transportation (Food
Systems Project Annual Report 2010). Due to the lack of supermarkets in northwest Lansing,
almost one-third of this area‘s residents shop for their groceries at places other than the major
supermarket chains: Kroger, Meijer, and L&L (Food Systems Project Annual Report 2010).

30

3.3 Questionnaire development
The decision to use a street-intercept methodology, randomly sampling patrons of food
retailers located in the Greater Lansing area, to understand how Greater Lansing residents
interact with their food environment was made based on previous research indicating that this
was the best way to obtain responses from segments of the population that have traditionally
been more difficult to reach with random-digit dialing telephone methodologies (Bush and Hair
1985; Miller et al. 1997). Previous literature identified that low-income groups, minority groups,
and those receiving nutritional assistance are often the most impacted by their surrounding food
environment (Morland et al. 2002a; Rose and Richards 2004; Baker et al. 2006b). Therefore, it
was essential to obtain sufficient responses from these more difficult to reach subpopulations to
understand how they interact with the Lansing food environment.
The benefits associated with a street intercept methodology including: increased response
rates from hard to reach segments of the population, higher response rates to sensitive questions,
and the ability to see firsthand the interaction that consumers have with their shopping
environment, made this method a clear choice to administer the questionnaire for this thesis.
Bush and Hair (1985) found that there was a consistent pattern of fewer refused questions and
more accurate responses when administering their survey in person. In-person surveys also had
the added benefit of allowing the researcher to obtain both more in-depth responses and
additional information that are potentially lost when conducting research over the phone (Bush
and Hair 1985). Survey participants are also more likely to divulge sensitive information, like
their income level, with an in-person surveying technique in comparison to a survey that was
conducted over the telephone (Miller et al. 1997). Previous research on food environments has
also utilized a street-intercept methodology. Baranowski et al. (2006) used this method to recruit
31

participants for their research in front of grocery stores, however, the actual interview for their
research took place over the telephone.
Once a street-intercept methodology was selected as the method to interact with research
participants a questionnaire was developed that was suitable for this particular methodology.
The questionnaire was designed in an effort to gain as much information as possible from
Greater Lansing residents within a short interview period of five minutes. The questionnaire was
divided into six sections: interviewer notes, transportation, consumer behavior and shopping
habits, diet and use of produce, location of residence and demographics. A copy of the final
survey instrument used in this research can be found in Appendix A.
3.3.1 Interviewer notes
The questionnaire instructed the researcher to take notes and record the location of the
survey, date and time the survey was conducted, as well as the weather at the time of the
interview. The notes from the interviewer were collected in order to ensure that an adequate
sample of the Lansing retail customers had been obtained and that there was no bias within the
sample towards individuals who shop at a particular time of day or day of the week.
3.3.2 Transportation
Questions related to access to transportation were asked in an attempt to understand the
various transportation options available to Greater Lansing residents. These questions were
designed to determine the level of access Greater Lansing residents had to automobiles, the mode
of transportation that the respondent used to get to the retail location and whether or not the
shopping trip was exclusively to go to the retail location or if the trip had multiple destinations.
This information was used to help determine how Greater Lansing residents access their food
environment.
32

3.3.3 Consumer behavior
The next section in the questionnaire was designed to help understand consumer behavior
and shopping habits. Greater Lansing residents were asked if they were the primary food
preparer in their household and the types of retail locations where they typically purchase food
items. This information was obtained in an effort to understand how involved the respondent
was with the food purchasing in their household. Questions that were related to shopping
behavior were used to determine the type of retail stores that Greater Lansing residents shop for
their food products at.
3.3.4 Diet and use of produce
The use of fresh produce in the diet of Greater Lansing residents was measured by
questions that focused on the number and type of fresh produce items purchased, the use of
frozen and canned produce items, and gardening activities. The information provided in this
section of the survey was used to help gauge the use of fresh produce items in the Greater
Lansing residents‘ diets.
3.3.5 Location of residence
Location of residence for Greater Lansing residents was determined by asking
participants to provide the address of their primary residence. If survey participants were not
comfortable providing their physical address then they were presented with the option of
providing the intersection that is located nearest to their primary place of residence.
3.3.6 Demographic information
Questions regarding demographic information were asked at the end of the survey. Since
survey participants can feel uncomfortable disclosing demographic information, the
questionnaire was designed in order to ensure that respondents completed the bulk of the survey
33

before they were asked to divulge personal information. The demographic information
participants were asked to disclose included: gender, year of birth, number of people residing in
their household, ethnicity, participation in food assistance programs and annual household
income.
3.3.7 Pretesting and IRB approval
Pretesting of the questionnaire was performed on graduate students in the geography
department in order to ensure that questions were easy to understand and answer. Once
questionnaire design was completed the questionnaire was submitted to Michigan State
University‘s Institutional Review Board (IRB) for review. IRB approval was granted on May
28, 2010 (IRB #10-491) with Dr. Kirk Goldsberry as the primary investigator. When IRB
approval was granted the process of selecting study sites began.
3.4 Selection of study sites
The research previously conducted by Duvall et al. (2010) and Goldsberry et al. (2010)
was used to help inform methodological decisions and the survey instrument implementation.
Study site selection was based on the classification of 94 food retail establishments performed by
Duvall et al. (2010). Duvall et al. (2010) found that the majority of retail stores in Lansing that
sold fresh produce items were supermarkets, generalist grocers and convenience stores. Previous
research, however, has also suggested that there is significant variability in products stocked at
food retail stores depending on where the store was located and whether or not there were other
retail stores located nearby (Guptill and Wilkins 2002; Duvall et al. 2010). Therefore, a variety
of different categories of food retail store types were selected based not only on the category of
store but on the relative importance of that store within the food environment.

34

Stratified random sampling was used to select the 30 initial stores to contact based on the
list of food retailers that provided fresh produce compiled by Duvall et al. (2010). Two
additional farmers markets that were not included in the list compiled by Duvall et al. (2010)
were added to the list in order to gain greater perspective on how Greater Lansing residents
interact with their food environment. Research has indicated that stratified random sampling is
on the average just as accurate as random sampling (Cochran 1946), while still ensuring the
inclusion of places deemed important by the researcher. A stratified random sampling
methodology was selected for store survey locations in order to ensure certain food retail stores
were included in the analysis, while still maintaining a level randomness in the sample
population. Certain stores were specifically selected to survey based on their footprint in the
Lansing food environment. The list of preselected stores included the East Lansing Meijer, the
Okemos Meijer, and the Kroger located in the Frandor mall. The remaining stores to contact
were selected by generating random numbers. Initially, two farmers markets, four convenience
stores, two discounters, two ethnic food stores, four grocers, two liquor stores, two organic
specialists, and six supermarkets were selected to contact.
Initial contact of the selected study sites began on June 3, 2010. Soon after contact of the
selected study sites began it became apparent that getting permission from retailers, especially
supermarkets and grocers, would be very difficult and many more stores would need to be
contacted in order to ensure an adequate sample. Meijer corporate offices very quickly
responded that no solicitation was allowed on any of their property, which meant that Meijer
store property could not be used for this research. Quality Dairy, which is an important provider
of fresh produce products in the Lansing area as well as the major grocery chain in the region,
also declined to participate in the project. Unfortunately, the lack of willingness from many of
35

the major retailers in the area forced the selection of study sites in some cases to be based more
on retailers that were willing to participate in the project then on a complete stratified random
sampling of area food retailers. A strong effort was made, however, to ensure that there was
diversity in both the geographic location and the type of retailers that were surveyed. Ultimately
40 retail stores were contacted to participate in the research and twelve study sites were
eventually selected to survey. The list of study sites included two 7-Eleven convenience stores,
Vallarta Supermarket, Carl‘s Supermarket, the Lansing City Market, Sav-On Market, Rainbow
Party Store, Better Health, Foods for Living, Horrocks Farm Market, and two Kroger
Supermarkets (Figure 3.1).
Figure 3.1: Location of study sites

36

The list of selected study sites covers a wide variety of types of retail stores from party
stores to full-scale supermarkets. In terms of categories of produce retailers that were established
by Duvall et al. (2010) the list of study sites included, three supermarkets (Frandor Kroger,
Okemos Kroger, and Horrocks Farm Market), two discount grocery stores (Vallarta Supermarket
and Carl‘s Supermarket), one farmers market (Lansing City Market), two organic specialists
(Better Health and Foods for Living), two liquor stores (Sav-On Market and Rainbow Party
Store) and two convenience stores (East Mt Hope Ave 7-Eleven and East Grand River Avenue 7Eleven).
3.5 Sampling method
Once permission was granted by the manager of a selected retail store, times when the
store would be able to be surveyed were established. Numerous stores offered to allow the
research to be conducted inside and some stores even offered to allow the research to be
conducted in the produce department to watch customer shopping behavior. However, as a result
of the fact that not all stores offered this level of access the research always took place outside by
the store entrance in order to maintain a level of control in the results. During the designated
periods of time when I was allowed at the stores, generally between one to two hours,
participants were selected as their entered or left the retail store. Potential participants were
asked if they wanted to participate, informed of the $100 gift card drawing, and were given the
information necessary for them to make informed consent. If potential participants gave
informed consent an interview was then conducted. Conducting the survey took between four
and seven minutes depending on how long it took participants to respond to the questions. At
least fourteen interviews were conducted at each of the different food retail stores and each retail
establishment was surveyed for a total of at least three hours. Some stores were surveyed over a
37

longer period of time if they had less foot traffic because it took longer to get the desired number
of interviews.
The actual process of administering the survey took place from June 5, 2010 until July 8,
2010. Greater Lansing residents who were shopping at study locations were surveyed from
10am until 6:30pm on all seven days of the week. The time in which Greater Lansing residents
were surveyed was often dictated by the store managers and when they would allow the research
to take place although an effort was made to ensure that there was a diversity of different times at
each store surveyed. Most of surveying for this project took place during sunny weather
conditions (67%), although Greater Lansing residents were surveyed on cloudy (28%), rainy
(1%), and overcast days (4%). Unfortunately, it was difficult to administer surveys on rainy
days as all of the interviews took place outside and potential survey participants were not
interested in standing in the rain. A total of 185 interviews were conducted for this research
project. Once sampling was completed the responses were coded and entered into Microsoft
Excel.
3.6 Statistical analysis
Data analyses were conducted in order to obtain general information about the dataset,
establish associations between the amount of calculated access Greater Lansing residents had to
fresh produce and the use of fresh fruits and vegetables in their diet, determine variables that
influenced Greater Lansing residents‘ dietary choices to inform future models of accessibility,
and analyze the impact that access had on those who receive nutritional assistance. In order to
answer the primary research objective of this thesis a variety of different statistical methods were
employed including basic statistics, correlation analysis, backwards stepwise regression,
geographically weighted regression (GWR), and two-sample t-tests.
38

A review of the data using basic statistics was conducted in order to ensure that the data
were an accurate representation of the greater Lansing population. The information obtained
from this review was also used to help answer the research question; do Greater Lansing
residents believe that their food environment is adequate? Responses from participants were
analyzed in order to understand how Greater Lansing residents perceived their current food
environment.
To conduct the data analysis for this research a number of different inferential statistical
models were used. Inferential statistical models were originally designed to handle experiments
that took place under controlled conditions (Wrigley 1983) and were based on a very specific set
of assumptions. Controlled geographic experiments are often the exception rather than the rule
and that it is important to take into account the nature of the data when using inferential
statistical models (Gould 1970; Goodchild 2009). Goodchild (2009, p.441) argues that
quantitative analysis in the social sciences should move in a new direction, ―…that treats natural
experiments, spatial dependence, and spatial heterogeneity as the norm‖. Since there was no
way to overcome the fact that the data obtained in this research did not perfectly meet the
assumptions of inferential statistical models diagnostic tools were utilized before each model was
employed in order to understand how the data fit with the assumptions of the model. These tools
were also essential in understanding the potential problems and limitations of the models that
were created (Wrigley 1983).
3.6.1 Calculation of access scores
A method was established in order to determine the level of access each survey
participant had to fresh produce items. Research subjects were asked to disclose either their
primary residence or an intersection that was located close to their home and from that
39

information geographic coordinates were obtained using Google Maps. Once the geographic
coordinates were obtained for the primary residence of each survey participant, the data were
uploaded into ArcMap along with the raster files that contained the accessibility scores of the
five models of the food environment in Lansing established by Goldsberry et al. (2010). Each
individual primary residence was given an access score based on the raster file for each of the
five models that were calculated by Goldsberry et al. (2010). Out of the 185 individuals who
were surveyed a total of 183 addresses were obtained. Each address was given five access scores
that could be used as a variable in further statistical analysis.
3.6.2 Correlation
Correlation analysis was used to measure the association between calculated access to
fresh produce and the use of fresh fruits and vegetables in the diets of Greater Lansing residents.
The information obtained from this data analysis was used to help answer the research question;
is there an association between calculated access to fresh produce products and the use of fresh
produce in the diets of Greater Lansing residents? All of the correlation analysis for this research
was performed using the MYSTAT statistical software package. In order to understand the
association between calculated access to fresh produce and the use of fresh fruits and vegetables
in the diets of Greater Lansing residents correlation analysis was performed between the three
variables that measured the consumption of fresh produce products and the five access variables
that were obtained from the research conducted by Goldsberry et al. (2010). The results of this
analysis were used to determine whether or not the level of access that was calculated by the
models created by Goldsberry et al. (2010) actually predicted the dietary choices of Greater
Lansing residents.

40

3.6.3 Backwards stepwise regression
Backwards stepwise regression was used in order to determine the variables that had the
most influence on the consumption of fresh produce items by all Greater Lansing residents and
those residents that receive nutritional assistance. The results obtained from this data analysis
were used to help answer the research question; are there variables other than the retail food
environment that influence the dietary behavior of Greater Lansing residents? Regression
analysis was one of the first statistical techniques that was used in geographic research (Hartman
and Hook 1956). In terms of geographic analysis, regression is often used to help describe
relationships between variables, interpolate between observations, test hypotheses, and
extrapolate information (Mather and Openshaw 1974). For this thesis bivariate backwards
stepwise regression analysis was used to help establish models that predicted the consumption of
fresh produce by Greater Lansing residents. All bivariate regression analysis that was conducted
for this research was performed using the MYSTAT statistical software package. Three models
were created to predict the three variables that measured the consumption of fresh produce items
by all Greater Lansing residents.
All of the models that were established for this research were examined to ensure that
they met the assumptions of a linear regression model. The assumptions of linear regression
models were outlined by Poole and O‘Farrell (1970) and include (1) the values are obtained
without measurement error, (2) the relationship between the variables is linear, (3) all conditional
distributions have a mean of 0, (4) there is no hetero-scedasticity present in the residuals, (5) the
residuals are independent of each other, (6) there is no multicollinearity present between the
independent variables, and (7) the distribution of all the variables are normal. In the context of
geographic research, however, it is often difficult to ensure that all of these assumptions are
41

completely met. Gould (1970, p.446) wrote that ―given the fact that the assumptions seldom
hold individually, and that the chances of them holding severally is virtually zero, it is difficult
not to think of traditional, non-spatial inferential statistics as totally irrelevant as correspondence
rules in geographic enquiry‖. Gould felt that in geographic inquiry satisfying the assumptions of
a linear model is nearly impossible and therefore it is important to take that information into
account and make decisions based on your individual data. When establishing the models for this
thesis an effort was made to ensure that all of the assumptions of a linear regression model were
met, however, when the assumptions could not be met an effort was made to explain why that
particular model was still an accurate representation of the data presented and how the results of
these models might be limited based on the fact that they violated the assumptions of the model.
Stepwise regression models are incredibly useful because the conditions of the model are
established before the model is created, enabling the researcher to ensure that their data meets the
assumptions of a linear regression model to the best of their ability. The backwards stepwise
regression process results in a traditional ordinary least squares (OLS) regression equation:
Equation 3.1: OLS regression
Y = β0 + β1X1 + β2X2… + βnXn + ε
An ordinary least squares regression equation seeks to predict the value of a variable (Y) based
on the value of other independent variables (X). Independent variables (n) are selected for entry
into the model based on their statistical fit with the dependent variable and not previously
defined hypotheses. In the case of backwards stepwise regression modeling all of the initial
candidate variables are entered into the model and then eliminated based on whether or not they
meet the criteria that were established by the researcher. Coefficients (β) measure the strength
and type of relationship that each independent variable has with the dependent variable and the
42

residuals (ε) are a measure of the random error present in the model. In the case of this research
backwards stepwise regression was useful because it was important to determine what the
variables that predicted the consumption of fresh produce were, as previous research was not
entirely clear on what those variables might be.
3.6.4 Geographically weighted regression
Geographically weighted regression (GWR) analysis was used mainly as an exploratory
technique to measure the variation of regression coefficients over space and to help further
answer the research question; are there variables other than the retail food environment that
influence the dietary behavior of Greater Lansing residents? The information obtained from this
analysis was used to further understand how the physical location of Greater Lansing residents‘
home residence influenced their dietary behavior and to better understand the models that were
created from the backwards stepwise regression process. The analysis for this research was
performed using both ArcMap and R statistical software programs. In traditional regression
analysis geographic relationships between coefficients are often lost and smoothed over. GWR
allows researchers to further explore their data and identify possible variables that may be
missing from the model, as well as spatial variations that may be difficult to introduce into the
model but are worth recognizing (Brunsdon et al. 1998).
A traditional regression model makes the assumption that the parameters of the model are
the same across the entire (Equation 3.2).
Equation 3.2: Traditional regression
Y = β0 + β1X1 + ε
For a traditional regression equation each independent variable has the same coefficient across
all observations, which in the process can potentially smooth out some of the spatial variation
43

from the model. In the case of GWR an estimation of the model takes place at each observation
with other data points that are located near the observation achieving a higher weight.
Equation 3.3: Geographically Weighted Regression
Y(u,v) = β0(u,v) + β1X1(u,v) … + βnXn(u,v) + ε(u,v)
The result of a GWR model is that each individual observation has its own parameters and
coefficients measuring the strength and type of relationship that the independent variables in the
model have with the dependent variable, with (u,v) referring to the geographic coordinates of the
individual data points. This provides additional information about spatial variation within the
dataset. For this research the three regression equations that were established using backwards
stepwise regression were reexamined using geographically weighted regression. This additional
analysis was performed to obtain additional information on the validity of the backwards
stepwise regression models and to better understand the spatial variation in dietary behavior
among Greater Lansing residents.
3.6.5 Two-sample t-test
Finally, two-sample t-tests were employed in order to understand how the population of
those who receive nutritional assistance differed from the population who did not receive
nutritional assistance. The information that was obtained from this analysis was used to answer
the research question; do Greater Lansing residents who receive nutritional assistance interact
with the food environment differently than Greater Lansing residents who do not receive
nutritional assistance? Two sample t-tests are used to asses if the means of two different
populations vary significantly, in the case of this research the population of residents who
receive nutritional assistance and the population of residents who do not. The two sample t-test
was used to determine if the mean travel time to the retail store Greater Lansing residents were
44

surveyed at, the frequency Greater Lansing residents visited different types of retail stores, and
the consumption of fresh produce items differed between Greater Lansing residents who received
nutritional assistance and Greater Lansing residents who did not. The primary hypotheses for
these tests are that the mean travel time to store surveyed at is greater for those who receive
nutritional assistance than those who do not, that those who receive nutritional assistance shop at
convenience and liquor stores for food items then those who do not, and that those who receive
nutritional assistance consume less fresh produce items then those who do not.
The primary assumptions of this statistical model are that the data were collected using
random sampling, the samples are independent of each other and the data that is being examined
are interval/ratio data. There are two different models of two-sampled t-tests that can be
employed, separate variance and pooled variance. In the case of this research pooled variance
was used when the variance for those that receive nutritional assistance and those that did not
was approximately equal. If the variance for those that receive nutritional assistance and those
who did not receive nutritional assistance was not equal separate variance was used. The null
hypothesis for the two-sample t-tests that were employed was that there is not a significant
difference in the means for the variable selected for the population who received nutritional
assistance and the population who did not receive nutritional assistance. The alternative
hypothesis is that the observed mean for those that receive nutritional assistance is significantly
different than for those who did not receive nutritional assistance. The information obtained
from this analysis will be useful to help inform future models of access to food items in the
Lansing area and to understand how certain subpopulations may interact differently with their
local food environment.

45

CHAPTER 4: RESULTS AND ANALYSIS

4.1 Introduction
The results of the survey revealed important new information about the Lansing food
environment and helped to achieve the primary objective of this thesis to understand the
behavioral interactions between Greater Lansing residents and their food environment. Most
other research of food environments examines the food environment as purely a geographic
entity with assumptions that fail to take into account how local residents actually interact with
the food environment. The results and analyses in this chapter shed light on how Greater
Lansing residents actually interact with their food environment.
4.2 General survey results
Overall the survey sample of Greater Lansing residents appeared to be a good
representation of the Lansing population when compared to 2009 American Community Survey
(ACS) data (Table 4.1). For the most part the demographic information obtained from the
survey population was in line with the Lansing demographic information available from the
ACS. The mean age for survey participants was 44 years old, which is slightly higher than the
mean age of Greater Lansing residents reported by the 2009 ACS (34 years), however, part of
this discrepancy is a result of the fact that children under the age of 18 were not allowed to
participate in this research but were included in the 2009 ACS data.

46

Table 4.1: Demographic results

Survey Participants
Male
Female
Age (median)
White (non-Hispanic)
African-American
Hispanic
Ethnicity
Asian
Other
Refused
Household size (median)
Children younger than 18

49.2%
50.8%
44
70%
10%
7%
2%
10%
2%
2.7
27%

Gender

2009 American
Community Survey
48.3%
51.7%
34
81%
8%
5%
3%
3%
N/A
2.4
28%

There did appear to be a slight skew in the data as a high proportion of survey
participants (14%) reported that they earned less than $10,000 per year, which is not surprising
as one-fifth of the surveyed population reported some form of receiving nutritional assistance.
One-fifth of survey participants also did not know their annual household income or refused to
disclose that information. A diverse array of locations were reported as their primary residences
(Figure 4.1).

47

Figure 4.1: Primary residence of participants

4.2.1 Travel behavior
A private automobile was the primary mode of transportation that survey participants
used to gain access to all different categories of stores surveyed, although eight out of the twelve
stores surveyed were accessible by public transportation. A substantial majority of those
surveyed had regular access to an automobile (88%), with the average household having access
to two automobiles. A higher percentage of pedestrians were surveyed shopping at liquor (35%)
and convenience stores (22%) than other categories of food stores, which indicates that stores
that sell few produce items and convenience stores may serve a more local pedestrian based
48

community (Table 4.2). Health food stores (93%) had the highest percentage of customers that
accessed the store by automotive transportation, which indicates that these types of stores may
serve a much wider geographic area.
Table 4.2: Travel time based on retail store type

Store category

Avg. Travel
Time
(minutes
reported)
7.8

Auto

Public

Pedestrian

72%

3%

22%

4.6

81%

--

13%

Farmers market

12.5

81%

--

13%

Health food store

13.5

93%

3%

3%

5.8

59%

3%

35%

10.3

83%

13%

4%

Convenience
Discounter

Liquor
Supermarket

On average participants indicated that it would take approximately 10.4 minutes to travel
from the store at which they were surveyed shopping to their primary residence. However these
travel times varied depending on the type of food retail store where they were surveyed (Table
4.2). Participants were willing to travel the longest to health food stores (13.2 minutes),
supermarkets (12.4 minutes), and farmers markets (12.3 minutes). The shortest reported travel
time was to discount grocery stores (6.1 minutes) and to liquor stores (7.2 minutes).
Respondents (71%) indicated that they would visit multiple other destinations on the
shopping trip they were on when surveyed before returning to their primary residence. Most
survey participants (58%) reported that the food retail store they were surveyed at was closer to
their primary residence than other locations that they frequented. This finding indicates that food
retail stores in close proximity to a respondent‘s primary residence may have more importance in
their overall food environment. However, a significant number of participants (32%) did
49

indicate that the store they were shopping at when surveyed was nearer to location that they
frequented other than their primary residence including their work (21%), child‘s school or
daycare (2%), church (1%), gym (1%), or other location that they regularly travel to (7%).
Interestingly, 11% of participants reported that the store was not near any place associated with
their daily activities, which demonstrates that some individuals are willing to travel outside of
their regular routine to shop for food items.
4.2.2 Shopping behavior
Most participants did not seem to feel that access to adequate shopping was a problem in
the neighborhood surrounding their primary residence with a strong majority (82%) indicating
that it was only a ―minor problem‖ or ―not a problem‖. Only a small percentage (7%) of
participants felt that access to adequate shopping was a ―very serious problem‖ in their
residential neighborhood. Despite the fact that most participants did not feel that access to
adequate shopping was a problem participants did not seem to indicate that the majority of their
food shopping was done in their immediate neighborhood. A majority of participants (57%)
indicated that only ―some‖ or ―none‖ of their shopping was done within a 20 minute walk or
about a mile away from their home and only 22% of participants indicated that ―all‖ or ―almost
all‖ of their shopping was done near their primary residence. This finding suggests that while
Greater Lansing residents are generally satisfied with their food environment most residents are
not doing their food shopping in close proximity to their homes. More than four-fifths of those
surveyed indicated that they were the primary shopper in their household.
The supermarket was the most commonly visited food retail location and had the highest
frequency of visits (Table 4.3). Almost all participants indicated that they regularly shopped for
food items at supermarkets, with 94% indicating that they shopped for food items at a
50

supermarket at least once a month and 63% at least once a week. The average number of trips to
a supermarket per month reported by participants was 4.7, which indicates that participants
frequent supermarkets on a regular basis. Based on the survey results the supermarket is clearly
an essential component of the Lansing food environment.
Table 4.3: Consumer behavior based on retail store category

Mean number
of times
shopped at for
food items per
month:

Supermarket
Restaurant
Convenience store
Fast-food
restaurant
Grocery store
Health food store
Farmers market
Liquor store
Department store
Discounter /
warehouse club

Store category:

Percent who
never shop at
this category:

Percent who
shop at this
category at least
once a week:

Percent who
shop at this
category 3 or
more times per
week:

4.7
4.4
4.2

6%
16%
52%

63%
27%
28%

10%
12%
10%

3.7

27%

39%

10%

3.6
1.8

30%
64%

39%
15%

8%
3%

1.5
1.5
0.8

52%
82%
66%

22%
12%
5%

3%
5%
1%

0.7

76%

5%

1%

Most participants indicated that they ate out frequently, consuming fast-food (73%) and
eating at restaurants (84%) at least once per month. A small percentage of respondents indicated
that they consumed food from fast-food restaurants (10%) and restaurants (12%) more than 3
times per week. Unfortunately, restaurants and fast-food retailers were not included in the
models of the food environment that were established by Duvall et al. (2010) and Goldsberry et
al. (2010).

51

Alternative food stores did not appear to be a major component of the Lansing area food
environment with less than half of participants indicating that they shopped at farmers markets
(48%) and health food stores (36%) for food items. The subpopulation of participants that do
shop at these types of stores does so relatively frequently with 45% of that group shopping at
farmers markets and health food stores at least once a week. Less than a quarter of respondents
reported gardening at home (22%) or in a community garden plot (2%).
4.1.3 Dietary behavior
Overall respondents ―strongly agreed‖ or ―agreed‖ that they are careful about what they
eat (73%) and have a healthy (78%) and nutritious (73%) diet (Table 4.4). Almost all
participants (96%) reported that they consumed fresh produce products. Those who did not
consume fresh produce items cited the ease of preparing other items, taste preferences, and
availability as reasons that they did not consume fresh fruits and vegetables. Respondents
indicated that they prepared meals with fresh fruits and vegetables on average 23 times per
month, which is less than daily. Fresh produce items were the most commonly purchased fruit
and vegetable items with the average participant purchasing 7.5 fresh, 1.9 canned, and 1.7 frozen
fruit and vegetable items in a typical week. Most respondents preferred to prepare meals with
fresh fruits and vegetables (86%) in comparison to frozen (8%) and canned (6%) fruit and
vegetable items. Less than half of survey participants indicated that they purchased organic fruit
and vegetable items (46%).

52

Table 4.4: Participant diet perceptions

Strongly
agree
―I am a healthy eater‖
―I am someone who eats in
a nutritious manner‖
―I am someone who is
careful about what I eat‖

Agree

Neither
agree nor
disagree

Disagree

Strongly
Disagree

22%

56%

9%

12%

2%

22%

51%

10%

15%

2%

24%

49%

14%

12%

2%

4.3 Correlation analysis
Correlation analysis was used mainly as an exploratory technique to understand the
association between calculated access to fresh produce items and the use of fresh fruits and
vegetables in the diets of Greater Lansing residents. Three variables were used to measure the
use of fresh produce in the diets of Greater Lansing residents; the number of meals prepared with
fresh produce items each week, the estimated number of fresh produce items purchased each
week, and the listed number of fresh produce items purchased each week. These three variables
were then compared to the five access scores that were obtained from the analysis performed by
Goldsberry et al. (2010). The Pearson correlation values between access to fresh produce
products and the reported use of fresh fruits and vegetables by respondents indicate that there is
no relationship present between modeled access to fresh produce items and the reported use of
fresh fruits and vegetables in the diets of Greater Lansing residents (Table 4.5). The
consumption of fresh produce items by participants was not influenced by their calculated access
to fresh fruit and vegetable items. The strongest relationship based on Pearson correlation values
was between the cumulative distance method of access and the consumption of fresh produce;
however this relationship was still weak, with the highest Pearson correlation value being only
53

0.196. These results indicate that the models of the food environment established by Goldsberry
et al. (2010) do not predict how many fresh fruit and vegetable items a resident of Greater
Lansing will reportedly consume in a month.
Table 4.5: Pearson correlation values between modeled access to fresh produce and dietary
behavior

-0.005
-0.009
-0.027

Number of Fresh
Produce Items
(estimated)
Purchased each
Week
-0.043
-0.006
-0.083

0.007

-0.039

-0.034

0.196

0.169

0.115

Number of Meals
Prepared with
Fresh Produce
each Week
Automobile (weighted)
Automobile (count)
Pedestrian (weighted)
Pedestrian (count)
Cumulative Distance
Method

Number of Fresh
Produce Items
(listed) Purchased
each Week
-0.076
-0.048
-0.066

4.4 General population backwards stepwise regression models
Since correlation analysis indicated that there was no association between calculated
access to fresh produce products and the use of fresh fruits and vegetables in the diets of Greater
Lansing residents the next step was to understand what other variables might influence the
consumption of fresh produce products. Three separate backwards stepwise regression models
were established for each of the variables in the dataset that measured the consumption of fresh
produce in order to identify the variables that were particularly important in predicting the
number of fresh fruit and vegetable items consumed by the population.
For each of the three backwards stepwise regression models, all eligible variables from
the survey results were initially entered. A list of these variables can be found in Appendix B.
All eligible variables were entered into to the model because there was no clear understanding as
54

to which variables might influence the use of fresh produce in the diets of Greater Lansing
residents. The same criteria were established for all three backwards stepwise regression
models. The tolerance criteria for entry into the final regression model was purposefully kept
low (1e-012) because of the fact that there was likely to be high amount of correlation among the
independent variables that were entered into the model and it was important to keep all possible
variables eligible for entry. The probability for removal from the model was set to 0.15. An
initial analysis of each variable‘s distribution was conducted to check for normality. All of the
variables had normal distributions and no transformations were performed.
4.4.1 Model #1: Number of different produce items (listed) purchased each week
The first backwards stepwise regression model that was created used the listed number of
different produce items purchased each week as the dependent variable. The final model
employed four independent variables: how nutritious a respondent believed that their diet was,
gender, number of household members, and whether or not a household received nutritional
2

assistance. The final regression model did not have a large explanatory power (R = 0.181),
although the F-Ratio was significantly large (10.816) to reject the null hypothesis at a 99%
confidence level (p-value<0.001).

55

Table 4.6: Number of different fresh produce items (listed) purchased each week:
Backward stepwise regression results

Effect (Xn)
Constant (β0)
How nutritious
respondent believed their
diet was
Gender
Number of household
members
Whether or not a
household receives
nutritional assistance

1.539

Std.
error
1.750

-0.959

0.260

-0.260

.933

-3.696

0.000

1.628

0.521

0.218

0.944

3.121

0.002

0.266

0.182

0.105

0.901

1.467

0.144

1.922

0.666

0.209

0.882

2.886

0.004

Coefficient

Std.
Tolerance
coefficient
0.000

t

p-value

.879

.380

The coefficients of variables (Table 4.6) behaved as would be expected within the final
regression model. The more nutritious a respondent believed that their diet was the more likely
they were to purchase a large variety of fresh produce products. The coefficients of the variables
also indicated that women were more likely to list a larger number of fresh produce products that
they purchased each week. The coefficient of the regression model indicated that the larger the
household size the more different fresh produce items were purchased each week. Finally
households who received nutritional assistance purchased a less diverse variety of produce
products.

56

Figure 4.2: Number of different fresh produce items (listed) purchased each week: Model
residuals

The regression model established some observations that were outliers or exhibited
leverage; however, based on analysis of the individual observations there was no reason to
remove these observations from the analysis. The residuals of the model appear to be
homoscedastic and there is no serial correlation based on the scatter plot and histogram of
residuals. There appears to be a slight but not overly consequential left-skew to the residuals of
the model. There did not appear to by any multicollinearity present between the independent
variables. A map of the residuals (Figure 4.2) of the model indicated that there was no spatial
bias and the model was not predicting some areas substantially better than other areas. The
57

residuals appeared to be evenly distributed. While there is evidence that the model violated
some of the assumptions of a linear regression model there was not enough evidence to indicate
that these results were not valid.
4.4.2 Model #2: Number of different produce items (estimated) purchased each week
The second backwards stepwise regression model used the estimated number of different
fresh produce items purchased each week as the dependent variable. The final model resulted in
the inclusion of five independent variables: whether or not participants viewed neighborhood
access as a problem, how nutritious participants believed that their diet was, frequency of visits
to farmers market, income, and whether or not participants received nutritional assistance. This
2

final regression model that was established did not have a large explanatory power (R = 0.198);
however the F-Ratio was significantly large (9.870) to reject the null hypothesis at a 99%
confidence level (p-value<0.001).
Table 4.7: Number of different fresh produce items (estimated) purchased each week:
Backward stepwise regression results

Effect (Xn)
Constant (β0)
Neighborhood
access problem
Nutritious diet
Farmers market
frequency
Income
Nutritional
assistance

Std.
Tolerance
coefficient
0.000

Coefficient

Std. error

t

p-value

7.891

2.712

2.909

0.004

-.0854

0.460

-0.126

0.969

-1.857

0.065

-1.846

0.436

-0.292

0.936

-4.237

0.000

0.321

0.177

0.127

0.918

1.818

0.071

0.375

0.164

0.160

0.913

2,289

0.023

2.304

1.124

0.146

0.881

2.050

0.042

Again the coefficients of variables behaved as should be expected within the model
(Table 4.7) despite the fact that only three of the five variables included in the final regression
58

model had a relationship that was significant. According to the coefficients in the final
regression model those who felt their diet was nutritious were more likely to estimate that
purchased a greater variety of fresh produce products. Those who reported a higher income and
did not receive nutritional assistance were also more likely to estimate that they purchased a
diverse variety of fresh produce products. Respondents who indicated that they lived in a
neighborhood where access to adequate shopping was a problem and respondents who did not
shop frequently at farmers markets purchased a less diverse variety of fresh produce items.
The results and residuals of the model were examined in order to ensure none of the
assumptions of the model were violated. The distribution of the residuals had a slight left-skew,
which indicates there may be some heteroskedasticity in the residuals and may cause some doubt
about the reliability of using a ordinary least squares regression model to examine this
relationship. A map of the residuals (Figure 4.3) of the model indicated that the model was not
predicting some areas significantly better than others and the residuals appeared to be evenly
distributed.

59

Figure 4.3: Number of different fresh produce items (estimated) purchased each week:
Model residuals

4.4.3 Model #3: Number of times meals are prepared with fresh produce items in a week
The third backwards stepwise regression model that was established for this research
used the number of times that meals were prepared with fresh produce items in a week as the
dependent variable. Overall this final model had the highest explanatory power of all of the
models that were developed, most likely as a result of the number of independent variables that
were able to gain entry into the model. The final model resulted in the entry of eight
independent variables; whether or not respondents viewed their neighborhood access as a
problem, how nutritious respondents believed that their diet was, frequency of visits to fast food
60

restaurants, frequency of visits to farmers market, frequency of visits to department stores,
2

gender, age, and household size. This regression model had a larger explanatory power (R =
0.345); however the overall explanatory power of the model was still not very large. The FRatio was significantly large (12.925) to reject the null hypothesis at a 99% confidence level (pvalue=0.000).
Table 4.8: Number of times meals are prepared with fresh produce items each week:
Backward stepwise regression results

Effect (Xn)
Constant (β0)
Neighborhood access
problem
Nutritious diet
Fast food frequency
Farmers market
frequency
Department store
frequency
Gender
Age
Household size

24.610

Std.
error
4.015

-1.566

0.623

-0.153

0.981

-2.513

0.013

-3.911
-.0296

0.641
0.131

-0.412
-0.154

0.792
0.780

-6.098
-2.258

0.000
0.025

0.472

0.237

0.123

0.945

1.989

0.048

0.404

0.257

0.100

0.900

1.572

0.118

2.586
0.091

1.190
0.038

0.134
0.150

0.948
0.904

2.173
2.372

0.031
0.019

1.004

0.411

0.153

0.921

2.440

0.016

Coefficient

Std.
Tolerance
coefficient
0.000

t

p-value

6.130

0.000

The coefficients of variables (Table 4.8) behaved as anticipate within the final regression
model. According to the model those who felt their diet was nutritious were more likely to
consume fresh produce. Those respondents who lived in a neighborhood that they believed had
adequate access to shopping, shopped infrequently at fast food restaurants, shopped frequently at
farmers markets, were female, older and lived in a large household were also more likely to
consume fresh produce items. More frequent shopping at department stores for food items was
associated with more frequent meal preparation with fresh produce items.
61

The results and residuals of the model were examined in order to ensure none of the
assumptions of the model were violated. Based on the analysis of a plot and histogram of the
residuals there did not appear to be heteroskedasticity present in the residuals of the model. The
tolerances of the variables also indicated that there was not significant multicollinearity in the
model either. A map of the residuals of the model indicated that the model was not predicting
some areas significantly better than others and the residuals appeared to be evenly distributed.
Figure 4.4: Number of times meals are prepared with fresh produce items each week:
Model residuals

62

4.4.4 Overall analysis of general population backwards stepwise regression models
Backwards stepwise regression analysis using the variables that were generated for this
research was unable to create a regression model that explained a substantial amount of the
variance in the consumption of fresh produce items by Greater Lansing residents. For this
analysis none of the accessibility score variables were entered into any of the models, which is
not surprising based on the Pearson correlation values that were generated. The independent
variables that appeared in at least two of the backward stepwise regression models that were
created included: how much of a problem respondents felt that access was in their neighborhood,
how nutritious respondents felt that their diet was, how often respondents shopped at farmers
markets, gender, the number of people in a household, and whether or not a resident received
nutritional assistance. Unfortunately, many of the variables that had high explanatory power
within the established regression models will not prove to be incredibly useful when creating
improved models of the Lansing food environment. While it is certainly interesting that how
nutritious a person believes their diet is, how much access people believe they have to shopping
in their neighborhoods, and how frequently a person shops at farmers markets are statistically
significant predictors of some of the variance in the consumption of fresh produce items, this
information will not be very helpful for developing predictive models of the food environment
because it is not available in large scale public datasets such as the Census and ACS.
Gender and the size of household could provide some important information for
informing future models of the food environment. The most important variable identified by
these regression models was that nutritional assistance, which would predict individual
consumption of fresh produce items. Previous literature has identified that people who receive
nutrition assistance are at risk for a consuming an unhealthy diet (Gunderson and Oliveira 2001;
63

Rose and Richards 2004; Algert et al. 2006) and the results that were obtained from the
backward stepwise regression models appear to support this finding. As a result a deeper
analysis of how Greater Lansing residents who receive nutritional assistance interact with their
food environment was conducted to fully understand the environmental factors that may
influence the dietary behavior of this subpopulation. The results of this analysis will be displayed
later in this chapter.
4.5 Geographically weighted regression
GWR enables the further examination of regression models to understand and identify
other possible variables that might be missing from the model as well as to identify spatial
variations that might be present but difficult to enter into the model as an independent variable
(Brunsdon et al. 1998). As a result, GWR was employed using variables established using
ordinary least squares backwards stepwise regression to help further explain and understand the
geographic variation in the consumption of fresh produce by Greater Lansing residents, as well
as test the validity of the regression models that were established. In order to perform GWR on
the survey data obtained for this thesis, the data were spatially aggregated into census block
groups, a necessary step in order to use GWR analysis on the data. The responses of all survey
participants who lived in a census block group were averaged in order to create an overall mean
response for the block. While aggregating the data introduces a plethora of different problems to
the analysis including boundary issues and the modifiable areal unit problem, the aggregation of
data at this stage was necessary in order to conduct GWR analysis and further explore the spatial
variation that might be present in the consumption of fresh produce by Greater Lansing residents.
An adaptive kernel with varying bandwidths was used to set the number of neighbors used in
analysis as the study region was irregular and it was necessary for all observations to have near
64

neighbors to aid with the estimation of the model. The number of neighbors that were used for
the GWR was set to six.
4.5.1 Model #1: Number of different produce items (listed) purchased each week
The regression model established using backwards stepwise regression analysis that used
the number of different produce items (listed) purchased each week as the dependent variable
2

had very little explanatory power (R = 0.181) as an OLS regression model. The four
independent variables that were used in this model were; how nutritious a respondent believed
that their diet was, gender, number of household members, and whether or not a household
received nutritional assistance. When that same regression model was employed using GWR,
which enables a separate estimation of the parameters of the model at each observation, the
2

overall explanatory power of the model remained the same (R = 0.181). It is apparent that the
data collected for this research does not have the ability to explain the variation in the listed
number of different produce items purchased each week by respondents.
There is a clear indication based on the results of the GWR that the regression model
that was originally established using backwards stepwise regression analysis for the listed
number of produce items purchased each week has significant geographic bias in its
explanatory power (Figure 4.5). There is a clear visible trend in the explanatory power of
the model, which was able to predict up to 25% of the variance in the listed number of fresh
produce items purchased each week in the western portions of the study area and was only
able to predict 16% of the variance in the eastern part of the study area. This finding
indicates that there might be some spatial variation that is taking place in the interaction
between Greater Lansing residents and their food environment, which was not captured in
the data for this research. An examination of the spatial variation in the regression
65

coefficients for the four independent variables further indicated that there clearly was a
spatial dimension to the listed number of fresh produce items purchased each week.
2

Figure 4.5: Number of produce items purchased each week (listed) GWR R values
(predictive power of number of produce items purchased)

4.5.2 Model #2: Number of different produce items (estimated) purchased each week
The regression model established using backwards stepwise regression analysis with the
estimated number of different produce items purchased each week as the dependent variable also
66

2

had very little explanatory power (R = 0.198). The five independent variables that were used in
this model were; whether or not respondents viewed neighborhood access as a problem, how
nutritious respondents believed that their diet was, frequency of visits to farmers market, income,
and whether or not respondents received nutritional assistance. When that same regression
model was employed using GWR the overall explanatory power of the model increased only
2

slightly (R = 0.223).
Again, there is a clear indication that the regression model established to explain the
estimated number of produce items purchased each week has significant geographic bias in its
explanatory power (Figure 4.6). In comparison to the previous model analyzed with GWR the
geographic bias in the predictive power of the model appears to be significantly stronger. While
there does continue to be an east-west bias in the predictive power of the model there also
appears to be significant variation between the southern and northern portions of the study area.
The GWR model was able to predict up to almost 27% of the variation in the southwestern
census blocks of Lansing, while the model was only able to predict 13% of the variation in
eastern Lansing. The results obtained from this GWR analysis continue to indicate that there
might be some spatial variation that is taking place in the interaction between Greater Lansing
residents and their food environment and that there are explanatory variables that are missing
from the data obtained for this research. An examination of the spatial variation in the regression
coefficients continued to indicate that there was a spatial dimension to the estimated number of
fresh produce items purchased each week, specifically between the southwestern and eastern
Greater Lansing area. The strength of the regression coefficients also appeared to be associated
with the trends that were visible in the geographic predictive power of the model.

67

2

Figure 4.6: Number of produce items purchased each week (estimated) GWR R values
(predictive power of number of produce items purchased)

4.5.3 Model #3: Number of times meals are prepared with fresh produce items in a week
The regression model with the number times a week that meals are prepared with fresh
2

produce items as the dependent variable had the most explanatory power (R = 0.345) of all of

68

the regression models that were established for this research using backwards stepwise
regression. The eight independent variables that were used in this model were; whether or not
Figure 4.7: Number of times meals are prepared with fresh produce items in a week
2
GWR R values (predictive power of number of times meals are prepared with fresh
produce items in a week)

69

respondents viewed their neighborhood access as a problem, how nutritious respondents believed
that their diet was, frequency of visits to fast food restaurants, frequency of visits to farmers
market, frequency of visits to department stores, gender, age, and household size. When that
same regression model was employed using GWR the overall explanatory power of the model
2

actually decreased (R = 0.234), which indicates the OLS regression model may be over
predicting the relationship potentially as a result of the modifiable areal unit problem. The trend
of significant variation in the predictive power between western (33%) and eastern (20%)
portions of the study area continued with this regression model (Figure 4.7).
4.5.4 GWR analysis
Based on the analysis of all three regression models using GWR there is a clear
indication that there is some spatially related phenomenon occurring in the regression models.
An analysis of the spatial distribution of the regression coefficients also continues to establish a
clear difference in what influences the consumption of fresh produce products by residents of the
western and eastern portions of the study area. One variable that could potentially be missing
from the data obtained for this research and may be necessary in future analysis of how Greater
Lansing residents interact with their food environment could potentially be educational
attainment, which was not included in this research. This finding may also indicate that the
relationship between the independent variables and the consumption and use of fresh produce
items is not linear. As a result, regression modeling techniques may not be appropriate to
examine the relationship between the behavioral choices of Greater Lansing residents and the use
of fresh produce items in their diets.

70

4.6 Analysis of population who receive nutritional assistance
The results of this research have indicated that receiving nutritional assistance is
associated with a decreased consumption of fresh produce products. Previous research has
indicated that environmental factors have an amplified influence on the dietary choices of
participants in nutritional assistance programs (Rose and Richards 2004) and that this
subpopulation has been previously identified in the literature as being less likely to have
sufficient access to fresh produce items (Algert et al. 2006). Since the initial review of the
dataset indicated that those who receive nutritional assistance in Lansing may interact with the
food environment differently than the general population in Lansing who do not receive
nutritional assistance, an in-depth analysis of this subpopulation and their interaction with the
food environment was performed.
Out of the entire sample for this research 37 (roughly one-fifth of) respondents indicated
that they were currently receiving some form of nutritional assistance. The responses from this
subgroup indicate that the behavior of those who receive nutritional assistance varies from those
who do not receive nutritional assistance, specifically in a number of ways that may impact this
subpopulations ability to achieve a healthy and balanced diet. Only 62% of those who received
nutritional assistance had access to an automobile in comparison with 94% of those who did not
receive nutritional assistance.
In terms of shopping behavior, the group of respondents that received nutritional
assistance was more likely to visit a store as a unique trip (57%) in comparison to those who did
not receive nutritional assistance (22%). This subgroup was also more likely to be surveyed
shopping at a store that was located in their residential neighborhood (76%). The mean travel
time from the store at which they were surveyed to their home for those who received nutritional
71

assistance was 7.3 minutes, which was shorter than the population who did not receive
nutritional assistance whose average travel time to a store was 11.4 minutes. However, those
who received nutritional assistance did not appear to believe that access to adequate shopping
was a problem in their neighborhood, with 70% indicating that access to adequate shopping was
only a ―minor problem‖ or was ―not really a problem‖. Despite indicating that access was not a
problem in their neighborhood, most of the respondents who receive nutritional assistance (59%)
did not report shopping for the bulk of their food items within a mile of their home. While those
who receive nutritional assistance tend to believe that their food environment is satisfactory they
do not appear to do the majority of their food shopping in the neighborhood surrounding their
primary residence.
The nutritional assistance subpopulation was also less confident than those who do not
receive nutritional assistance that the diet they were consuming was healthy. While the
population of respondents who received nutritional assistance felt that their diet was healthy
(68%) less respondents reported that they were careful about what they ate (54%) and that they
had a nutritious diet (51%) then the population who did not receive nutritional assistance (Table
4.9). Most respondents who received nutritional assistance would prefer to prepare meals with
fresh fruits and vegetables (73%). However, more than a quarter of respondents indicated that
they would rather prepare meals with frozen (16%) and canned (11%) fruit and vegetable items,
which was greater than respondents who did not receive nutritional assistance (5% and 6%
respectively). Less than one-third of respondents who receive nutritional assistance reported
purchasing organic fruit and vegetable items.

72

Table 4.9: Dietary perceptions of those who receive nutritional assistance in comparison to
those who do not
Neither
agree nor Disagree
disagree

Strongly
agree

―I am a healthy eater‖

―I am someone who
eats in a nutritious
manner‖
―I am someone who is
careful about what I
eat‖

Nutritional
assistance
No nutritional
assistance
Nutritional
assistance
No nutritional
assistance
Nutritional
assistance
No nutritional
assistance

Agree

Strongly
Disagree

14%

54%

11%

16%

5%

24%

56%

9%

11%

1%

22%

30%

19%

27%

3%

22%

56%

8%

12%

1%

24%

30%

19%

24%

3%

24%

53%

13%

9%

1%

4.6.1 Correlation analysis
Correlation analysis was performed in order to understand whether or not the relationship
between the consumption of fresh produce items and previously established models of the
Lansing food environment (Goldsberry et al. 2010) was stronger for those who receive
nutritional assistance than those who do not receive nutritional assistance. However based on the
results obtained from correlation analysis (Table 4.10) there also does not appear to be a
substantial association between calculated access to fresh produce items and the use of fresh
fruits and vegetables in the diets of those who receive nutritional assistance. While the
relationship between calculated access and the use of fresh produce in the diets is stronger for
those who receive nutritional assistance in comparison to those who do not receive nutritional
assistance, there is still not a large amount of predictive power. There appears to be only a slight
indication that the built environment as represented in the models established by Goldsberry et
73

al. (2010) has more of an influence on those who receive nutritional assistance than those who do
not based on the analysis of the Pearson correlation results.
Table 4.10: Pearson correlation vales between calculated access and dietary behavior for
those who receive nutritional assistance and those who do not

Number of Meals
Prepared with Fresh
Produce each Week

Number of Fresh
Produce Items
(estimated) Purchased
each Week
No
Nutritional
nutritional
assistance
assistance

Nutritional
assistance
Drive
(weighted)
Drive (count)
Pedestrian
(weighted)
Pedestrian
(count)
Cumulative
Distance
Method

No
nutritional
assistance

0.251

-0.038

0.231

0.105

-0.027

0.142

Number of Fresh
Produce Items (listed)
Purchased each Week
Nutritional
assistance

No
nutritional
assistance

-0.023

0.261

-0.060

0.136

0.014

0.244

-0.049

0.000

0.107

-0.028

0.023

0.039

0.067

0.069

0.141

0.010

0.008

0.066

-0.171

0.008

-0.121

0.007

-0.045

0.008

4.6.2 Difference in travel time
In order to understand if the observed differences between the population of those who
received nutritional assistance and those who do not receive nutritional assistance were
statistically significant analysis of the dataset using two-sample t-tests was performed. The first
variable that was examined using a two-sample t-test was the amount of time that it took for the
respondent to travel to the retail store where they were surveyed from their primary residence.
The responses for this particular variable by survey participants were examined to ensure that the
data did not violate the assumptions of the model. After it was determined that this variable
could be analyzed using a two-sample t-test hypotheses were generated:
74

H0: There is no significant difference in the mean travel time from retail store to
home for the population who receive nutritional assistance and the population
who do not.
HA: There is a significant difference in the mean travel time from retail store to
home for the population who receive nutritional assistance and the population
who do not.
Based on the fact that the variance in this variable for those that received nutritional assistance
and those that did not was not equal a separate variance model two-sample t-test was used.
Table 4.11: Mean and standard deviation for perceived travel time from home to retail
location

7.270

Standard
deviation
7.974

11.399

9.204

Mean
Nutritional assistance
No nutritional assistance

The results of the two-sample t-test indicated that the difference in the mean perceived
travel time from home to the retail store surveyed at for those who receive nutritional assistance
and those who do not was significant. The t-value was large enough (-2.716) to reject the null
hypothesis at a 99% confidence level (p-value=0.009). This finding indicates that the population
of participants who receive nutritional assistance interact differently with their food environment
than participants who do not receive nutritional assistance. Respondents who received
nutritional assistance were significantly more likely to shop at food stores that were located
closer to their home and only traveled 7.3 minutes to the store where they were surveyed in
comparison to the rest of the population who traveled 11.4 minutes.

75

4.6.3 Difference in food retail store frequency
The ten variables (Table 4.12) which measured the frequency that Greater Lansing
residents shopped at different categories of food retailers to purchase food items were examined
to see if there were significant differences between the population of respondents that receive
nutritional assistance and the population that does not. The responses for all ten variables were
examined to ensure that the data did not violate the assumptions of the two sample t-test model.
After it was determined that these variables could be analyzed using a two-sample t-test
hypotheses were generated:
H0: There is no significant difference in retail store shopping frequency for the
population who receive nutritional assistance and the population who do not.
HA: There is a significant difference in retail store shopping frequency for the
population who receive nutritional assistance and the population who do not.
Finally each variable was individually examined to determine whether or not pooled or separate
variance models should be used.
The results of the two-sample t-tests indicated that there were some differences in
shopping behavior between the population of respondents who receive nutritional assistance and
the population that does not. Although for the majority of food retail store categories the null
hypothesis was accepted and there were no statistically significant differences in shopping
behavior, there were some significant differences in shopping behaviors between the two
separate populations (Table 4.12). The surveyed population that received nutritional assistance
were more likely to shop at convenience stores and liquor stores for food items at a 99%
confidence level. The respondents that did not receive nutritional assistance were more likely to
purchase food items at restaurants (95% confidence level) and farmers markets (99% confidence
76

level). Based on these findings it continues to appear that there are significant differences in
how the population of Greater Lansing residents who receive nutritional assistance interact with
their food environment in comparison to the population who does not receive nutritional
assistance.
Table 4.12: Food retail store frequency two-sample t-test results
Nutritional
assistance
mean

No nutritional
assistance
mean

Convenience store

9.703

2.764

2.833

.007

Farmers market

0.486

1.806

-4.028

.000

Liquor store

5.405

0.507

3.194

.003

Restaurant

2.757

4.875

-2.319

.021

Department store

1.108

0.708

1.170

.245

Failed to reject

Discounter

0.622

0.694

-0.231

.817

Failed to reject

Fast food

4.649

3.417

1.222

.227

Failed to reject

Grocery store

2.919

3.875

-1.404

.163

Failed to reject

Health store

1.405

1.896

-0.446

.657

Failed to reject

Supermarket

4.730

4.681

0.065

.948

Failed to reject

Variable

t-value p-value

H0
Rejected at a 99%
confidence level
Rejected at a 99%
confidence level
Rejected at a 99%
confidence level
Rejected at a 95%
confidence level

4.6.4 Difference in the consumption of fresh produce
Finally the three variables that measured consumption of fresh produce products were
examined to determine if there were statistically significant differences in dietary behaviors
between the respondents who received nutritional assistance and those that did not. The
responses for these three variables were examined to ensure that the data did not violate the

77

assumptions of the two-sample t-test. After it was determined that these variables could be
analyzed using a two-sample t-test hypotheses were generated:
H0: There is no significant difference in fresh produce consumption between the
population who receive nutritional assistance and the population who do not.
HA: There is a significant difference in fresh produce consumption between the
population who receive nutritional assistance and the population who do not.
The variance of all three variables was examined to determine whether or not pooled or separate
variance models should be used.
Based on the results of the two-sample t-test it is clear that there are differences in the use
of fresh produce in diets between the respondents that receive nutritional assistance and the
respondents that do not (Table 4.13). For two of the three fresh produce consumption variables
the difference in means between the population of respondents who receive nutritional assistance
and the population of respondents who do not was statistically significant. The two variables
that were statistically significant were measuring the diversity of different fresh produce items
purchased. For both the estimated and listed number of fresh produce items purchased in a week
the population of respondents who received nutritional purchased fewer different items then the
population who did not receive nutritional assistance. These results indicate that the population
who are not receiving nutritional assistance are consuming a more diverse variety of fruits and
vegetables then those who do receive nutritional assistance. According to the 2010 Dietary
Guidelines for Americans one of the key elements of achieving a healthy and balanced diet is
consuming a diet with a diverse variety of different fruit and vegetable items. Based on this
result it is apparent that the population of respondents who are receiving nutritional assistance

78

are not consuming as diverse a diet as those who do not receive nutritional assistance, which may
have an impact on their overall health.
Table 4.13: Consumption of fresh produce two-sample t-test results

Variable
Number of times a month
respondent prepares meals
with fresh produce
Number of fresh produce
items purchased in a week
(estimated)
Number of fresh produce
items purchased in a week
(listed)

No
Nutritional
nutritional
assistance
assistance
mean
mean
21.162

23.097

t-value

p-value

H0

1.222

.227

Failed to reject

4.514

8.347

-2.319

.021

4.108

6.507

2.833

.007

79

Rejected at a
95% confidence
level
Rejected at a
99% confidence
level

CHAPTER 5: DISCUSSION

5.1 Research question #1: Association between calculated access to and the use of fruits and
vegetables in the diets of Greater Lansing residents
A person‘s decision on what to eat is incredibly personal and often unique to the
individual. Research has indicated that there is not one variable that can completely explain how
an individual makes dietary decisions and there are a plethora of different factors that can
potentially influence an individual‘s dietary decisions (Lytle 2009; Sobal and Bisogni 2009). As
a result, it may be incredibly difficult to isolate whether or not physical access to food retailers
impacts the dietary choices of residents (White 2007). The results of this research indicate that
there is not a straightforward relationship between calculated access to fresh produce and the use
of fresh fruits and vegetables in the diets of Greater Lansing residents, as the survey data
indicated that there was no correlation between calculated access to fresh produce and the dietary
behavior of Greater Lansing residents.
There may be a number of different reasons why no association was found between
calculated access to fresh produce and the use of fresh produce in the diets of Greater Lansing
residents. First it is possible that the accessibility models established by Goldsberry et al. (2010),
that were used as representations of the food environment for this research, do not capture the
entire interaction taking place between Greater Lansing residents and their food environment.
Results from this survey indicate that access through the available transportation network may
only be a small part of how an individual interacts with the food environment and other
important variables that influence dietary behaviors are not included in these models. The results
of the GWR analysis also indicated that the relationship between the variables obtained from the
80

survey instrument and the consumption of fresh produce items by Greater Lansing residents is
not linear and more complex models should be used to understand the relationship.
5.2 Research question #2: Variables that influence dietary behavior
This research found that it was very difficult to identify individual variables that had a
large amount of explanatory power over the consumption of fresh fruits and vegetables, which is
supported by previous research that has suggested that a variety of different factors may
influence dietary behavior (Ford and Dzewaltowski 2008). The food environment as represented
by Goldsberry et al. (2010) does not appear to have any explanatory power over the dietary
related behavior of Greater Lansing residents. When all of the variables that were obtained from
this research were entered into a backwards stepwise regression model there were only a few
variables that had an association with dietary behavior; how much of a problem Greater Lansing
residents felt that access was in their neighborhood, how nutritious Greater Lansing residents felt
that their diet was, how often a respondent shopped at a farmers market, the gender of a
respondent, the number of people in a household, and whether or not a respondent received
nutritional assistance. Unfortunately, many of the variables that had high explanatory power
within the established regression models are not incredibly helpful when establishing improved
models of the food environment as they are not readily available in public large scale datasets,
such as the Census and ACS. Further research must be done to better understand the dietary
behaviors of Greater Lansing residents and how that influences their interaction with the food
environment, as the variables from the survey instrument for this research were not able to
predict the dietary behavior of Greater Lansing residents.
Another key finding from this thesis is that not all food retail stores have an equal impact
on the Lansing food environment. While organic and health food stores often provide a diverse
81

variety of produce items to the surrounding community (Duvall et al. 2010) these stores are not
frequently visited by the general population of Lansing, with only 15% of respondents reporting
that they shop at this category of stores at least once a week. In comparison supermarkets, which
also provide a diverse variety of fresh fruit and vegetable items, were visited at least once a week
by 63% of the survey population. This finding is incredibly important as this information can be
used to inform and create more representative models of the Lansing food environment. The
results of this research clearly indicate that supermarkets are the primary location where Greater
Lansing residents obtain food items. The information regarding the frequency of visits to certain
retail food stores will enable researchers to give retail store weights based on their overall impact
on the Lansing food environment. Updating models of the Lansing food environment with this
type of information could potentially facilitate a more informed model of how Greater Lansing
residents are actually impacted by their food environment.
While some research has suggested that the popularity of CSAs, farmers markets, and
other direct-selling systems has increased in recent years resulting in an increased consumption
of fresh fruits and vegetables among the general population (Guptill and Wilkins 2002), this
research has found that may not necessarily be the case. There is some indication in the data that
was obtained for this research that the frequency one shops at a farmers market may be
associated with a greater consumption of fresh produce items. Currently, however, only 52% of
the Lansing population report shopping at farmers markets and only 22% of the population does
so at least once a week. Increasing the attractiveness of farmers markets with populations who
are at risk for consuming an unhealthy diet may be an important step towards combating obesity
in Lansing and more research must be performed to further understand this phenomenon. While
alternative food environments are certainly an essential part of the Lansing food environment the
82

findings of this research indicate that at the moment they are only a small part of how Greater
Lansing residents obtain food items on a large scale.
5.3 Research question #3: How Greater Lansing residents view their food environment
Greater Lansing residents do not appear to believe that the food environment in Lansing
is a significant problem. A majority of Greater Lansing residents (82%) did not feel that access
to shopping was a problem in their neighborhood, which may indicate that the food environment
in Lansing may exist in part as a reflection of consumer demand in the area. This finding is
supported by previous research which has suggested that food retail outlets do not sell items that
will not be purchased (Baker et al. 2006a). In fact, despite indicating that access to adequate
shopping was not a problem in their neighborhood most Greater Lansing residents (57%) did not
report shopping for their food items at a food retail store within a mile of their home. The lack
of demand for fresh produce products in these neighborhoods may be reflected in the lack of
stores selling fresh fruits and vegetables in the area. Solutions to the nutritional problems
plaguing Greater Lansing residents may need to include approaches to incentivize stores in
neighborhoods at risk for obesity to sell fresh produce items, as well as to provide educational
programs within those same neighborhoods to help increase knowledge of the necessity to
consume fresh produce items to increase the demand for these items.
5.4 Research question #4: Difference between Lansing residents who receive nutritional
assistance and those who do not
The population of Greater Lansing residents who receive nutritional assistance has
statistically significant differences in food environment then the general population the observed
interactions between residents and the environment are statistically significant. The association
between calculated accessibility and dietary behavior becomes stronger and there does appear to
83

be some level of environmental influence on dietary behavior for this subpopulation. GWR
analysis also indicated that there was a strong spatial dimension in the predictability of the
consumption of fresh produce items by Greater Lansing residents. The results of this research
indicate that not all groups of the population may use the food environment the same. The
behavioral information from Greater Lansing residents that was obtained for this research may
help provide insight into how Greater Lansing residents interact with their food environment in
order to enable researchers to create more accurate models of the Lansing food environment in
the future.
According to the 2010 Dietary Guideli0nes for Americans one of the essential parts of
achieving a healthy diet is consuming a diverse variety of fresh fruits and vegetables. Greater
Lansing residents reported that they purchased more than seven different fresh fruit and
vegetable products in an average week and that fresh was by far the most common way that they
purchased fruit and vegetable items. This research has found, however, that those who receive
nutritional assistance are not purchasing as diverse of an array of fresh produce products, only
four different items a week, which is statistically significantly different from the general
population. Although, this finding is in no way is an absolute measure of dietary practices it
does indicate that Greater Lansing residents who do receive nutritional assistance are consuming
a less diverse diet than those who do not. Jago et al. (2007) found that long term exposure to
fresh produce items may in fact influence dietary practices. If the Greater Lansing residents who
receive nutritional assistance are not purchasing a diverse array of fresh produce products then it
is possible that their children may grow up without the knowledge to purchase fresh produce
products on their own. Therefore it is of the upmost importance to focus on this subpopulation
when analyzing the food environment in Lansing.
84

Previous research has found that shopping behavior can be variable depending on the
individual and every person interacts with their food environment differently (Whelan et al.
2002; Yoo et al. 2006). These findings by previous researchers are supported by this thesis,
which found that the population of Greater Lansing residents who receive nutritional assistance
interact with the food environment differently than those who do not receive nutritional
assistance. The population of Greater Lansing residents who receive nutritional assistance are
significantly more likely to shop for food items at convenience stores and at liquor stores.
Convenience and liquor stores are more likely to sell products that are energy rich (i.e.
carbohydrates, sugars, breads, pasta) and often less expensive than fresh produce items, which
makes them potentially more desirable to individuals on a limited income (Cummins and
Macintyre 2002; Morland et al. 2006). This particular research finding is worrisome as a diet
that is high in energy rich food items has been associated with obesity (Liese 2009). When
establishing representations to help understand the food environment in Lansing it is important to
include behavioral information on how subpopulations interact with the food environment
differently in order to fully understand how local residents are interacting with their food
environment.
One of the primary assumptions of the Lansing food environment models developed by
Goldsberry et al. (2010) is that Greater Lansing residents are willing to travel 10 minutes to
reach a food retail destination within the Lansing food environment. This assumption was
validated by the findings of this thesis. The average travel time that it took survey participants to
reach the store location where they were surveyed was 10.4 minutes, which confirms the use of a
10 minute threshold. However, an additional important finding was that respondents who
received nutritional assistance traveled on average only 7.3 minutes to the store at which they
85

were surveyed in comparison to those that do not receive nutritional assistance who traveled 11.4
minutes, which was a statistically significant difference. This difference in travel time between
the two populations could be indicative of a larger issue with generating models of the food
environment. Greater Lansing residents may interact with the food environment in very different
ways. The findings of this research indicate that the population of Greater Lansing residents who
receive nutritional assistance have a much smaller food environment accessible to them than
those who do not. Therefore, the 10 minute threshold used by Goldsberry et al. (2010) might
actually be overestimating the number of fresh produce items available to Greater Lansing
residents who receive nutritional assistance.
The finding that the population of Greater Lansing residents who receive nutritional
assistance interact with the food environment differently is additionally supported by the fact that
the Greater Lansing residents who receive nutritional assistance were more likely to shop at retail
locations in their residential neighborhoods (76%) than the general population (58%). This
important difference in shopping behavior may be a result of the fact that those who receive
nutritional assistance are less likely to own automobiles, with only 62% of those surveyed who
received nutritional assistance having regular access to an automobile. While research has
shown that local shopping options do not necessarily decrease the reliance of neighborhood
residents on their automobile (Handy and Clifton 2001) the lack of transportation options
available to Lansing resident who receive nutritional assistance may have an impact on the food
environment available to them.
5.5 Importance of examining the food environment as a whole
The findings of this thesis highlight the importance of examining the food environment as
a whole and not only focus on the food environment near where a person resides. Food choice
86

decisions are often situational and are made based on the environment in which a person is in at
that particular point in time (Sobal and Bisogni 2009). The interaction between Greater Lansing
residents and their food environment appears to support this finding. Not all Greater Lansing
residents shop for food items in neighborhoods that are close to their primary residence, with
42% of survey participants indicating that the retail store where they were shopping at when
surveyed was closest to a location other than their primary residence.
5.6 Limitations of research
While the findings of this thesis have established a new understanding of how Greater
Lansing residents interact with their food environment this thesis does not suggest that the same
data is directly comparable to regions outside of Lansing. Each urban area has its own unique set
of characteristics and problems, and the analysis presented here is merely a discussion of the
phenomena occurring within the food environment of Lansing. The data for this research was
collected during two summer months, June and July and as a result the analysis provided in this
thesis may be slightly biased as the data was collected during a season when produce items were
readily available. The proportion of Greater Lansing residents that obtain fresh produce items
from outside the food retail environment at places such as gardens or local farmers markets may
be significantly higher during the summer months. In order to help mitigate the impact of the
increased availability of fresh produce items an effort was made to only survey places that were
available to Greater Lansing residents year round. However, there is no way to be certain that
the findings that were obtained from survey results obtained during summer months would be the
same as findings from a survey that took place over winter months.
The survey methodology used for this research may have also introduced some bias into
the results. This study is based on a sample of 185 residents of Lansing, Michigan, which is only
87

a small portion of the 291,000 estimated people who reside in the study area. An effort was
made to ensure that the sample of Greater Lansing residents obtained for this research was as
representative of the overall population as possible by constantly reviewing the demographics of
the sample population to ensure that they approximately matched the census distribution of the
Lansing population. In addition, according to Goldsberry et al. (2010) there are 94 retailers of
fresh produce products in the Lansing area, however, only 12 retailers were surveyed for the
purpose of this research. The list of 12 retailers surveyed also does not include and Meijer or
Quality Dairy stores, which are both significant parts of the Lansing food environment. As a
result, the participants surveyed in this research are patrons of only a small portion of Lansing
retail establishments. This may in turn bias my results as the entire Lansing food environment
was not surveyed. Finally, while there are many positive benefits to conducting in-person
research, the presence of the researcher during the interview can cause the respondent to give
answers that they believe to be more socially acceptable (Tyebjee 1979). This could cause some
potential problems for this research as there may be a social stigma associated with eating poorly
and obesity. This may cause participants to answer questions how they believe that they should
answer and not provide responses that are representative of their actual behaviors.
There are a number of different factors which may influence an individual‘s dietary
choices including culture, social capital, and price that were not included in the survey and data
analysis for this research (Ball et al. 2006; Ford and Dzewaltowski 2008; Sobal and Bisogni
2009). While information such as education, more in depth daily dietary choices, the actual
items purchased during the shopping trip, and year round shopping habits may have been useful
in this analysis these variables were not included in the research because of the scope of the
project and time limitations on the survey questionnaire. Previous research has also found that
88

social support is significantly correlated with the purchase of fruits and vegetables (Baranowski
et al. 2006) and there was no measure of this factor at all in my data analysis. In retrospect the
questions related to the consumption of fresh produce items could have been improved upon
dramatically. There were also many different ways that a person can consume fruits and
vegetables that were not measured in the survey designed for this research. The quality of fresh
fruits and vegetables available may also be a factor that influences the decision to purchase fresh
produce items. Stores located in urban areas, neighborhoods with low socioeconomic statuses
and where food is secondary have been found to offer lower quality fresh fruit and vegetable
items (Cummins et al. 2009; Koro et al. 2010). However, no measure of food quality was
included in this particular analysis of the Lansing food environment.
Finally the primary focus of this research study was on the use of fresh produce items in
the diets of Greater Lansing residents. While previous research has indicated that fresh produce
items are a good measure to judge the nutrition on an individual‘s overall diet (Morland et al.
2002a; Bertrand et al. 2008) there are many other components that make up a healthy and
balanced diet. Since the focus of this research was primarily on the consumption of fresh
produce items a large part of an individual‘s dietary behavior was missing from the analysis of
the overall Lansing food environment. Future research into the Lansing food environment
should include a comprehensive approach that examines all components of dietary behavior.
Unfortunately, the models of the Lansing food environment established by Duval et al.
(2010) and Goldsberry et al. (2010) that were used as the background measure of access for this
research did not include restaurants within their definition of the overall food environment. The
results obtained from this research clearly indicate that restaurants are a significant component of
where Greater Lansing residents obtain food items, with a significant number of residents
89

obtaining food items from fast-food restaurants (39%) and restaurants (27%) at least one time a
week. Information about the locations of fast-food restaurants and restaurants, however, is
completely missing from the model of the Lansing food environment that was used for this
research. This information will be essential to creating more comprehensive models of the
Lansing food environment that completely capture how Greater Lansing residents interact with
their food retail environment and is a major limiting factor in the models that were used of the
Lansing food environment for this research.
Finally, most of the quantitative analysis that was conducted for this thesis was done with
survey data that may have been subject to errors in data collection. This type of data is not
considered to be ideal for inferential statistical models as the survey was not administered under
perfect experimental conditions as is required by the assumptions of many of the models
employed in this research (Wrigley 1983). This fact, however, does not negate the information
that has obtained from the analysis that was conducted for this research. The findings presented
in this thesis offer many insights into the behaviors of Greater Lansing residents and are
important towards informing future studies of the Lansing food environment and more
importantly creating better representative models of the Lansing food environment.

90

CHAPTER 6: CONCLUSION

6.1 Modeling the Lansing food environment
The primary objective of this research was to understand the interaction between Greater
Lansing residents and their food environment. The findings of this thesis have clearly indicated
that there is not a simple and easily generalizable way to understand the interaction that Greater
Lansing residents have with their food environment. As a result establishing accurate and allencompassing models of the Lansing food environment will require further research and much
more sophisticated modeling techniques in order to fully capture the complete interaction. Lytle
(2009) found that much of the previous research conducted on food environments has been,
―…dominated by research using unsophisticated study designs and has frequently failed to see
the role of social and individual factors and how they interrelate with the physical environment‖
(p.S134). As research concerning food environments moves forward it is important to make sure
that social and individual decision making factors are included in representations of how people
interact with the built environment, a fact which is strongly supported by the findings of this
research. This research indicates that there is no relationship between calculated access to fresh
produce and the use of fresh fruits and vegetables in the diets of Greater Lansing residents.
Previous models of the Lansing food environment (Goldsberry et al. 2010) can be significantly
improved upon using information about the behavioral interactions Greater Lansing residents
have with their food environment gained through this research. Improving the representations of
the food environment in Lansing may afford researchers greater insight into the relationship
between the environment and dietary behavior.

91

While the models that were developed by Goldsberry et al. (2010) present an excellent
initial examination of the Lansing food environment, new models can be developed utilizing
some of the information gained from this research in order to further advance our understanding
of the Lansing food environment as a whole. In particular gravity based models or probabilistic
accessibility measures that introduce the variable of attractiveness of a particular location and the
probability that a specific location will be visited could be utilized to provide a more
comprehensive understanding of the Lansing food environment (Hansen 1959; Wang 2000).
The findings of this research have indicated that almost all Greater Lansing residents shop for
food items at supermarkets (94%), while they tend to do less of their shopping for food items at
other food retail providers. The use of a gravity based models or probabilistic accessibility
measures will allow researchers to give supermarkets a greater weight within the food
environment model to more accurately represent their footprint in the community. Stores such as
organic and health food stores that stock a large number of produce items but serve a smaller
client base could be weighted down to decrease their impact on the overall food environment.
Information on how far Greater Lansing residents are willing to travel to different categories of
stores could also be used to better represent how Greater Lansing residents actually interact with
the food environment.
One of the primary research questions of this thesis was; do Greater Lansing residents
who receive nutritional assistance interact with the food environment differently than those who
do not? This research has clearly indicated that there are differences in shopping behavior
between Greater Lansing residents who receive nutritional assistance and Greater Lansing
residents who do not. More specifically, the Greater Lansing residents who received nutritional
assistance were more likely to shop at convenience and liquor stores for food items and were less
92

likely to purchase a diverse array of fresh produce products, believe they had a nutritious diet, or
travel to stores that were not located near their home. These findings indicate that it is important
to focus attention on the subpopulation who receive nutritional assistance and to make sure that
models of the food environment are capturing how they interact with their surrounding
environment, as well as the interaction of the general population with their food environment.
Often those who receive nutritional assistance are constrained by the modes of transportation
available to them and their socioeconomic status. While the food environment may be accessible
for the majority of Greater Lansing residents, it is essential that the constraints and impacts that
the food environment has on those who do receive nutritional assistance are further examined.
Models such as relative accessibility measures (Golledge 1993) could be used to understand the
different impact that the food environment has for those who receive nutritional assistance and
those who do not. A relative accessibility measure would enable researchers to understand the
relative access to a food retail store from a given location for a person who receives nutritional
assistance versus a person who does not (Church and Marston 2003). This type of model would
provide more detailed information to researchers on the complex differences between how
Greater Lansing residents interact with their food environment.
Finally more complex agent based models could be utilized to better incorporate the
different behavioral interactions that have been measured in this research into updated models of
access in the Lansing food environment. Another one of the primary research questions
proposed by this thesis was: are there variables other than the food retail environment that
influence dietary behavior? Regression models indicated that few of the variables from the
survey instrument had significant explanatory power. This finding was further confirmed
through GWR analysis, which further indicated that this may be because the relationship
93

between dietary behavior and the variables presented might not be spatially constant. An
approach using agent based modeling allows for the simulated interactions of different agents
within the environment and enables the researcher to measure the impact of an agents decisions
on the system as a whole. This methodological approach could enable food environment
researchers to understand what the implications might be on the overall food environment if a
particular store were to close or if a new store were to be added into a community that was
determined to be at risk. By using this type of model a focus could be placed on the different
market basket of goods an agent who receives nutritional assistance might purchase in a food
environment in comparison to an agent who does not. This form of modeling could enable the
researchers to easily incorporate more actual behaviorally based information into food
environment research and potentially create more accurate representations of the food
environment.
6.2 Policy implications
Until relatively recently food has been left out of the planning of American cities for the
most part (Pothukuchi and Kaufman 1999) and because economic and political processes have
favored the development of suburban neighborhoods many residents of the urban core have been
left behind (Morland et al. 2006). Research has indicated that ―…the public health significance
of the local food environment on overweight and obesity is not a function of any independent
effect of the environment on individuals; rather, it is a function of its casual role in facilitating or
constraining individual choices and opportunities‖ (Morland et al. 2006, p.338). While having
achievable access to a supermarket is an important part of providing those who are nutritionally
at risk with access to fresh fruits and vegetables (Lavin 2005) what is more important is
identifying who the populations are nutritionally at risk are and understanding how their local
94

food environment is impacting their dietary behaviors. Achieving a greater understanding of
actual behavioral information is an essential step towards informing public policy that will effect
change in dietary behavior.
When discussing food environments it is important to remember that the food retail
industry is incredibly large, profitable, business and food retail outlets do not sell items that will
not be purchased (Baker et al. 2006a). In addition to education programs that help inform
Greater Lansing residents about how to eat a healthy and balanced diet some research has
suggested that there may be the additional need to use policy approaches to incentivize stores in
neighborhoods at risk for obesity to sell fresh produce items (Krebs-Smith et al. 2010).
However, before people start gravitating towards these sorts of policy solutions a much greater
understanding of how residents interact with their food environment must be achieved. Research
has only begun to identify how residents utilize and interact with their food environment and it
would be premature to assume that the development of a new grocery store would alter the
dietary habits of residents who reside in the surrounding community. It is incredibly easy for
researchers to fall in a ―local trap‖ and fail to understand how the entire food environment
functions as a whole (Born and Purcell 2006). The findings of this thesis indicate that most
Greater Lansing residents are satisfied with their current food environment. Just because a food
store is not located at a near distance to where a person lives does not mean that food store is
inaccessible to that person, especially as a result of the fact that people function in a variety of
different environments over the course of the day. This research has demonstrated that in
Lansing consumers often loop together their shopping trips and sometimes shop on the way
home from work. Focusing only on an individual‘s residential context does not really give a
complete picture of how residents are interacting with the food environment and current food
95

environment research does not contain enough information about the variety of different contexts
that urban residents function in a given day to accurately inform large scale policy solutions.
6.3 Future research
Food environments are incredibly complex and therefore require a diverse research
approach to help find solutions to the current dietary problems in America (Bertrand et al. 2008).
This research attempts to help come to a better understanding of the Lansing food environment
and how local residents interact with it in order to inform the development of better geographic
models of the food environment in the future. The results of this research clearly indicate that
the food environment is complex and is not necessarily easy to create accurate representations of.
Food environments and dietary behavior are constantly associated with a variety of different
factors, including behavioral preferences, socioeconomic status, culture, availability of food
items, cost of food items and quality of food items. As a result when researchers attempt to
model the food environment an equally complex model that takes into account all of these
different factors should be used. Current geographic models of the food environment are very
simple and have only used distance and the transportation network as limiting factors in where
urban residents are able to purchase food items. Updated models of the food environment should
be established using more advanced modeling techniques, such as agent based modeling, to fully
capture the complex interaction that takes place between Greater Lansing residents and their
food environment. Further research should look to examine and incorporate behavioral
information into their analysis of local food environments before jumping to broad conclusions
about the impact of the food environment on dietary behavior. Developing a comprehensive
understanding of how Greater Lansing residents interact with their food environment will be

96

essential in understanding how to find effective solutions to the obesity and obesity related health
issues.

97

APPENDICES

APPENDIX A: SURVEY INSTRUMENT
Survey # __________
SECTION 1: INTERVIEWER NOTES
Location of survey:
Store name: __________________________________
Store category: ________________________________
Address: _____________________________________
Date of survey: ______________________________________
Time interview was conducted: __________________________
Weather at the time of interview:
a.
b.
c.
d.
e.

Cloudy
Sunny
Rainy
Overcast
Other comments____________________________

SECTION 2: TRANSPORTATION
1.) What mode of transportation did you use for this shopping trip?
a.
b.
c.
d.
e.
f.
g.

Drove myself in a personal vehicle
Ride in car/van/truck of family or friends
Public transportation
Taxi
Walk
Bike
Other (Specify) _____________________________

2.) Do you have regular access to an automobile?
a. Yes
b. No

99

2a.) If Q2=Yes: How many automobiles do you have access to in your household?
a.
b.
c.
d.

1
2
3
4

3.) Did you travel to this location directly from your home?
a. Yes
b. No
3a.) If Q3=Yes: How long did it take you to travel here today?

3b.) If Q3=No: How long would it take you to travel directly home from this location?

4.) Is this store closest to your:
a. Home
b. Work
c. Child‘s school or daycare
d. Church
e. Gym
f. None
f. Other (specify): _____________________________
5.) Do you plan on traveling directly home after shopping at this location?
a. Yes
b. No
SECTION 3: SHOPPING HABITS
6.) How much of a problem would you say that lack of access to adequate shopping is in your
neighborhood?
a. Very serious problem
b. Somewhat serious problem
c. Minor problem
d. Not really a problem

100

7.) About how much of your household food shopping would you say is done within a 20 minute
walk or about a mile away from your home?
a. All or almost all of it
b. Most of it
c. About half of it
d. Some of it
e. None or almost none of it
8.) What is the name and location of the primary store where you typically do your food
shopping?
Name: ________________________________________________________
Location: ______________________________________________________
9.) Are you the primary food preparer in your household?
a. Yes
b. No
SECTION 4: DIET / USE OF PRODUCE
Please indicate how much you agree with each of the following statements by choosing whether
you strongly agree, agree, neither agree nor disagree, disagree, or strongly disagree.
Strongly
agree
10a.) I am a healthy eater
10b.) I am someone who eats in
a nutritious manner
10c.) I am someone who is
careful about what I eat

Agree

Disagree

Strongly
Disagree

2

Neither
agree nor
disagree
3

1

4

5

1

2

3

4

5

1

2

3

4

5

101

11.) How often do you purchase food items from each of the following types of locations?
a. Fast-food restaurants?
_______ times per day
_______ times per week
_______ times per month
b. Restaurants (not including fast-food)?
_______ times per day
_______ times per week
_______ times per month
c. Convenience stores?
_______ times per day
_______ times per week
_______ times per month
d. Grocery stores (Goodrich‘s Shop-Rite, Horrocks, L&L etc. )?
_______ times per day
_______ times per week
_______ times per month
e. Supermarkets (Meijer, Kroger, Safeway etc.)?
_______ times per day
_______ times per week
_______ times per month
f. Farmers markets?
_______ times per day
_______ times per week
_______ times per month
g. Health food stores?
_______ times per day
_______ times per week
_______ times per month

102

Party Stores/Liquor stores?
_______ times per day
_______ times per week
_______ times per month
h. Department stores (Target, Walmart etc.)?
_______ times per day
_______ times per week
_______ times per month
i. Discounters / Warehouse clubs?
_______ times per day
_______ times per week
_______ times per month
12.) Do you consume fresh fruit and vegetable items in your diet?
a. Yes
b. No
12a.) IF Q12= YES: How often do you prepare meals with fresh fruits and vegetables?
_______ times per day
_______ times per week
_______ times per month
12b.) IF Q12= YES: In a typical week how many fresh fruit and vegetable items do you
purchase?
12c.) IF Q12= YES: What fresh fruit and vegetable items do you typically purchase?
Apples
Asparagus
Avacados
Bananas
Basil
Blueberries
Broccoli
Carrots
Celery
Cherries

Cilantro
Cucumbers
Garlic
Grapes
Green beans
Green pepper
Lemons
Lettuce
Mushrooms
Oranges

Onions
Parsley
Peppers
Potatoes
Raspberries
Radish
Strawberries
Squash
Tomatoes
Zucchini
103

12e.) IF Q12=NO: Why do you choose not to consume fresh fruit and vegetable items?

13.) In a typical week how many canned fruit and vegetable items do you purchase?

14.) In a typical week how many frozen fruit and vegetable items do you purchase?

15.) If given the choice would you rather prepare meals with…?
a. Canned fruit or vegetables
b. Fresh fruit or vegetables
c. Frozen fruit or vegetables
16.) Do you purchase organic fresh fruit and vegetables items?
a. Yes
b. No
17.) Do you have your own garden at home?
a. Yes
b. No
18.) Do you have a community garden plot?
a. Yes
b. No
19) IF Q17=Yes or Q18=Yes: Do you grow fruits and/or vegetables in your garden?
a. Yes
b. No
19a.) If Q19=Yes: Do you consume the fruits and/or vegetables you grow?
a. Yes
b. No

104

SECTION 4: LOCATION OF RESIDENCE
20.) What is your home address (Non PO Box Address)?
20a.) If Q20= REFUSED: What is the intersection that is located nearest to your home?

SECTION 5: DEMOGRAPHICS
21.) What is your gender?
a. Male
b. Female
c. Don‘t know / refused
22.) What is your age?
23.) How many people are permanently living in your household?
a.
b.
c.
d.
e.
f.
g.
h.
i.

1
2
3
4
5
6
7
8+
Don‘t know / refused

24.) Do you have any children under the age of 18 residing in your household?
a. Yes
b. No
c. Don‘t know / refused
25.) What is your ethnicity?
a. White
b. Hispanic
c. African American
d. Asian
e. Other (specify) __________________________
f. Don‘t know / refuse
105

26.) What is your annual household income?
a. < $10,000
b. $10,000 - $20,000
c. $20,000 - $30,000
d. $30,000 - $40,000
e. $40,000 - $60,000
f. $60,000 - $80,000
g. $80,000 - $100,000
h. > $100,000
i. Don‘t know / refused
27.) Does anyone in your family receive food stamps/SNAP (Supplemental Nutritional
Assistance Program)?
a. Yes
b. No

106

APPENDIX B: LIST OF VARIABLES INCLUDED IN BACKWARDS STEPWISE
REGRESSION MODEL

Level of
Measurement:

Min:

Max:

Mean:

Standard
deviation:

Nominal

1

2

1.124

0.331

Number of available automobiles in
household

Ratio

1

23

2.191

1.961

Reported travel time to store

Ratio

1

60

10.413

9.048

How much of a problem respondent
felt access to adequate shopping was
in their neighborhood

Ordinal

1

4

3.368

0.935

Amount of shopping respondent did
within a 20 minute walk or a mile
away from their home

Ordinal

1

5

3.400

1.623

Whether or not respondent was the
primary shopper in their household

Nominal

1

2

1.184

0.388

How healthy respondent felt their diet
was

Ordinal

1

5

2.162

0.953

How nutritious respondent felt their
diet was

Ordinal

1

5

2.232

1.014

How careful respondent was about
what they ate

Ordinal

1

5

2.178

0.987

How frequently respondent visited
fast-food restaurants

Ratio

0

28

3.659

4.986

How frequently respondent visited
restaurants

Ratio

0

28

4.378

4.979

Variable:

Whether or not a respondent had
access to an automobile

107

How frequently respondent visited
convenience stores

Ratio

0

84

4.157

10.059

How frequently respondent visited
grocery stores

Ratio

0

28

3.630

5.037

How frequently respondent visited
supermarkets

Ratio

0

28

4.652

4.076

How frequently respondent visited
farmers markets

Ratio

0

12

1.543

2.509

How frequently respondent visited
health food stores

Ratio

0

84

1.766

6.932

How frequently respondent visited
liquor/party stores

Ratio

0

28

1.484

4.845

How frequently respondent visited
department stores

Ratio

0

28

0.788

2.369

How frequently respondent visited
discount stores

Ratio

0

28

0.679

2.522

How frequently respondent purchased
canned fruit and vegetable items

Ratio

0

20

1.852

3.391

How frequently respondent purchased
frozen fruit and vegetable items

Ratio

0

20

1.747

3.001

Whether or not a respondent
purchased organic items

Nominal

1

2

1.538

0.500

Whether or not a respondent gardened

Nominal

1

2

1.709

0.456

Gender

Nominal

1

2

1.508

0.501

Interval

18

87

44.297

15.977

Age

108

Number of people residing in
household

Interval

1

9

2.681

1.471

Whether or not children under the age
of 18 resided in their household

Nominal

1

2

1.709

0.479

Whether or not a respondent was a
minority

Nominal

1

2

1.293

0.456

Ordinal

1

9

5.440

2.732

Nominal

1

2

1.796

0.404

Income
Whether or not their household
received nutritional assistance

109

REFERENCES

110

REFERENCES

Adebawo, O., Salau, B., Ezima, E., Oyefuga, O., Ajani, E., Idowu, G., Famodu, A., et al. (2006).
Fruits and vegetables moderate lipid cardiovascular risk factor in hypertensive patients.
Lipids in Health and Disease, 5(14).
Algert, S. J., Agrawal, A., & Lewis, D. S. (2006). Disparities in access to fresh produce in low
income neighborhoods in Los Angeles. American Journal of Preventative Medicine,
30(5), 365-370.
American Community Survey Office. (2009). American Community Survey. Retrieved February
10, 2011, from http://www.census.gov/acs/www/.
Andreyeva, T., Blumenthal, D. M., Schwartz, M. B., Long, M. W., & Brownell, K. D. (2008).
Availability and prices of foods across stores and neighborhoods: the case of New Haven,
Connecticut. Health Affairs, 27(5), 1381 -1388.
Baker, E. A., Schootman, M., Barnidge, E., & Kelly, C. (2006a). The role of race and poverty in
access to foods that enable individuals to adhere to dietary guidelines. Preventing
Chronic Disease, 3(3), A76.
Baker, E. A., Kelly, C., Barnidge, E., Strayhorn, J., Schootman, M., Struthers, J., & Griffith, D.
(2006b). The garden of Eden: acknowledging the impact of race and class in efforts to
decrease obesity rates. American Journal of Public Health, 96(7), 1170-1174.
Ball, K., Timperio, A. F., & Crawford, D. A. (2006). Understanding environmental influences on
nutrition and physical activity behaviors: where should we look and what should we
count? The International Journal of Behavioral Nutrition and Physical Activity, 3.
Baranowski, T, Missaghian, M, Broadfoot, A., Watson, K, Cullen, K, Nicklas, T, Fisher, J., et al.
(2006). Fruit and vegetable shopping practices and social support scales: a validation.
Journal of Nutrition Education and Behavior, 38(6), 340-351.
Beaulac, J., Kristjansson, E., & Cummins, S. (2009). A systematic review of food deserts, 19662007. Preventing Chronic Disease, 6(3), A105.
Bell, J., & Burlin, B. M. (1993). In urban areas: many of the poor still pay more for food.
Journal of Public Policy & Marketing, 12(2), 268-270.
Bertrand, L., Thérien, F., & Cloutier, M. (2008). Measuring and mapping disparities in access to
fresh fruits and vegetables in Montréal. Canadian Journal of Public Health, 99(1), 6-11.
111

Blaikie, P., & Brookfield, H. (1987). Land Degradation and Society. Routledge Kegan & Paul.
Booth, K. M., Pinkston, M. M., & Poston, W. S. C. P. (2005). Obesity and the built environment.
Journal of the American Dietetic Association, 105(5), S110-117.
Born, B., & Purcell, M. (2006). Avoiding the local trap. Journal of Planning Education and
Research, 26(2), 195 -207.
Brunsdon, C., Fotheringham, S., & Charlton, M. (1998). Geographically weighted regressionmodeling spatial non-stationarity. Journal of the Royal Statistical Society. Series D (The
Statistician), 47(3), 431-443.
Burns, L. D., & Golob, T. F. (1976). The role of accessibility in basic transportation choice
behavior. Transportation, 5(2), 175-198.
Bush, A. J., & Hair, J. F. (1985). An assessment of the mall intercept as a data collection method.
Journal of Marketing Research, 22(2), 158-167.
Butler, J. S., & Raymond, J. E. (1996). The effect of the food stamp program on nutrient intake.
Economic Inquiry, Economic Inquiry, 34(4), 781-98.
Campbell, M. C. (2004). Building a common table: the role for planning in community food
systems. Journal of Planning Education and Research, 23(4), 341 -355.
Casagrande, S. S., Wang, Y., Anderson, C., & Gary, T. L. (2007). Have Americans increased
their fruit and vegetable intake? The trends between 1988 and 2002. American Journal of
Preventive Medicine, 32(4), 257-263.
CDC (2010). Obesity and Overweight for Professionals. Retrieved July 14, 2011, from
http://www.cdc.gov/obesity/defining.html.
Charreire, H., Casey, R., Salze, P., Simon, C., Chaix, B., Banos, A., Badariotti, D., et al. (2010).
Measuring the food environment using geographical information systems: a
methodological review. Public Health Nutrition, 13(11), 1773-1785.
Cheadle, A., Psaty, B. M., Curry, S., Wagner, E., Diehr, P., Koepsell, T., & Kristal, A. (1991).
Community-level comparisons between the grocery store environment and individual
dietary practices. Preventive Medicine, 20(2), 250-261.
Chung, C., & Myers, S. L. (1999). Do the poor pay more for food? An analysis of grocery store
availability and food price disparities. Journal of Consumer Affairs, 33(2), 276-296.
Church, R. L., & Marston, J. R. (2003). Measuring accessibility for people with a disability.
Geographical Analysis, 35(1), 83-96.
112

Clarke, G., Eyre, H., & Guy, C. (2002). Deriving Indicators of access to food retail provision in
British cities: studies of Cardiff, Leeds and Bradford. Urban Studies, 39(11), 2041 -2060.
Clifton, K. J. (2004). Mobility strategies and food shopping for low-income families. Journal of
Planning Education and Research, 23(4), 402-413.
Cochran, W. G. (1946). Relative accuracy of systematic and stratified random samples for a
certain class of populations. The Annals of Mathematical Statistics, 17(2), 164-177.
Cummins, S., & Macintyre, S. (2002). A systematic study of an urban foodscape: the price and
availability of food in greater Glasgow. Urban Studies, 39(11), 2115 -2130.
Cummins, S., & Macintyre, S. (2006). Food environments and obesity—neighbourhood or
nation? International Journal of Epidemiology, 35(1), 100 -104.
Cummins, S., Smith, D. M., Taylor, M., Dawson, J., Marshall, D., Sparks, L., & Anderson, A. S.
(2009). Variations in fresh fruit and vegetable quality by store type, urban-rural setting
and neighbourhood deprivation in Scotland. Public Health Nutrition, 12(11), 2044-2050.
Diez-Roux, A. V., Nieto, F. J., Caulfield, L., Tyroler, H. A., Watson, R. L., & Szklo, M. (1999).
Neighbourhood differences in diet: the Atherosclerosis Risk in Communities (ARIC)
Study. Journal of Epidemiology and Community Health, 53(1), 55-63.
Donkin, A. J., Dowler, E. A., Stevenson, S. J., & Turner, S. A. (2000). Mapping access to food in
a deprived area: the development of price and availability indices. Public Health
Nutrition, 3(1), 31-38.
Drewnowski, A., & Specter, S. E. (2004). Poverty and obesity: the role of energy density and
energy costs. The American Journal of Clinical Nutrition, 79(1), 6-16.
Duvall, C. S., Howard, P. H., & Goldsberry, K. (2010). Apples and Oranges? Classifying Food
Retailers in a Midwestern US - City Based on the Availability of Fresh Produce. Journal
of Hunger & Environmental Nutrition, 5(4), 526-541.
Flegal, K. M., Carroll, M. D., Ogden, C. L., & Curtin, L. R. (2010). Prevalence and trends in
obesity among US adults, 1999-2008. JAMA: The Journal of the American Medical
Association, 303(3), 235-241.
Food Systems Project Annual Report. (2010). NorthWest Initiative.
Ford, P. B., & Dzewaltowski, D. A. (2008). Disparities in obesity prevalence due to variation in
the retail food environment: three testable hypotheses. Nutrition Reviews, 66(4), 216-228.
Franco, M., Diez Roux, A. V., Glass, T. A., Caballero, B., & Brancati, F. L. (2008).
Neighborhood characteristics and availability of healthy foods in Baltimore. American
113

Journal of Preventive Medicine, 35(6), 561-567.
French, S. A., Story, M., & Jeffery, R. W. (2001). Environmental influences on eating and
physical activity. Annual Review of Public Health, 22(1), 309-335.
Gale, S. (1972). Inexactness, Fuzzy Sets, and the Foundations of Behavioral Geography.
Geographical Analysis, 4(4), 337-349.
Glanz, K., & Hoelscher, D. (2004). Increasing fruit and vegetable intake by changing
environments, policy and pricing: restaurant-based research, strategies, and
recommendations. Preventive Medicine, 39 Suppl 2, S88-93.
Glanz, K., Sallis, J. F., Saelens, B. E., & Frank, L. D. (2005). Healthy nutrition environments:
concepts and measures. American Journal of Health Promotion: AJHP, 19(5), 330-333.
Glanz, K., & Yaroch, A. L. (2004). Strategies for increasing fruit and vegetable intake in grocery
stores and communities: policy, pricing, and environmental change. Preventive Medicine,
39(Supplement 2), 75-80.
Goldsberry, K., Duvall, C. S., Howard, P. H., & Stevens, J. E. (2010). Visualizing nutritional
terrain: a geospatial analysis of pedestrian produce accessibility in Lansing, Michigan,
USA. Geocarto International, 25(6), 485.
Golledge, R. G. (1993). Geography and the Disabled: A Survey with Special Reference to Vision
Impaired and Blind Populations. Transactions of the Institute of British Geographers,
New Series, 18(1), 63-85.
Golledge, R. G. (2008). Behavioral Geography and the Theoretical/Quantitative Revolution.
Geographical Analysis, 40(3), 239-257.
Goodchild, M. F. (2009). What problem? Spatial autocorrelation and geographic information
science. Geographical Analysis, 41(4), 411-417.
Gould, P. (1969). Spatial diffusion. AAG Resource Paper no. 4 presented at the Association of
American Geographers, Washington, DC.
Gould, P. (1970). Is statistix inferens the geographical name for a wild goose? Economic
Geography, 46, 439-448.
Gundersen, C., & Oliveira, V. (2001). The food stamp program and food insufficiency. American
Journal of Agricultural Economics, American Journal of Agricultural Economics, 83(4),
875-87.
Guptill, A., & Wilkins, J. L. (2002). Buying into the food system: trends in food retailing in the
US and implications for local foods. Agriculture and Human Values, 19(1), 39-51.
114

Hägerstrand, T. (1970). What about people in regional science? Papers of the Regional Science
Association, 24(1), 6-21.
Handy, S. L., & Niemeier, D. A. (1997). Measuring accessibility: an exploration of issues and
alternatives. Environment and Planning A, Environment and Planning A, 29(7), 11751194.
Handy, S. L., & Clifton, K. J. (2001). Local shopping as a strategy for reducing automobile
travel. Transportation, 28, 317-346.
Hansen, W. G. (1959). How accessibility shapes land use. Journal of the American Institute of
Planners, 25(2), 73.
Hartman, G. W., & Hook, J. C. (1956). Substandard urban housing in United States: A
quantitative analysis. Economic Geography, 32(2), 95-114.
Hedley, A. A., Ogden, C. L., Johnson, C. L., Carroll, M. D., Curtin, L. R., & Flegal, K. M.
(2004). Prevalence of overweight and obesity among US children, adolescents, and
adults, 1999-2002. JAMA: The Journal of the American Medical Association, 291(23),
2847 -2850.
Hendrickson, D., Smith, C., & Eikenberry, N. (2006). Fruit and vegetable access in four lowincome food deserts communities in Minnesota. Agriculture and Human Values, 23(3),
371-383.
Heynen, N. C. (2006). Justice of eating in the city: the political ecology of urban hunger. In the
Nature of Cities: Urban Political Ecology (pp. 129-142). New York, NY: Routledge.
Hodgart, R. L. (1978). Optimizing Access to Public Services. Progress in Human Geography, 2,
17-48.
Howard, P. H., & Fulfrost, B. (2007). The density of retail food outlets in the central coast region
of California: associations with income and Latino ethnic composition. Journal of
Hunger & Environmental Nutrition, 2(4), 3-18.
Inagami, S., Cohen, D., Finch, B., & Asch, S. (2006). You are where you shop: grocery store
locations, weight, and neighborhoods. American Journal of Preventive Medicine, 31(1),
10-17.
Jago, R., Baranowski, Tom, & Baranowski, J. C. (2007). Fruit and vegetable availability: a micro
environmental mediating variable? Public Health Nutrition, 10(7), 681-689.
John, J., Ziebland, S., Yudkin, P., Roe, L., & Neil, H. (2002). Effects of fruit and vegetable
consumption on plasma antioxidant concentrations and blood pressure: a randomized
controlled trial. The Lancet, 359(9322), 1969-1974.
115

Joshipura, K J, Ascherio, A, Manson, J E, Stampfer, M J, Rimm, E B, Speizer, F E, Hennekens,
C. H., et al. (1999). Fruit and vegetable intake in relation to risk of ischemic stroke.
JAMA: The Journal of the American Medical Association, 282(13), 1233-1239.
Joshipura, Kaumudi J., Hu, F. B., Manson, Joann E., Stampfer, Meir J., Rimm, Eric B., Speizer,
Frank E., Colditz, G., et al. (2001). The effect of fruit and vegetable intake on risk for
coronary heart disease. Annals of Internal Medicine, 134(12), 1106 -1114.
Kamphuis, C. B. M., Giskes, K., de Bruijn, G.-J., Wendel-Vos, W., Brug, J., & van Lenthe, F. J.
(2006). Environmental determinants of fruit and vegetable consumption among adults: a
systematic review. The British Journal of Nutrition, 96(4), 620-635.
Kantor, L. S. (1998). A dietary assessment of the U.S. Food supply: comparing per capita food
consumption with food guide pyramid serving recommendations. United States
Department of Agriculture, Economic Research Service.
Koro, M. E., Anandan, S., & Quinlan, J. J. (2010). Microbial quality of food available to
populations of differing socioeconomic status. American Journal of Preventive Medicine,
38(5), 478-481.
Krebs-Smith, S. M., Reedy, J., & Bosire, C. (2010). Healthfulness of the U.S. food supply: little
improvement despite decades of dietary guidance. American Journal of Preventive
Medicine, 38(5), 472-477.
Laraia, B. A., Siega-Riz, A. M., Kaufman, J. S., & Jones, S. J. (2004). Proximity of supermarkets
is positively associated with diet quality index for pregnancy. Preventive Medicine, 39(5),
869-875.
Larsen, K., & Gilliland, J. (2008). Mapping the evolution of ―food deserts‖ in a Canadian city:
supermarket accessibility in London, Ontario, 1961-2005. International Journal of
Health Geographics, 7, 16.
Larsen, K., & Gilliland, J. (2009). A farmers‘ market in a food desert: evaluating impacts on the
price and availability of healthy food. Health & Place, 15(4), 1158-1162.
Larson, N. I., Story, M. T., & Nelson, M. C. (2009). Neighborhood environments: disparities in
access to healthy foods in the U.S. American Journal of Preventive Medicine, 36(1),
74.e10–81.e10.
Lavin, M. (2005). Supermarket access and consumer well-being: The case of Pathmark in
Harlem. International Journal of Retail & Distribution Management, 33(5), 388-398.
Liese, A. D., Weis, K. E., Schulz, M., & Tooze, J. (2009). Food intake patterns associated with
incident type 2 diabetes. Diabetes Care, 32(2), 263-268.
116

Lytle, L. A. (2009). Measuring the food environment: state of the science. American Journal of
Preventive Medicine, 36(4), S134-S144.
Mather, P., & Openshaw, S. (1974). Multivariate methods and geographical data. Journal of the
Royal Statistical Society. Series D (The Statistician), 23(3/4), 283-308.
Miller, H. J. (2010). Measuring space-time accessibility benefits within transportation networks:
basic theory and computational procedures. Geographical Analysis, 31(1), 1-26.
Miller, K. W., Wilder, L. B., Stillman, F. A., & Becker, D. M. (1997). The feasibility of a streetintercept survey method in an African-American community. American Journal of Public
Health, 87(4), 655-658.
Moore, L. V., & Diez Roux, A. V. (2006). Associations of neighborhood characteristics with the
location and type of food stores. American Journal of Public Health, 96(2), 325-331.
Morland, K., Wing, S., & Diez Roux, A. (2002a). The contextual effect of the local food
environment on residents‘ diets: the atherosclerosis risk in communities study. American
Journal of Public Health, 92(11), 1761-1767.
Morland, K., Wing, S., Diez Roux, A., & Poole, C. (2002b). Neighborhood characteristics
associated with the location of food stores and food service places. American Journal of
Preventive Medicine, 22(1), 23-29.
Morland, K., Diez Roux, A. V., & Wing, S. (2006). Supermarkets, other food stores, and obesity:
the atherosclerosis risk in communities study. American Journal of Preventive Medicine,
Morland, K., & Filomena, S. (2007). Disparities in the availability of fruits and vegetables
between racially segregated urban neighbourhoods. Public Health Nutrition, 10(12),
1481-1489.
Neumann, R. P. (2005). Making Political Ecology. New York: A Hodder Arnold Publication.
Norton, W. (1997). Human Geography and Behavior Analysis: An Application of Behavior
Analysis to the Explanation of the Evolution of Human Landscapes. Psychological
Record, 47(3), 439-460.
Ohri-Vachaspati, P., & Leviton, L. C. (2010). Measuring food environments: a guide to available
instruments. American Journal of Health Promotion, 24(6), 410-426.
Openshaw, S. (1983). The modifiable areal unit problem. Geo Books (Vol. 38). Norfolk:
Norwick.
Papas, M. A., Alberg, A. J., Ewing, R., Helzlsouer, K. J., Gary, T. L., & Klassen, A. C. (2007).
The built environment and obesity. Epidemiologic Reviews, 29(1), 129 -143.
117

Pearce, J., Witten, K., & Bartie, P. (2006). Neighbourhoods and health: a GIS approach to
measuring community resource accessibility. Journal of Epidemiology and Community
Health, 60(5), 389 -395.
Piachaud, D., & Webb, J. (1996). The price of food: missing out on mass consumption. London:
Suntory and Toyota International Centres for Economics and Related Disciplines.
Poole, M. A., & O‘Farrell, P. N. (1971). The Assumptions of the linear regression model.
Transactions of the Institute of British Geographers, 52, 145-158.
Pothukuchi, K., & Kaufman, J. L. (1999). Placing the food system on the urban agenda: The role
of municipal institutions in food systems planning, 16(2), 213-224.
Pothukuchi, K. (2004). Community food assessment: a first step in planning for community food
security. Journal of Planning Education and Research, 23(4), 356 -377.
Putnam, J., & Allshouse, J. (1997). Food consumption, prices, and expenditures 1970-95.
Statistical Bulletin No. 939. Washington, DC: USDA Economic Research Service.
Robbins, P. (2004). Political ecology: a critical introduction. Malden, MA: Wiley-Blackwell.
Rose, D., & Richards, R. (2004). Food store access and household fruit and vegetable use among
participants in the US food stamp program. Public Health Nutrition, 7(8), 1081-1088.
Sherry, B., Blanck, H., Galuska, D., Pan, L., Dietz, W., & Balluz, L. (2010). Vital signs: statespecific obesity prevalence among adults - United States, 2009 (Morbidity and Mortality
Weekly Report No. 59). Centers for Disease Control and Prevention.
Sloane, D. C., Diamant, A. L., Lewis, L. B., Yancey, A. K., Flynn, G., Nascimento, L. M.,
McCarthy, W. J., et al. (2003). Improving the nutritional resource environment for
healthy living through community-based participatory research. Journal of General
Internal Medicine, 18(7), 568-575.
Sobal, J., & Bisogni, C. A. (2009). Constructing food choice decisions. Annals of Behavioral
Medicine, 38(S1), 37-46.
Swinburn, B., Egger, G., & Raza, F. (1999). Dissecting obesogenic environments: the
development and application of a framework for identifying and prioritizing
environmental interventions for obesity. Preventive Medicine, 29(6), 563-570.
Swyngedouw, E., & Heynen, N. C. (2003). Urban political ecology, justice and the politics of
scale. Antipode, 35(5), 898-918.
Thorpe, K. E., Florence, C. S., Howard, D. H., & Joski, P. (2004). Trends: The impact of obesity
on rising medical spending. Health Affairs, Supplemental Web Exclusives: W4-480-6.
118

Turrell, G., Patterson, C., Oldenburg, B., Gould, T., & Roy, M.-A. (2003). The socio-economic
patterning of survey participation and non-response error in a multilevel study of food
purchasing behaviour: area- and individual-level characteristics. Public Health Nutrition,
6(02), 181-189.
Tyebjee, T. T. (1979). Telephone survey methods: the state of the art. The Journal of Marketing,
43(3), 68-78.
U.S. Census Bureau (2000). Census. Retrieved April 20, 2011, from
http://www.census.gov/main/www/cen2000.html.
U.S. Department of Health and Human Services and U.S. Department of Agriculture. (2010).
Dietary Guidelines for Americans. Retrieved February 10, 2011, from
http://health.gov/dietaryguidelines/.
Usual Dietary Intakes: Food Intakes, US Population, 2001-04. (2010). Risk Factor Monitoring
and Methods Branch Website. Retrieved February 10, 2011, from
http://riskfactor.cancer.gov/diet/usualintakes/pop/.
Vinkeles Melchers, N. V. S., Gomez, M., & Colagiuri, R. (2009). Do socio-economic factors
influence supermarket content and shoppers‘ purchases? Health Promotion Journal of
Australia: Official Journal of Australian Association of Health Promotion Professionals,
20(3), 241-246.
Wachs, M., & Kumagai, T. G. (1973). Physical accessibility as a social indicator. SocioEconomic Planning Sciences, Socio-Economic Planning Sciences, 7(5), 437-456.
Wang, F. (2000). Modeling commuting patterns in Chicago in a GIS - environment: a job
accessibility perspective. The Professional Geographer, 52(1), 120.
Wang, M. C., Kim, S., Gonzalez, A. A., MacLeod, K. E., & Winkleby, M. A. (2007).
Socioeconomic and food-related physical characteristics of the neighbourhood
environment are associated with body mass index. Journal of Epidemiology and
Community Health, 61(6), 491 -498.
Whelan, A., Wrigley, N., Warm, D., & Cannings, E. (2002). Life in a ―food desert.‖ Urban
Studies, 39(11), 2083 -2100.
White, M. (2007). Food access and obesity. The International Association for the Study of
Obesity (pp. 99-107). Newcastle, UK: Institute of Health and Society.
Wolf, E. (1972). Ownership and Political Ecology. Anthropological Quarterly, 45(3), 201-205.
Wrigley, N. (1983). Quantitative methods. Progress in Human Geography, 7(4), 567 -577.
119

Wrigley, N., Warm, D., Margetts, B., & Whelan, A. (2002). Assessing the impact of improved
retail access on diet in a ―food desert‖: a preliminary report. Urban Studies, 39(11), 2061
-2082.
Wylie, C., Copeman, J., & Kirk, S. F. L. (1999). Health and social factors affecting the food
choice and nutritional intake of elderly people with restricted mobility. Journal of Human
Nutrition and Dietetics, 12(5), 375-380.
Yoo, S., Baranowski, Tom, Missaghian, Mariam, Baranowski, Janice, Cullen, Karen, Fisher, J.
O., Watson, Kathy, et al. (2006). Food-purchasing patterns for home: a grocery storeintercept survey. Public Health Nutrition, 9(3), 384-393.
Zenk, S. N., Schulz, A. J., Hollis-Neely, T., Campbell, R. T., Holmes, N., Watkins, G.,
Nwankwo, R., et al. (2005a). Fruit and vegetable intake in African Americans income
and store characteristics. American Journal of Preventive Medicine, 29(1), 1-9.
Zenk, S. N., Schulz, A. J., Israel, B. A., James, S. A., Bao, S., & Wilson, M. L. (2005b).
Neighborhood racial composition, neighborhood poverty, and the spatial accessibility of
supermarkets in metropolitan Detroit. American Journal of Public Health, 95(4), 660667.

120