Citation
Examining the potential relationships between social capital, built environment and physical activity

Material Information

Title:
Examining the potential relationships between social capital, built environment and physical activity a mixed methods study
Creator:
Hill, Jennie Lynn
Publication Date:
Language:
English
Physical Description:
xv, 150 leaves : ; 28 cm.

Subjects

Subjects / Keywords:
Human ecology ( lcsh )
Social capital (Sociology) ( lcsh )
Obesity -- Social aspects ( lcsh )
Physical fitness ( lcsh )
Human ecology ( fast )
Obesity -- Social aspects ( fast )
Physical fitness ( fast )
Social capital (Sociology) ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (M.A.)--University of Colorado Denver, 2009.
Bibliography:
Includes bibliographical references (leaves 137-150).
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Jennie Lynn Hill.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
464210844 ( OCLC )
ocn464210844

Full Text
EXAMINING THE POTENTIAL RELATIONSHIPS BETWEEN
SOCIAL CAPITAL, BUILT ENVIRONMENT AND PHYSICAL
ACTIVITY: A MIXED METHODS STUDY.
By
Jennie Lynn Hill
B.S., University Nebraska Kearney, 1998
M.S., Kansas State University, 2000
A thesis submitted to the
University of Colorado Denver
In partial fulfillment
Of the requirements for the degree of
Doctor of Philosophy
Health and Behavioral Sciences
2009


2009 by Jennie Lynn Hill
All rights reserved.


This thesis for the Doctor of Philosophy
degree by
Jennie Lynn Hill
has been approved
by


Hill, Jennie Lynn (Ph.D., Health and Behavioral Sciences)
Examining the potential relationships between social capital, built
environment and physical activity: a mixed methods study.
Thesis directed by Professor Deborah S. Main
ABSTRACT
Many contextual factors have been studied to determine the
influence of physical and social environmental factors on physical
inactivity and obesity. Within this body of literature, social capital
has been linked to improved health and physical activity, yet
findings are inconsistent. There remains a lack of consensus on
the best way to measure and conceptualize social capital as it
relates to health outcomes. The overall aim of this dissertation
project is to describe variability of neighborhood social capital and
increase understanding of potential relationships between social
capital, built environment and physical activity using a two-phase
explanatory mixed methods design. Methods: Neighborhood (GIS
and walking audits) and individual-level (telephone surveys) data
were collected from 950 adults residing in five urban
neighborhoods. Independent variables included perceived (from
surveys) and objective (from walking audits) indicators of social
capital while dependent variables included self-reported physical
activity, self-rated health, and BMI. Relationships were examined
using multi-level modeling with individuals nested within
neighborhood block groups. For the qualitative component, nine
focus groups were conducted in study neighborhoods to explore a
Bourdieu-based definition of social capital. Data were recorded,
IV


transcribed, and organized into general themes and then reviewed
by a community advisory group to ensure validity. Results:
Participants who report higher perceptions of social capital (i.e.
social cohesion) and lower rated incivilities have 57% greater odds
of meeting recommendations for physical activity. Similarly, social
capital indicators, perceptions of safety and social cohesion, were
associated with higher self-rated health and lower BMI. The focus
group themes aligned with the general social capital themes of
trust, shared values and norms of reciprocity, it was more difficult
for groups to relate to social capital when conceptualized as a
resource, but the groups did provide various examples and ways of
talking about social capital. Conclusions: There are relationships
between multiple indicators of social capital and physical activity,
self-rated health, and BMI. However, the qualitative data suggest
that the current definition of social capital may be incomplete and
could contribute to inconsistent relationships between social capital
and health.
This abstract accurately represents the content of the candidates
thesis. I recommend its publication.
Signed
Deboran S. Main
V


DEDICATION
I dedicate this dissertation to Paul and Maggie for your unwavering
support and understanding throughout this process.
VI


ACKNOWLEDGEMENT
I would like acknowledge the role and efforts of the Taking
Neighborhood Health to Heart (TNHTH) research team and the
community coalition. Without their hard work, support and
dedication this project would not have been possible
VII


CONTENTS
Figures.......................................... xii
Tables........................................... xiv
CHAPTER
1. INTRODUCTION.............................. 1
Specific Aims............................. 3
Organization of the Dissertation.......... 3
2. THEORECTICAL BACKGROUND................... 4
Relevant Theories......................... 4
Social Ecological Models.................. 4
Ecological models and physical
activity............................ 7
Contextual analysis: Considerations for
ecological models......................... 10
Social Capital.................................. 12
A brief history of social capital......... 12
Social capital theory in public health
literature................................ 14
Conceptual models for social capital in
public health............................. 17
Bourdieu-based models............... 19
3. LITERATURE REVIEW............................... 24
Neighborhoods and health........................ 24
Neighborhoods and physical activity............. 25
Social capital and health....................... 27
Social capital, physical activity
and obesity............................... 30
Gaps in the literature......................... 32
Gaps this study will address.................... 34
4. QUANTITATIVE METHODS & RESULTS............ 36
Section 4.1 Study Design Overview............... 38
Rationale for mixed methods............... 38
viii


Section 4.2 Quantitative Methods................... 40
Neighborhood and Participant selection.. 40
Study neighborhoods................... 42
Sampling frame: walking audits.... 42
Sampling frame: household
participants (individual)............. 42
Data Collection Methods..................... 44
Neighborhood variables................ 45
Walking audits........................ 45
Household participant surveys..... 46
Section 4.3 Data Management and Measures....... 48
I ndependent Variables...................... 50
Neighborhood variables...................... 50
Territoriality........................ 50
Incivilities.......................... 50
Condition of alleyways................ 51
Other built environment variables. 51
Individual level variables.................. 51
Social capital indicators............. 52
Perceived safety...................... 52
Social cohesion....................... 52
Civic participation................... 53
Health resources...................... 53
Dependent Variables......................... 54
Health status......................... 54
BMI................................... 54
Physical activity.... ................ 54
Section 4.4 Specific Aims and Hypotheses....... 55
Section 4.5 Analyses and Results of
Specific Aim #1.................................... 57
Neighborhood and participant
characteristics............................. 57
Participation rate.................... 57
Participants.......................... 57
Describe social capital within
study neighborhoods......................... 59
Differences in perceived social capital. 60
Differences in objective social capital. 61
Differences between neighborhoods
IX


on dependent variables................. 71
Self-rated health.................... 63
BMI.................................. 63
Physical activity.................... 63
Summary: Describe current levels
of social capital........................... 64
Potential Relationships between social capital
and self-rated health, BMI,
and physical activity....................... 65
Specific Aim #1: Hypothesis Testing............... 67
Model #1: Diversity and
perceived social capital.................... 67
Models 2-5: Self-rated health, physical activity,
BMI, and perceived social capital........... 67
Summary Specific Aim #1........................... 68
Hypothesis 1a............................... 68
Hypothesis 1b............................... 73
Discussion.................................. 76
Section 4.6 Specific Aim #2....................... 76
Models 5-7: Objective social capital
and outcomes................................ 76
Models 8-10: Combined social capital
variables and physical activity............. 76
Neighborhood design and physical
activity & BMI.............................. 77
Summary Specific Aim #2........................... 78
Hypothesis 2a............................... 78
Hypothesis 2b & c........................... 78
5. QUALITATIVE METHODS & RESULTS................ 87
Mixed Methods Design: Qualitative................. 87
Study Design Overview............................. 87
Focus of qualitative phase.................. 87
Methods........................................... 88
Study Location and Participant Selection..... 88
Study timeline and location................. 88
Sampling frame: Focus Groups................ 88
Data Collection Methods........................... 89
Description of focus group questions... 89


Research staff and processes.......... 89
Incentives............................ 90
Audio recording and transcription..... 90
Member check.......................... 90
Analyses and Results........................ 90
Participation Rate.................... 91
Participants.......................... 91
Data Analysis......................... 91
Results of Analyses......................... 93
Theme: Traditional definitions of
social capital.............................. 93
Trust, norms of reciprocity, and
shared values......................... 93
Civic participation................... 96
Turnover of neighborhood.............. 96
Crime................................. 97
Diversity............................. 98
Theme: Resource-based definition
of social capital........................... 99
Resources for physical activity....... 99
Resources for Healthy Eating................. 100
Other resources....................... 101
Not-for-profits/community
organizations.................. 101
Role of businesses.................... 102
Library............................... 102
People as resources............ 102
Theme: Bonding, bridging
and linking social capital......................... 104
Bonding...................................... 104
Bridging..................................... 104
Linking...................................... 106
Member Check....................................... 107
Connections to Quantitative Data................... 108
Discussions of Qualitative Results................. 108
Resource based definition.................... 109
Social Networks......................... 112
XI


6. CONCLUSIONS & FUTURE DIRECTIONS............. 114
Generalizations................................. 115
Limitations of the study........................ 119
Generalizability.......................... 119
Recruitment............................... 119
Cultural and Ethnic....................... 120
Walking audits............................ 120
Contributions of the study...................... 121
Characterizing neighborhoods
On social capital......................... 121
Physical activity and social capital...... 121
Exploration of definitions of social capital... 121
Implications for Future Research................ 122
APPENDICES...................................... 124
Appendix A................................ 124
Map of Study Area................... 125
Appendix B................................ 126
Detailed description of walking audit
methods............................. 127
Audit data collection tool.......... 132
Appendix C................................ 133
TNHTH Recruitment Flowchart..... 134
Appendix D................................ 135
Recruitment for focus groups.... 136
BIBLIOGRAPHY.................................... 137
xii


FIGURES
2.1 The ecological model of physical activity (EMPA).... 9
2.2 Conceptual framework for social capital
presented in Islam et al......................... 19
2.3 Carpianos proposed conceptual model for
neighborhood-level social capital and individual
health outcomes.................................. 21
4.1 Illustration of conceptual model for neighborhood level
social capital and physical activity............. 37
4.2 Explanatory design: Follow-up Explanations
Model (QUAN emphasized).......................... 38
4.5 Conceptual diagram for multi-level data analysis.. 44
5.1 Explanatory Design: Follow-up Explanations
Model (QUAN emphasized).......................... 88
5.3 Flowchart representing the themes, codes and
meaning units around social capital described by
focus groups participants................... 95
5.4 Representation of Maslows Hierarchy of Needs..... 111
xiii


TABLES
4.3 The Taking Neighborhood Health to Heart
(TNHTH) neighborhoods........................... 41
4.4 Summary information for the TNHTH study......... 43
4.6 Summary of Measures for Dissertation............ 48
4.7 Household recruitment rates by neighborhood..... 57
4.8 Participant socio-demographic characteristics
by neighborhood................................. 58
4.9 Individual level means and standard deviations for
independent and dependent variables
by neighborhood................................... 60
4.10 Block group mean and standard deviation for
objective indicators of social capital and built
environment by neighborhood....................... 62
4.11 Unadjusted regression analysis for perceived
social capital variables and outcomes variables. 66
4.12 Final model with covariates, social capital
and outcomes...................................... 66
4.13 Model 1: Unadjusted and adjusted multi-level
models for perceived social capital and diversity. 69
4.14 Model 2: Unadjusted and adjusted multi-level
analysis for self-rated health.................... 70
4.15 Model 3: Unadjusted and adjusted multi-level
analysis for BM1.................................. 71
XIV


4.16 Model 4: Unadjusted and adjusted multi-level
analysis for physical activity......................... 72
4.17 Model 5: Unadjusted and adjusted multi-level
analysis for self-rated health......................... 81
4.18 Model 6: Unadjusted and adjusted multi-level
analysis for BMI....................................... 82
4.19 Model 7: Unadjusted and adjusted multi-level
analysis for physical activity......................... 83
4.20 Model 8. Coefficients for final model perceived
social capital and physical activity................... 84
4.21 Model 9: Coefficients for final model objective
social capital and physical activity................... 85
4.22 Model 10: Coefficients for final model perceived and
objective social capital and physical activity........ 86
5.2 Focus group participant socio-demographic
characteristics by neighborhood........................ 93
XV


CHAPTER 1: INTRODUCTION
Regular physical activity is related to a myriad of physical and
mental health benefits, yet the majority of the U S. population does
not meet recommended guidelines for physical activity. (5) A lack
of regular physical activity and poor nutrition create an energy
surplus that is driving the obesity epidemic within this country.
Strategies focused on individual behavior change and
determinants have dominated the research literature yet, thus far,
these individual strategies have yet to slow down the obesity
epidemic.(6;7)
Research has begun to focus on identifying contextual factors that
may be related to physical activity and obesity. For example, levels
of physical activity have been related to neighborhood
characteristics including the connectivity of street systems, mixed
use areas, and the proximity of parks and recreational areas. (8-12)
The Task Force on Community Preventive Services conducted a
review of physical activity interventions and found stronger support
for environmental and policy approaches to increase physical
activity than individual-level approaches.(13) In a multilevel
analysis consistent with the U.S. Preventive Task Force review,
Ewing et al. found modest yet significant associations between
U.S. county-level urban sprawl and diminished time spent walking,
and an increased risk of obesity.(12) These are just a few
examples that support a shift in physical activity research from
focusing on the individual-level theories to more comprehensive,
social ecological approaches that include both individual and
environmental components.(14)
To date, social ecological frameworks, when applied to physical
activity, has primarily focused on physical environmental changes
without considering various broader social constructs often present
social ecological models-constructs like social
l


cohesion, social support, or social capital.(15) Research on social
capital has received considerable attention, particularly in terms of
its relationship with health. For example, low levels of social capital
have been related to higher levels of mortality using state level
data sets.(16) Given the strong relationship between certain
health behaviors and all-cause mortality, there is potential that
social capital may be a significant correlate of physical activity.
Although social capital has been linked to health outcomes in a
number of settings, the magnitude, type, and level of these
relationships have been inconsistent in the literature.(17;18)'
(1; 17; 19-26) These inconsistencies make it difficult to draw
conclusions about social capital as a useful construct in public
health research. For example, there is evidence to support a
hypothesis that material or economic resources (e g. city
recreation centers or improvements in parks) may function to
facilitate the development of social capital. It also may be
hypothesized that this social capital spreads beyond the
boundaries of one neighborhood and into adjacent
neighborhoodsleading to health improvements across
communities. It could also be hypothesized that disparities in
resource allocation between neighborhoods could increase social
capital in one neighborhood while having little impact or
dampening impact on social capital in adjacent neighborhoods. To
date, available evidence neither supports nor refutes these
hypotheses. Thus, while social capital or some aspect of social
capital is likely to influence health outcomes, there is a need for
research to address the current conceptual and methodological
issues in the extant literature more systematically.
The concepts of social capital, social cohesion, social support, and
even social networks are often used interchangeably or without
clear conceptual distinction.(27) This unclear distinction hinders
measurement, analyses, and interpretation of results across
2


studies. Several researchers have called for more detailed and
fundamental work in the conceptualization and operationalization
of social capital in public health. (28;29)
The overall aim of this dissertation project is to describe
neighborhood social capital and further understand potential
relationships between social capital, built environment, and
physical activity using an explanatory mixed metiiods design.
Specific Aims of this project are:
1. To characterize current levels of social capital in the study
neighborhoods.
2. To determine potential relationships between physical activity,
built environment, and social capital within these neighborhoods.
3. To conduct focus groups within the study neighborhoods to
improve and expand understanding of quantitative findings (e.g.
levels of social capital)
Organization of the Dissertation
This dissertation consists of seven chapters. Chapters 1-3
introduce the purpose of the study and provide a theoretical
background and review of the literature. Chapters 4 and 5 detail
the methods, analysis, and results of the quantitative and
qualitative phases, respectively. The final chapter includes a
discussion of the findings, limitations, and implications of this
dissertation research along with recommendations for future
research. A comprehensive bibliography and appendices provide
additional information for the reader.
3


CHAPTER 2: THEORETICAL BACKGROUND
Relevant Theories
Ecological models have begun to gain favor in public health
research instead of the previous emphasis on individual agency
models of behavior change.(30) The emphasis of these models
is based on a universal underlying hypothesis: behavior and
health outcomes are the result of the interaction between an
individual and her/his environment Researchers in a variety of
disciplines are beginning to acknowledge the complexity of health
behavior and behavior change by examining multiple levels of
influences on individuals. According to Kawachi and Berkman,
social capital is an ecological characteristic that resides in the
social relationships and structures external to the individual.(21)
Similarly, Bourdieu defines social capital as a resource available to
individuals based on the social networks in which they reside.(31)
Therefore, a social ecological model is appropriate for research
examining the possible relationships between social capital and
health behaviors such as physical activity.
Social Ecological Models
Social ecological approaches to health behavior have their roots in
ecological systems theory.(15) Ecological systems theory posits
hierarchical relationships whereby characteristics of the
environment at one level can moderate or mediate environmental
characteristics at other levels. For example, a persons level of
physical activity may be limited by neighborhood factors such as
no sidewalks or parks. Over time, several psychologists have
theorized the extent and direction of the environments influence
on individual behavior. In the 1930s, Kurt Lewin was one of the
first psychologists to theorize that perceptions of environment
external to the individual had bearing on behavior.(32) Lewins
4


work is the basis for current psychosocial models that place human
behavior in active and reciprocal relationships with the
environment. (33) Roger Barker became one of the first
researchers to hypothesize that the environment may directly
affect current and future physiological and behavioral
outcomes.(34) Urie Brofenbrenner provides an example of an
ecological model that incorporates both physiological and social
components.(15) He proposed a multilevel model that emphasized
interactive systems that influence behavior. Those systems include
microsystems, an individuals immediate physical or social
environment; mesosystems, which link individuals to the outside
world in day-to-day activities (e.g. school or work); exosystems,
which indirectly affect individuals (e.g. community organizations);
and the macrosystem, which entails the larger cultural environment
(e.g. policy, media, and norms).(15)
Social ecological frameworks add explanatory value to individual
models of health behavior by considering factors outside the
individual that may influence behavior.(11;30;35) These
frameworks acknowledge the reciprocal relationship among
individuals and their socio-physical milieu.(36;37) A socio-
ecological model nests the individual within a larger
household/family, community, and socio-cultural context.
Importantly, this nesting allows researchers to test hypotheses
about each level as well as the interaction of factors between
levels. Sallis and Owen outline five common tenets of ecological
models: 1) there are multiple dimensions that influence health
behaviors; 2) these dimensions interact with each other; 3) there
are multiple levels of influences within the environment; 4)
environmental factors directly influence behaviors; and, 5) various
environmental factors will influence specific behaviors differently,
so ecological models should specify which factors influence which
behaviors.(38)
5


A similar model proposed by McLeroy et al. includes five levels
that factor into a given behavioral outcome(39): intrapersonal,
interpersonal, institutional, community, and public policy. The
intrapersonal level includes characteristics of the individual such as
attitudes and skills. The interpersonal level considers formal and
informal social networks and social support systems immediate to
the individual (e.g. family). A level just beyond these immediate
social ties includes the social institutions and organizations with
which the individual interacts (e.g. school). The norms, rules, and
regulations of these organizations influence behavior choices. The
next two levels include community and public policy. Broader
community organizations may influence individuals and policy
while laws and regulations constrain individual behavior in some
ways.
These various social ecological models illustrate the influence of
environmental and social contexts on individual behavior. These
influences may be positive or negative and may vary according to
factors such as time (length and frequency of contact) and
susceptibility of the individual (social norms and motivation to
comply). Using a social ecological model to study behaviors such
as physical activity and healthy eating, allows Teal world
investigations into how these complex behavior patterns occur in
the context of the larger environment in which the individual lives.
6


Ecological models and physical activity
Traditionally, behavioral models used to predict physical activity
and nutrition focused on individual perceptions and beliefs. These
approaches have had limited success in promoting long-term
maintenance of physical activity.(4;15;40) At best, correlational
research has demonstrated that these individually-focused theories
explain approximately 30% of the variability in physical activity.(6)
In response, a number of researchers have moved away from
theories focusing solely on individual perceptions and beliefs to
social ecological approaches that emphasize the interaction
between an individual and the environment.
The Ecological Model of Physical Activity (EMPA) is a social
ecological model that uses Brofenbrenners conceptualization of
systems applied to physical activity.(4) In particular, EMPA
includes pathways for biological processes and physical ecology
when determining direct versus indirect roles of the environment
on physical activity participation. Figure 2.1 is a reproduction of the
EM PA. (4) The model conceptualizes influences on physical activity
as micro-, meso-, exo-, and macro-environmental. Examples of
the microsystem include places such as work, school, home, or a
park in which the individual would participate in physical activity
(e.g. playgrounds in a park). The mesosystem is the interaction of
two or more microsystems, and is the key to physical activity. For
physical activity to occur, there must be both existing infrastructure
(e g. walking trail) and support for that activity (e.g. encouragement
from a walking partner) to promote behavior. (4) Exosystems are
composed of processes between two or more mesosystems. In a
familial context, it might be the link between parent physical activity
at the workplace and a childs activity at school. Lastly, the
macrosystems are the sociocultural factors that create the context
in which the individual lives and interacts. These sociocultural
factors influence physical activity through pathways such as
7


normative behaviors (e.g. most people walk or bike for transport)
or climate (e.g. conducive to outdoor activity). The micro- and
macro-environmental elements are relatively stable, linked by
dynamic processes of the meso- and exo- systems.(4)
Spence and Lee highlight a number of testable hypotheses using
the EMPA. These hypotheses test the relationships of individual
and environmental factors and physical activity that align with more
generalized social ecological models. According to Spence and
Lee, the environment exerts a direct effect on physical activity, with
psychological factors mediating the direct effect of the
environment. Second, biological factors (e.g., heart rate response
to physical activity) influence physical activity. Third, biological
factors are more likely to serve as moderators of the relationship
between psychological factors and behavior than as potential
mediators. (4; 15)
Ecological and social ecological models provide a practical
heuristic when guiding investigations of how social and
environmental factors interact and influence an outcome such
as physical activity. Given the complexity of social ecological
models, there are some important methodological
considerations for research using such a model.
8


Figure 2.1 The ecological model of physical activity (EMPA).(4)
9


Contextual Analysis: Considerations for Ecological Models
The framing of this dissertation project using an ecological
framework necessitates a discussion of multilevel or contextual
analyses(41). One of the complexities inherent in using an
ecological framework is that it includes data collected, analyzed,
and interpreted at individual and group levels. It is likely that
individual-level and group-level data will be correlated, thereby
violating the basic assumption of independent samples required for
most statistical analyses. These contextual analyses require
different approaches and assumptions.
Diez-Roux defines contextual analysis as the study of the effect of
collective or group characteristics on individual-level outcomes
(41 ;42). A generic contextual hypothesis may be stated as
individual risk for disease depends on individual-level factors
(genetics, lifestyle, etc), group or aggregate-level factors (e.g.
healthcare, access to parks), and the interaction of these factors.
For example, individual risk for cardiovascular disease may
depend on individual health behaviors (e.g. smoking, physical
inactivity) and a lack of resources (no safe places to walk in
neighborhood; inadequate access to health care). Additionally,
individual risk may be tied to genetic makeup, potential, and the
interaction of factors at each of these levels.
An additional concern is in drawing inferences to an individual
based on their inclusion or proximity to a given group. A common
fallacy occurs when causal inferences are drawn to individuals
based on group-level data. The opposite event, atomistic fallacy,
involves drawing inferences at the group-level based on data
collected at the individual-level. To avoid such fallacies, the level of
data collection and measurement should match the level at which
the inferences are to be drawn.
10


In defining and collecting data, an important conceptual distinction
exists between group-level and aggregate-level measures(41).
Group-level variables may be derived or integral. Derived
variables, also called aggregate or analytical variables, summarize
individual characteristics. A typical expression of derived variables
includes mean, median, percentages, or other distribution
variables. Derived variables will often have a matching individual-
level variable(41). An illustration of this would be individual income
and median block group income. In contrast, integral variables
represent characteristics of the group not derived directly from the
individual members of that group. Examples of integral variables
include availability of healthcare, political systems, or population
density. Integral variables do not have an analogue at the
individual-level.
The creation or selection of variables at the appropriate level for
measurement and analysis is critical for socio-ecological studies
(41). A common example is that of a community or neighborhood.
Individuals may declare neighborhood or community membership
with little regard to actual physical boundaries. However, as a
group-level variable, a neighborhood typically relies on
geographical boundaries. In reality, most individuals exist within
and identify with a variety of overlapping social and geographical
contexts. However, our aggregate measures need to be based on
a stable and at times arbitrary definition. Again, these data may be
correlated with one another, thus complicating the determination of
potential causal pathways.
Testing models that illustrate these links between social and
physical structure and health is important(42). For this dissertation,
the definition of neighborhood will be based on the Statistical
Neighborhoods defined by the City and County of Denver, linked to
Census Tracts. Group-level data are aggregated at level of the
block group. To address the ecological fallacy, data will be
11


collected at both the individual- and neighborhood-level. A two-
level hierarchal model will be used to test the effect of
neighborhood-level factors upon individual-level factors.
Social Capital
Social capital as a concept in public health is relatively new,
however it has a long history in the disciplines of criminology,
sociology, and political science.(1;17;21;27;43) The introduction of
social capital in public health has led to both enthusiastic support
and criticism. The criticism centers around the lack of theoretical
testing in the public health context and lack of clear
conceptualization, including definition, measurement, and unit of
analysis.(18)
A brief history of social capital
Major social theorists such as Durkheim, Marx, and Bourdieu have
shown that society is not simply the sum of the individuals, and
that the factors that influence well being at the population level
cannot be reduced to the individuals of that population.(21) In his
seminal book, Suicide, Durkheim began an inquiry into societal-
level determinants of health by asking why some societies stay
sick and others do not.(44) In Suicide, the level at which an
individual is integrated into society was shown to be protective of
overall health.(44) The works of Bourdieu and Marx extend this
idea by critically studying institutional, structural, and normative
behaviors that advance well being in some groups but not
others.(18;31;45;46) Disadvantaged groups in society may not be
able to act collectively because they lack the resources and capital
held by the more mainstream groups of a given society. This
capital may be economic, human, or social in nature. Bourdieu is
given credit for the first systemic analysis of social capital and a
theoretically sound definitional 7) Bourdieu defines social capital
as the aggregate of the actual or potential resources which are
linked to possession of a durable network of more or less
12


institutionalized relationships of mutual acquaintance or
recognition (Bourdieu 1985, p. 248; 1980). According to Bourdieu,
social capital is another advantage for privileged groups in a
society because it secures their power and position in society.
Although Bourdieu argues that possession of social capital is
reducible to economic capital, the process that brings on non-
material capital is not. Transactions resulting in social capital are
ambiguous and characterized by unspecified future obligations,
uncertain timeframes, and expectations of reciprocity. The nature
of these exchanges obscures otherwise plain market
exchanges.(17;31) Bourdieu emphasizes the complexity of
measuring social capital compared to other forms of capital (e.g.
economic) due to its intangible nature.(17)
Bourdieus conceptualization of social capital has influenced
researchers Glen Loury, James Coleman, and Robert Putnam.
Recent works by Coleman and Putnam have created resurgence
in the interest in social capital as a potential explanatory factor in
health inequities.(47;48) Loury, an economist, contends that
economic theories are too individualistic in basing human capital
on education and skills.(49) In the context of work-skill programs
designed to reduce racial inequalities, he argued that such
programs would be ineffective because among racial minorities,
poverty and low educational attainment are inherited from parents
and difficult to overcome. In turn, these children grow up to have
fewer social connections and networks within the labor market that
would allow them to move up in a free-society. While Loury makes
this anti-individualistic argument, he did little to define social capital
or refine the idea beyond the differential access to opportunities
through social connections for minority youth.(49)
Lourys work heavily influenced James Colemans definitions of
social capital in sociology. Colemans research centers on
adolescents, youth development, and the quality of relationships
13


between youth and their families, extended family, peers, and
communities. Coleman vaguely defines social capital as a
productive force in the social ties among people. (47;49) His
studies found single parent families and working mothers a central
cause of declining community cohesion and social capital.(47;49)
This view of social capital is popular in U.S. or North American-
based studies that analyzed family structure, school performance,
or youth delinquency.(49)
Finally, Robert Putnams influential work, Bowling Alone,
conceptualizes social capital as a characteristic of communities,
not a function of individuals. (26;48) According to Putnam, social
capital is the network that constitutes the civic community, peoples
sense of belonging to this network or civic community, and the
norms of cooperation and trust that govern the functions of these
networks. Levels of social capital are likely to increase (or decline)
due to economic growth.(20;26;49)
Social capitals resurgence in popularity is in part due to the
recognition that there are non-material or non-monetary
resources available to those who participate in social
networks. These resources are often unrecognized and
unmeasured sources of power and influence that may impact
outcomes, including health, for that individual.(21)
Social capital theory in public health literature
Following its rise in other disciplines, Szreter and Woolcock credit
social capitals entry into public health literature to Robert Putnam
and Richard Wilkinson.(18) Putnams 2000 book Bowling Alone
brought social capital to the forefront of public health literature in
the United States. By comparison, Wilkinson has a longer history
of work in comparative epidemiology on income and health
inequalities. As described by Szreter and Woolcock, there are
generally three different schools of thought in the public health
14


literature on social capital: social support networks, economic and
social inequalities, and political-economy or the neo-materialist
approach.(18) These schools or philosophies align with different
theoretical bases for the conceptualization of social capital. In an
editorial response to Szreter and Woolcock, Kawachi et al.
suggested that inequalities and political economy essentially make
the same argument; that material deprivation relative to others in
the network or society erodes the sense of community, trust, and
ultimately the social capital.(18;22) Kawachis studies of
inequalities and political economy will be included in the following
discussion outlining the two schools of thought on social capital.
The social support school of thought conceptualizes social capital
as the nature and extent of ones social relationships and
associated norms of reciprocity (Szreter & Woolcock; pg 651).(18)
Through mechanisms of social support, these social relationships
are hypothesized to be related to health outcomes. The vast
majority of empirical research on social capital and health has
measured levels of trust and norms of reciprocity via the
mechanism of social support.(16; 18;20;26;48;50) With this
conceptualization, social capital has been linked to improved child
development and adolescent well being, lower violent crime rates
and youth delinquency, reduced mortality, lower susceptibility to
depression and loneliness, increased mental health, and higher
perceptions of well being and self-rated health.(18) In urban and
rural neighborhoods with low levels of social capital, residents
report higher levels of stress, isolation, and reduced capacity to
respond to environmental health risks.(18)
Critics of the social support view of social capital have raised
methodological concerns primarily around the conceptualization
and measurement of trust and norms of reciprocity. Both trust and
norms of reciprocity are contextual, and therefore aggregation from
individuals or small groups to higher levels (e.g. state) may not be
15


appropriate.(18;21;51) Although research that is more recent
analyzes social capital within multilevel models, they are still
criticized for aggregating data from individuals to the community
(or state and national) level.(18;21;51) Critics question at which
level aggregation is no longer meaningful. From an analytic
perspective, Islam et al. demonstrated that the relationship
between social capital and health in fact varied as a function of the
level of aggregation and the type of analysis used.(20) A more
fundamental argument persists: If social capital as a construct is
context specific, aggregation beyond the group of interest, (e.g.
neighborhood or community) may be conceptually flawed. What
may be social capital in one group cannot be assumed at a higher
level of aggregation. As an example, for residents in a high crime
area, issues around safety and trust of neighbors may indeed best
represent that residents idea of social capital. However, a high-
income low-crime neighborhood may perceive civic involvement as
a primary component of social capital.(18;20;51) Further, social
capital through the social support view is thought to be overly
positive and disregards the potential negative effects of social
capital. Social capital may function in both inclusive and exclusive
ways while those in a network benefit greatly and those outside do
not. Social capital may be high among networks that are not
positive in nature such as organized crime or gangs.
According to the review by Szreter and Woolcock, Wilkinson led
the break from this social support view of social capital to an
inequalities conceptualization of social capital. In the inequalities
conceptualization, social capital is important to the constellation of
psychosocial factors related to socio-economic status.(18)
Wilkinson argues that in the most affluent societies, when
infectious disease is not the primary cause of death, an ever-
widening gap in socioeconomic levels is the root cause of poor
health outcomes. In these societies, there is a lessening of civic
involvement and an increase in intolerance and violence among
16


citizens accompanied by declining institutional support, which
leads to poor health outcomes. This perspective is not unique to
Wilkinson, as Marmot and colleagues have a long history of
physiological research linking psychosocial factors such as stress
and autonomy to biological outcomes.(52;53) Further, both
absolute material deprivation and relative deprivation contribute to
poor health outcomes. Relative deprivation is a factor when those
in ones family, network, or other social milieu obtain a standard of
material existence that others cannot.(54) The inequalities
perspective on social capital argues that as inequalities increase,
goodwill towards others erodes and health suffers as a
consequence; thereby placing social capital as a contributing
mechanism when considering the psycho-social milieu that may
impact biological outcomes.(16;52)
A key point in the Szreter and Woolcock article that brings both
sides of the social capital debate together is that the difference is
not over whether inequality is highly significant in accounting for
class variations in health experience in economically advanced
societies, but over the nature of the principal pathways of
causation involved. (quote from Szreter, 2004 p. 653).
Conceptual models for social capital in public health
In public health, variation exists in conceptualization of social
capital, measurement, level of measurement and unit of analysis.
For this dissertation, discussions of these differences are
limited to the common conceptual framework used in
empirical studies and the revival of a Bourdieu-based
conceptualization of social capital in recent public health
publications.
Social capital has been defined several ways.(21;26;28;46-48)
Common themes include the dimensions of trust, norms of
reciprocity, collective efficacy, participation in voluntary
organizations, and social integration for mutual benefit.(23)
17


Inherent in those themes are two key points: a) it is social and, b) it
is a public good. (21) Unlike other forms of capital, such as human
or economic capital in which the owner gamers the direct benefits,
the hypothesized benefits of social capital are publicly owned and
all should benefit from increased levels. Societies as a whole
benefit from high social capital and some argue that the lack of
individual benefits may lead to an underinvestment (e.g. civic
participation) in activities that build social capital. (21;48)
A graphical representation of the measurement and
conceptualization of social capital published in Islam et al.(20) is
shown in Figure 2.2. This is not a linear model; it simply illustrates
the proposed dimensions of social capital. Social capital has been
theorized to have both cognitive and structural aspects.(20;23)
Cognitive social capital consists of individual perceptions of norms,
values, attitudes, and beliefs. This conceptualization of social
capital is common public health literature.(20;21;48) Structural
social capital is contextual and measured by civic engagement and
societal activity. Although less common in the literature, structural
social capital includes the extent or density of social networks. In
addition to structural and cognitive, social capital has vertical and
horizontal pathways. Horizontal pathways include bonding and
bridging social capital. Bonding social capital includes relations in
homogenous groups and connects family members, neighbors,
and close friends. Bridging social capital occurs among dissimilar
groups. Vertical social capital or linking social capital is the
hierarchical or unequal relations among groups due to differences
in power and status. (20;21;23)
18


Figure 2.2. Conceptual framework for social capital presented in
Islam et al. (20)
Bourdieu-based models
Recent articles have cited Bourdieus concept of social capital as a
viable basis for advancement of social capital in public health
literature.(1;2;27;49;55) An alternative conceptualization and more
theoretically oriented framework of social capital was recently
published by Carpiano.(l) (See Figure 2.3). This conceptualization
addresses criticisms about lack of theoretical testing and unclear
definitions of social capital in public health research.(1;2;27;49;55)
A more sociological based view of social capital, such as
Bourdieus, focuses on the resources derived from social
networks. Bourdieu defines social capital as the aggregate of the
actual or potential resources which are linked to possession of a
durable network of more or less institutionalized relationships of
mutual acquaintance and recognition (Bourdieu, 1986, p. 248).
19


This conceptualization of social capital emerges from Bourdieus
consideration of how social class and inequalities are reproduced.
Like other forms of capital used to maintain or advance ones place
in society (e.g. economic, cultural), social capital may be used
alone or with other forms of capital to maintain social class. As
conceptualized by Carpiano, an individuals social capital depends
upon: (1) the size of network connections an individual can access
and use, and; (2) the amount and type of other capital possessed
by those to whom he/she is connected (e.g. economic, cultural,
symbolic). Further, social capital is not limited to the existence of
social cohesion. It also includes the actual or potential resources
possessed by social networks and the ability of individuals to draw
upon those resources. (1;2)
Carpianos conceptualization is a departure from Putnam and other
public health researchers.(l) A resource-based conceptualization
of social capital considers not only the cohesiveness of a
neighborhood, but whether neighbors rely on those relationships to
obtain resources or access that would not be available to them by
individual means. Therefore, social capital is a social process
distinct from social cohesion. The residents of a neighborhood may
be very tightly knit and socially cohesive (sharing similar values,
trust, and norms) however, they may not rely on each other to
leverage access or resources. Critics of Putnam and others often
cite a propensity for positive direction of social capital. In contrast,
Bourdieu recognizes that social capital may work in a negative
manner as well. Recognizing the exclusivity of those within the
network may impact those outside. To extend Bourdieus general
themes of social class, those who hold other forms of capital would
most likely enter the same networks, thus limiting the upward
mobility of other social classes by restricting access to social
capital.(1;31)
20


Figure 2.3. Carpianos proposed conceptual model for
neighborhood-level social capital and individual health
outcomes. (2)
Carpiano proposes the following conceptual distinctions (1 ;2):
(1) Structural antecedents: Within neighborhoods, these
antecedents are characteristics related to the structure of the
neighborhoods and surrounding areas. These factors influence the
type and strength of resources available within the neighborhood.
Examples of structural antecedents are ethnic makeup, percent
home ownership, median length of residency and income, and
adjacent neighborhood median income.
(2) Social cohesion: In this model, social cohesion is distinct from
social capital in that social cohesion must exist and acts as a
mediator between the structural antecedents and social capital.
For Carpianos model, social capital as defined by Putnam (trust,
21


norms of reciprocity, values etc.) is referred to as social cohesion.
Often, social cohesion or collective efficacy is measured as an
indicator of social capital. Carpiano considers it a distinct construct
that acts as the foundation from which social capital arises. In this
conceptualization, social cohesion may have direct effects on
health independent of social capital.(1 ;16;56)
(3) Social capital: Social capital is only the resource, actual or
potential, that is rooted within neighborhood social networks. This
construct is consistent with Bourdieu's conceptualization of social
capital. Carpiano proposes four forms of social capital: social
support, social leverage, informal social control, and participation
in neighborhood organizations. Each form of social capital is a
neighborhood-based resource that may influence health outcomes.
While Bourdieu and other researchers do not define different forms
of social capital, each would agree with Bourdieus conception that
social capital is a resource that functions to reproduce inequalities.
This resource-based definition of social capital has been applied in
other disciplines and found to be associated with non-health
outcomes such as crime, quality of life, and school performance.
(4) Outcomes of social capital: An outcome of social capital may
be a goal or benefit for the neighborhood network. That benefit
may be to individuals or the entire neighborhood. The outcome of
social capital may also be negative.
Relatively new, this model proposed by Carpiano has been tested
in a limited fashion.(2) However, it does advance
recommendations by Portes and others for theoretical refinement
in public health literature by separating proposed or related factors
from social capital itself to clearly define and operationalize the
construct.^ ;2;21 ;26;28) The strength of this theoretical
conceptualization is that it positions social capital as a mediator
between structural factors and health outcomes, and it separates
social cohesion and social capital as distinct constructs.
22


This dissertation project uses a Bourdieu-based definition of
social capital: the resources available to residents that may
promote or inhibit health behaviors such as physical activity.
Social capital is distinct from social cohesion in that
neighborhoods may be cohesive, but not rely on neighbors
for material and non-material resources that promote physical
activity and health. Carpianos conceptual model provided a
guide for hypothesis testing in the quantitative data.
Recognizing the limitations of the existing quantitative
measures and the potential of different types of social capital
(e.g. horizontal or linking) in community-based research,
Specific Aim #3 explores these aspects of social capital.
23


CHAPTER 3: LITERATURE REVIEW
Neighborhoods and Health
In countries where chronic conditions are the leading cause of
mortality and morbidity, the environment can have a dramatic, if
not direct, impact on health.(57;58) In contemporary public health
research, consideration of neighborhoods and their effect on health
is situated in 3 main dichotomies: (1) compositional or contextual
effects; (2) physical or social environmental effects; and (3) direct
or indirect effects on health.(58) Each pair leads to a different set
of hypotheses and conclusions.
Perhaps the most pervasive dichotomy relates to the compositional
versus contextual effects on health. Compositional effects are
differences in areas and individuals that can be explained by
differences among individuals in those areas; whereas contextual
effects are features of the social or physical environment that
influence the health of those exposed to that area.(58) (59) This
can be generalized in another way by considering whether poor
environments compound risk for poor health outcomes, or that
health appears to be lower in certain areas because those in poor
health tend to live in the same general location.
Another approach in public health literature distinguishes the
physical and social environment. Early research in neighborhoods
and health utilized geographic indicators such as density,
socioeconomic status, percent poverty, or percent home
ownership. These secondary indicators are capable of
characterizing a neighborhood but do not provide information on
the social aspects of that neighborhood. Research on social
environments has found that factors such as social support and
social networks can buffer the effects of poor environments.(58;59)
24


Finally, neighborhoods may have both direct and indirect effects on
health. Direct effects, such as toxins or air pollutants have an
adverse effect on health.(58) Physical and mental health may be
compromised in the form of physical attacks, stress, fear, and
anxiety. Indirect effects on health may include cumulative effects
over a life span spent in stressful or deprived areas that make
individuals more susceptible to chronic conditions. Indirect effects
may also occur due to avoidance of or inability to engage in
healthy activities (e.g. physical activity) or absence of healthy
options (e.g. fresh produce at the local market) in a given
neighborhood. If an elderly person feels unsafe walking in their
neighborhood, they may not engage in exercise. Further, they may
have less chance to interact with others leading to social isolation,
fear, or depression. A young mother who wishes to provide fresh
produce and dairy to her family may not have access to these
items at the local market. Instead, heavily marketed pre-packaged
snack foods, cigarettes, and soda pop may be her only choice.
These choices over time may adversely affect health outcomes.
What is lost in choosing a one-or-the other approach is the
consideration that each of these is interrelated and dependent. It is
not realistic to consider the geography of neighborhoods without
considering the social interactions among those residents. Both
compositional and contextual effects are important, because
people create places and places create people. These
dichotomies are artificial separations of symbiotic concepts.
Social ecological models are intuitive in studies of
neighborhoods and health as they allow for conceptualization
and testing of these various hypotheses in concert
Neighborhoods and physical activity
Physical activity research has previously focused on individual-
level factors in part due to the complexity of addressing social and
structural determinants that may be influencing individual behavior.
25


(60) Recently, the field of physical activity research has examined
the impact of social and physical environmental contexts that may
mediate physical activity behavior. (55)
Another way neighborhoods affect health is by their design and
features that may encourage physical activity. This has become a
primary focus in the fight against obesity.(11 ;12;61-67)
Environmental considerations for the promotion of physical activity
can complement individual behavior changes and lifestyle
modification interventions because they reach entire populations
and address multiple behaviors. These environmental strategies
may cause positive changes in behavior by creating a supportive
environment for physical activity and other health
behaviors.(63;68) Environments that foster physical activity
depend on an integration of land use, transportation, and
community education. Land use includes an appropriate balance
between density and open space and conscious design for non-
vehicular transportation such as mass transit, biking, or walking
for both transport and leisure. In addition to building environments
for transportation or leisure activity, it is important to create a
sense of walkability for obtaining goods, services, and
entertainment. The combination of environmental, community,
household, and individual factors have been shown to positively
influence health.(63;64;66)
Research funded by the Robert Wood Johnson Active Living
Research Initiative has provided evidence supporting the effects of
the physical (i.e. built) environment on physical activity. For
example, research indicates that people living in highly walkable
neighborhoods take twice as many trips by walking than those who
do not live in walkable neighborhoods.^ 1;63) People in mixed-use
neighborhoods were 35% less likely to be obese.(67) In research
sponsored by the CDC, investigators found that improving places
26


for activity can result in a 25% increase in percentage of people
who meet the minimum requirements for exercise.(63)
Proximity and access to recreation facilities or places to be
physically activity also matter. Adults who live near aesthetically
pleasing parks and spaces or near recreation facilities engage in
more physical activity.(69) Among low income neighborhoods,
adults who lived within a mile of a park reported 38% more
exercise sessions and were more likely to visit the park at least
once a week.(70) Access and availability of free resources for
physical activity differ by high versus low-income
neighborhoods.(71 ;72) National studies conducted in the U.S.
using geographic information systems found that neighborhoods
with higher Hispanic and African-American populations or a high
proportion of low-income residents are less likely to have public
parks. (73-75)
Thus, it is clear that the built environment influences physical
activity. However, the social environment also matters and to date
is not as well understood. Further, studies considering the
combined effects of physical and social environments are not
common.(63;64)
Social Capital and Health
The potential impact of social capital on health has been
conceptualized in several ways: a) compositional effects; b)
contextual effects; and c) larger socio-political impacts.
Compositional effects, as previously defined, are differences in
areas and individuals that can be explained by differences among
individuals in those areas.(58) Another possible mechanism linking
social capital to health is contextual effects. Contextual effects are
features of the social or physical environment that influence the
health of those exposed to that area. (59) For social capital, this
may be low availability of health-related access to care, toxic
27


hazards in environment, opportunities for physical activity, or lack
of grocery stores for healthful food options. These factors may
also have psychosocial effects such as fear, anxiety, violence, and
normative behaviors that may not be healthy in a persons
environment. (58)
In general, ecological or contextual studies have found some
support for social capital and health. For example, Kawachi et al.
found social capital to be related to all-cause and cause-specific
state-level mortality in the United States.(16) In a similar study,
Kawachi et al showed that residents of states that had overall low
levels of trust had worse self-rated health.(50) Putnam further
supports these findings by demonstrating that states that scored
high on the social capital index also scored high on a public health
index with lower age-adjusted all-cause mortality rates.(26) A
recent study from Australia found an association between low
social capital scores among the lowest income category and those
reporting poor physical health status.(76) Similarly, Swedish
researchers found a positive association between low social
contextual social capital and poor self-rated health after controlling
for individual-level social capital and socio-demographics.(77)
Despite these results, not all contextual studies have found a
relationship between social capital and health. Lynch and
colleagues found no evidence that characteristics such as trust,
control, and organizational membership were related to national
level health or international health.(78) This is similar to findings
by Poortinga who found that, when controlling for socio-economic
status, social capital was not related to self-rated health status at
the national level. A positive relationship between social capital
and health was found at the individual-level with ratings of trust
and civic participation related to higher individual ratings of
health.(79) Hooijdonk et al. found no relationship between
community social capital and all-cause mortality.(80) Examining
28


the diversity in community social capital by mortality for different
causes, demographic groups, and levels of urbanization
demonstrated weak associations with relative risks ranging from
0.92-1.09(80).
Compositional effects have been examined as well. Veenstra
studied social capital indicators of civic participation, social trust,
and sense of identity and found little evidence for compositional
effects on self rated health.(81) However, other studies (e.g.
Hyypaa and Maki) reported that mistrust, number of friends, and
group membership were related to self-rated health at the
individual-level.(82) Similar supportive findings of compositional
effects between social capital and self-rated health were found by
Schultz.(83) In a study of monozygotic and dizygotic twins, which
allowed the researchers to account for potential confounding
variables, individual-level social capital was associated with better
self-rated health.(84) Although social capital is positively related to
individual health outcomes in these studies, it is not clear that there
is an additive contextual effect on individual or population health
due to higher social capital.(79)
A review by Islam et al. examined contextual effects by
determining whether more egalitarian countries with high social
capital also had lower inequalities in health outcomes.(20) Their
review found that regardless of countries overall egalitarianism,
there was a positive association exists between social capital and
health in single-level studies. The results in studies accounting for
multilevel data were mixed. Fixed effect results across studies
demonstrated that a countrys degree of egalitarianism did not
modify the effect of social capital on health. This finding highlights
the importance and possible fallacy of drawing conclusions from
single-level studies on social capital and health because they do
not reflect the true complexity of this relationship across multiple
levels.(20) Further, categorizing social capital into more refined
29


constructs of bonding, bridging, and linking social capital may aid
in understanding social capital and its relationship to health.
Sundquist and Yang found that neighborhoods with low linking
social capital (as defined by voting in national elections) had a
higher risk of poor self-rated health than neighborhoods with the
highest levels of linking social capital.(20;51)
Kim and Kawachi combined compositional and contextual effects
into a single study in which they demonstrated a moderate
association between community social capital and self-rated health
within 40 U.S. communities.(85) Those with higher ratings of
community social capital had 4-9% lower odds of fair/poor health.
Moreover, being high on all three scales measuring community
social capital was associated with an 18% decrease in odds of
fair/poor health. In this study, individual-level social capital
measures were inversely associated with fair/poor health with a
protective effect of high participation in formal or religious groups,
volunteering, and high trust.(85)
An ecological or social ecological approach is appropriate for
the conceptualization and analysis of social capital to
untangle potential effects and more accurately determine the
impact social capital may have on health outcomes.
Social Capital, physical activity and obesity
Many aspects of social capital have been studied in relation to
physical activity, specifically walking. Dimensions of social capital,
such as trust and reciprocity, are often measured by proxies such
as crime rates or perceptions of safety, which may influence how
much a person is willing to interact in her or his environment.
Individuals living in stressful and hazardous environments are at
an increased risk for engaging in unhealthy behaviors, such as
physical inactivity. (55;86)
30


McNeill et al. identified a number of modifiable aspects of the
social environment that may influence physical activity. These
include: a) interpersonal relationships (e.g. social support and
social networks); b) social inequalities (e.g. socioeconomic position
and income inequality, racial discrimination); and c) neighborhood
and community characteristics (e.g. social cohesion and social
capital). Recent reviews of physical activity interventions found
strong support for social support interventions that increase
physical activity through avenues such as walking groups or buddy
systems.(87) Interpersonal relationships may influence physical
activity by providing social support, establishing healthy norms,
modeling healthy behaviors, and providing access to resources
such as facilities or trainers. (88) Social inequalities have been
related to physical activity through reduced resources in general.
Persons in lower socioeconomic positions are more likely to report
physical activity related to work and walking for errands or
transportation compared to higher income persons who are more
likely to engage in leisure time physical activity and sport. (89)
There is also support for the fact that those in lower income areas
have access to fewer free resources than high socioeconomic
counterparts. (71) Overall reductions in spending on parks,
sidewalks and other amenities in urban areas have been found,
particularly in low-income urban areas. (55) (90) Finally,
characteristics of communities and neighborhoods that potentially
influence physical activity are physical features (e.g. air quality,
sidewalks, and trash), community support services, and social and
cultural norms for physical activity. (55)
Specific to social capital and obesity, a literature search yielded
two epidemiological studies of this possible relationship. Kim et al.
examined U.S. state and county-level social capital in relation to
obesity and physical inactivity. (91) Secondary data analysis of
individual-level BRFSS data combined with Roper Social and
Political Trends data found that persons residing in states in which
31


the social capital score was above the median were less prone to
obesity and physical inactivity. Using similar methods, Holtgrave
and Crosby found that social capital was protective against
diabetes and obesity at the state level.(92) Notable limitations of
these studies include combinations of secondary data sets to test
the hypotheses. In some cases, the data were not from the same
source, collected during the same period or intended for these
particular research questions. The outcomes as presented by Kim
et al. vary at the county-level in which the model becomes
insignificant. Although limiting, they provide at least a reasonable
rational for the inclusion of social capital in examining community-
based measures of physical activity.(91;92)
An alternative outcome to physical activity is to focus on physical
inactivity. While there are some conceptual issues around physical
inactivity as an outcome, it is worth noting that at least one study
found associations between social capital and physical
inactivity.(93) In a sample of community dwelling adults, Mummery
et al. found those in the top quartiles of a social capital measure
were 57% and 67%, respectively less likely to be physically
inactive than those individuals in the lowest social capital
quartile.(93)
Gaps in the literature
There are many methodological and conceptual issues confronting
public health researchers interested in determining the influence of
social capital on health outcomes. (20;21;50) In the literature, there
is no consensus on what is the best conceptualization or
operational definition of social capital. Often a single item (e.g.
trust, civic participation) is used to capture or indicate social
capital. This becomes problematic due to the lack of a common or
consistent definition of social capital across studies. (20;43) Much
of this variation is due to a lack of theoretical testing or even a
clear articulation of how social capital might be related to health
32


outcomes. If it is multi-dimensional, it has not been tested to
determine which proposed dimensions of social capital are the
most salient or predictive or how they might change over time.
(20;21;23) No studies using health outcomes have used repeated
measures or pre-post study designs to test changes in social
capital over time, which might clarify how social capital develops.
The inability to test the proposed aspects of social capital is often
due to how the data are collected and measured. Most of the
empirical evidence is from cross-sectional national datasets and
often the social capital data are not from the same data collection
period or study in which the health outcomes were collection.
Thus, only correlations can be considered. In other fields, a more
robust theoretical framework for social capital exists.(1 ;43)
Due to the aforementioned fuzzy conceptualization, decisions
about at which level to measure social capital is also an issue.
Social capital is often defined as a public good with the benefits
accrued by individuals, families, communities and societies. (21)
However, there is not unanimous agreement among researchers
as to whether social capital is a group or individual good. (25) For
example, individuals concentrated in areas of low social capital
experience higher levels of social isolation.(25;56;81;94;95)
However, it is difficult to determine whether low social capital of the
area causes the social isolation or vice versa. Poortinga and
Subramanian both attributed the benefits of social capital are
attributable to social networks of the individuals.(79;96) Poortinga
found that social trust and civic participation were two dimensions
of social capital that strongly related to self-rated health at the
individual-level.(79) This relationship was not sustained at the
national level after accounting for individual socio-demographic
differences (25). In contrast, Subramanian found support at both
individual and community levels for relationships of social networks
and social capital.(78;79) (25;97)
33


Regardless of a group or individual good, most studies measure
social capital at the level of the individual and then aggregate that
score to an average for the community, state, or nation.(51 ;98)
Sundquist and Yang have advocated using a smaller scale area for
aggregation of data rather than aggregating social capital at the
community or state level.(51) Relationships between social capital
and health often disappear at the state level when controlling for
other factors. Measurement of social capital and aggregation of
data to a unit that is more meaningful to residents (e.g.
neighborhoods) might provide more consistent results and
understanding of potential impact on health outcomes. Social
capital and what that entails might be specific to at least the
neighborhood-level, it is possible that what defines social capital
and what interventions influence levels of social capital could vary
across neighborhoods, communities, or states.(51;58;81;94)
Gaps this study will address
As presented, the gaps in the current social capital literature
are sizable. It is impractical to suppose that any single
research study could address all issues raised in this review
of the social capital literature. However, social capital may
have potential to inform community-based or neighborhood-
level research on health outcomes. There is also some
preliminary evidence that it may influence obesity and
physical activity. (91 ;92) This dissertation project aims to
contribute to existing literature on social capital by
specifically addressing the following gaps:
First, this study uses primary data collection including
walking audits of the physical environment and telephone
surveys covering a variety of topics including social capital,
resources available, and self-rated health. To date, little
empirical support for social capital in public health used data
collected for the primary purpose of investigating social
34


capital and health. Second, this study more clearly defines
social capital at the outset, and uses a resource-based
definition of social capital, separating it conceptually from
social cohesion or social support Third, this study maintains
a contextually relevant unit of analysis of social capital as it
relates to physical activity and health outcomes. Finally,
recognizing the limitation of existing measures of social
capital, this study further explores the findings on social
capital through a follow-up qualitative component The goal of
the qualitative phase of the project is to explore in detail the
quantitative results as they relate to social capital and
physical activity. The quantitative results will determine the
final focus of the qualitative data, but one likely area is to
explore is the potential differences in defining social capital
among study neighborhoods. Also qualitative data collection
will help explore bridging and linking social capital as
potential pathways to unify the neighborhoods.
35


CHAPTER 4: QUANTITATIVE METHODS AND RESULTS
The purpose of this dissertation is to describe neighborhood social
capital and further understand potential relationships between
social capital, built environment, and physical activity in diverse yet
adjacent urban neighborhoods. The primary quantitative phase
includes analysis of data collected as part of the Taking
Neighborhood Health to Heart Study (TNHTH). The TNHTH study
used a community based participatory process to explore
cardiovascular disease outcomes in a set of urban neighborhoods.
The conceptual model for neighborhood level social capital and
health behaviors put forth by Carpiano guided the quantitative
analysis.(l) Figure 4.1 illustrates the conceptual model for
neighborhood level social capital and physical activity used for this
study.(1;2) Individual and environmental data collection included
telephone surveys, walking audits, and GIS-mapping and coding.
Methods included adapting, piloting, and refining previously
validated individual and environmental measures using
community-based participatory research. This chapter will cover
the quantitative methods, data collection, and results. To aid the
reader, the chapter is presented in the following sections:
4.1 Study Design
4.2 Quantitative Methods
4.3 Data Management and Measures
4.4 Specific Aims and hypotheses
4.5 Analyses and Results for Specific Aim #1
4.6 Analyses and Results for Specific Aim #2
36


Health Behavior
Physical Activity
BMI
Health Status
Self-Rated Health
Figure 4.1 Illustration of conceptual model for neighborhood level social capital and physical
activity. (1;2)
37


Section 4.1. Study Design Overview
Rationale for mixed methods
This explanatory sequential mixed method design is a two-phase
design with the qualitative phase building upon the quantitative
findings.(3) This type of mixed methods design typically
emphasizes the quantitative measures. A main advantage is the
sequential design allows a single researcher to conduct both
phases. The tradeoffs include a longer overall study timeline and
typically, a separate sampling frame is necessary for the qualitative
data collection. The explanatory mixed methods design has two
possible variations (as proposed by Creswell) in which the
qualitative research focuses on the follow-up explanation, or on the
participant selection model. This dissertation study utilized the
former, as depicted in Figure 4.2 (Creswell & Plano-Clark, 2007).
The quantitative findings provided the basis for the development of
the content for the focus groups described in the qualitative phase.
OUM JhXn 1 v QUAN s ^ ttdtysls i T OMAN results

qual Data collection qual Data analysis qual results
'Highlighted boxes indicate the focus on quantitative methods and results in this chapter.
Figure 4.2. Explanatory Design: Follow-up Explanations Model
(QUAN emphasized).(3)1
38


The rationale for using an explanatory mixed methods design
extends from the previously discussed gaps in the social capital
literature as it pertains to health outcomes. The primary data
collection for the overall TNHTH study allowed for a large random
sample within the study neighborhoods. It also allowed us to use
previously validated measures for social capital and outcomes of
physical activity and obesity. However, as noted in the introduction,
the current measurement tools used to assess social capital may
not be adequate to explore the study hypotheses related to social
capital, health, and physical activity. A priori, it was determined
that the qualitative phase will be restricted to clarifying the concept
of social capital and probing further to explore the complexities or
subtleties in social capital within the context of the study
neighborhoods.
39


Section 4.2 Quantitative Methods
The following methods section includes neighborhood and
participant selection, data collection methods, and a description of
the collected measures. The Colorado Multiple Institutional Review
Board (COMIRB) and the University of Colorado Denver Human
Subjects Review Committee (HSRC) approved all research
activities involving human participants. All members of the
research staff completed appropriate human subjects training and
project specific training as required.
Neighborhood and Participant Selection
Study Neighborhoods
Neighborhoods included in this project were participating in the
larger research project title, Taking Neighborhood Health to Heart
Study (TNHTH). Proximity to Stapleton redevelopment was a
selection criterion; therefore, each neighborhood shares at least
one physical border with Stapleton. Each neighborhood may also
border another study neighborhood. Please see Appendix A for
map of study neighborhoods (N=5).
The intentional selection of these particular neighborhoods allowed
researchers to take advantage of an occurring natural experiment
involving the Stapleton redevelopment. For decades, Stapleton
International Airport provided aviation services for the greater
Denver metropolitan area. In 1995, the airport was relocated to
outside city limits and the Stapleton site closed. Four long-standing
urban neighborhoods were adjacent to the Stapleton property.
These neighborhoods experienced decades of exposure to traffic,
noise, and air pollution. These neighborhoods include Northwest
Aurora, East Montclair, Northeast Park Hill, and Park Hill. These
neighborhoods are ethnically and economically diverse. (See
Table 4.3 for a brief description of each neighborhood).
40


Table 4.3. The Taking Neighborhood Health to Heart (TNHTH)
Neighborhoods.*_________________________________________
Pop Ethnicity % in Education (# Households) Poverty (< than HS education)
Northeast Park Hill 8,794 (2,744) 69% Afr.Am 24% Latino 24% 34%
Park Hill 19,202 (7,522) 36% Afr.Am 10% Latino 8% 12%
East Montclair 7,691 (4,218) 32% Afr.Am 32% Latino 22% 28%
Original Aurora 24,399 (7.773) 15% Afr.Am 58% Latino 27% 48%
Stapleton 6,446 (1.87.1J 3% Afr. Am 8% Latino 5% 4%
* Data from 2007 Taking Neighborhood Health to Head Study. Unless otherwise
indicated, data taken from 2007 Piton Foundation data and 2000 U.S. Census
data.
Prior to the airport relocation, a coalition of civic leaders, city
planners, residents, and foundations began working to create a
cohesive vision for reclaiming the former airport site for
redevelopment. The vision became the creation of a sustainable
mixed-use urban community. Based on Denvers historic
neighborhoods, it would have distinctive housing and
neighborhood shops on a walkable grid with tree-lined streets and
nearby offices, schools, and parks.(99) In 1998, the city selected
Forest City Enterprises, Inc as master developer to implement this
vision. The city requires a mixed-use urban design including
business and a variety of residential options ranging from low-
income apartments to single-family dwellings. Additionally, the
developer must dedicate more than 1/3 of the area to green space,
including greenbelts with biking and walking trails, pocket parks,
playgrounds, and other opportunities for active living. Another
founding principle was to recognize and accept the diversity of the
surrounding neighborhoods and keep the redevelopment
connected physically and socially to these neighborhoods. To
41


advance and ensure this vision carried through, the Stapleton
Foundation was established. Today, the Stapleton Foundation
continues to work with the developer and community members to
foster economic, social, and cultural connections to the
neighborhoods around Stapleton. In 2001, Stapleton became the
largest urban redevelopment in the United States.
(www.stapletondenver.com)(99)
Sampling Frame: Walking Audits (neighbortioocUblock group)
Data from the 2000 U.S. Census provided a basic map template to
build a sampling frame for this study. Sampling for the walking
audits used a combination of random and purposive strategies.
The goal of this combination was to ensure that each participating
household had an associated walking audit for their block. In
addition, we aimed to audit enough blocks within a neighborhood
to characterize the typical environment. Study neighborhoods were
divided into census tracts and blocks groups. Within block groups,
blocks were stratified as high/low density and randomly sorted
within each stratum. Walking auditors began by walking the first
25% of blocks within each randomly selected block group and
stratum. Walking audits began in advance of the household
recruitment by approximately 2 weeks. Once household sampling
began, auditors sampled additional blocks if the randomly selected
25% did not match the household distribution obtained through
door-to-door recruitment of households. See table 4.4 for details of
block audits.
The final number of blocks audited was 416. This is approximately
43% of all blocks possible within the study area. Exclusions for
blocks included large tracts of vacant or undeveloped land or
industrial or manufacturing areas. This was a very small portion of
the study area. Block level data have been aggregated to the block
group level for hypothesis testing.
42


able 4.4. Summary information for TNHTH Study.
N
Neighborhoods 5
Census tracts 15
Block Groups 60
Blocks* 416
Households* 950
*# of blocks with completed audit or household with completed survey
Sampling Frame: Household Participants (Individual)
For individual-level data collection, households were proportionally
recruited within neighborhoods and block groups for a total N=950.
Research staff hired and trained local members of study
neighborhoods to conduct door-to-door recruitment in each
neighborhood.
Recruiters began at a random spot on each block and proceeded
to walk in a randomly selected direction (north, east, west or south)
knocking at each house. If no one was home, recruiters left a door
tag with study information and contact phone numbers. For
efficiency of time and resources, recruiters did not return to all of
those houses originally selected. However, if interested, the door
tag contained information for the resident to call and inquire about
the study. Recruiters proceeded around a block until they reached
their quota. If unfilled, they proceeded to the adjacent block until
they met quota.
At the door, recruiters followed a standard script and using the
birthday method, chose the appropriate adults in the household to
complete the study. Once that member of the household agreed to
the telephone survey, that information was passed on to the survey
unit. The professional survey unit called each household to
conduct the telephone survey. The eligibility criteria for the
telephone survey required community-dwelling adults (18 yrs or
older) who spoke English or Spanish, lived full-time at the selected
residence, and gave their telephone or cell phone number.
43


Data Collection Methods
Individual-level data were collected using a telephone survey
administered by a professional survey unit in English or Spanish.
The survey took about 25 minutes and included questions about
overall health status, risk factors, and outcomes associated with
cardiovascular disease and other socio-demographic information.
Data used to characterize the built environment included primary
and secondary data collection. Census data and GIS-mapping and
coding provided the sampling frameworks, audit maps, and block
group level socio-demographics. Primary data collection included
the walking audit in each neighborhood. Methods included
adapting, piloting, and refining previously validated individual and
environmental measures to meet the purposes of this study. Figure
4.5 illustrates the conceptual model for multiple levels of data.
( Built environment Objective Social Capital \
Land use/density Territoriality
Walkability Incivilities
Type/Condition of Condition of
^ housing or buildings alleyways )

C Perceived Social Socio-
Capital demographics
Safety Race/Ethnidty
Social cohesion Age
Civic participation Gender
Health Resources Income
Education
Employment
V )
\
Self-Rated Health
Obesity (BMI)
Physical Activity
\______________________)
Figure 4.5 Conceptual diagram for multi-level data analysis.
44


Neighborhood Variables (Block-group)
The research team obtained 2000 Census data for each
neighborhood, except Stapleton. Maps created for each study
neighborhood included census tract, block group, and block
information. Other area-based data were obtained from the Piton
Foundation (www.piton.org). The Piton Foundation compiles data
for Denvers 77 neighborhoods based on the 2000 Census data.
These data include maps, graphs, neighborhood population, socio-
demographics, health status, median house price, median income,
educational reports, and crime for each neighborhood. This data
provided supplemental information for the study neighborhoods.
The Stapleton redevelopment project was not in existence during
the 2000 U.S. Census, therefore, census level statistics, maps,
and other similar data did not exist for Stapleton at the time of
census data collection. The developer provided the research team
with maps, demographic information, and other data sets
necessary to create the Stapleton maps. Replication of methods
used to reduce census data approximated block group and block
levels for the Stapleton data collection. To account for missing
socio-demographic census data for Stapleton, this study uses
individual socio-demographic information reported during the
household survey for all neighborhoods.
Walking Audits
To achieve an objective assessment of the built environment and
physical characteristics in each neighborhood, the research team
conducted walking audits. The hypothesized structural
characteristics included walkability, opportunities for physical
activity or recreation, grocery store or food outlets, and public
transport. Other aesthetic features hypothesized to be indicators of
social disorganization, safety, and crime included litter, graffiti,
vagrants, and the overall condition of houses and buildings. Other
features considered were density of area, lighting, parks, condition
45


and functionality of sidewalks, broken windows, and condition of
vacant lots or buildings. Auditors circled blocks matching
household recruitment methods (instead of sides or segments) to
maximize households with a matching block-level audit.
Previously validated measures were adapted and piloted to meet
the specific aims and needs of this research project (100; 101). A
detailed description of instrument development and pilot testing is
available in Appendix B. Hired auditors completed a two-day
training session conducted by the research team. In training, a
threshold of 60% agreement indicated that the trainees were
skilled enough to begin working in the field. All trainees reached
this goal on test blocks by the end of the second day. Research
staff conducted weekly quality checks in the first month and bi-
monthly checks throughout the audit timeline.
Kappas computed for all blocks with two audits were acceptable
across all blocks and neighborhoods (M=0.49-0.79). T-tests
indicated there were no consistent differences between pairs of
auditors or between blocks audited. Seventy-four percent of the
audited blocks contained at least one household survey.
For hypothesis testing, a single audit score per block was
necessary. Due to high inter-rater reliability, a random delete
strategy was appropriate. A computer generated 1-2 random
number list was assigned to all paired blocks and all ones were
deleted from the dataset. All single audits remained in the final
data set. For analyses, block group aggregate scores have been
included for all computed variables and scales from individual audit
items.
Household Participant Surveys
A professional survey unit was hired to conduct the household
surveys from July of 2007 to January of 2008. Recruiters passed
household information onto the survey unit who then proceeded
46


with follow-up calls. The phone survey took about 25 minutes to
complete. Callers from the survey unit attempted calls during the
preferred times and would attempt up to 12 call backs before
eliminating the household as a potential participant. The survey
unit had the capacity to conduct surveys in Spanish, with 156 (16%
of sample) participants completing the survey in Spanish.
47


Section 4.3 Data Management and Measures
All data collection took part under the larger TNHTH study. Table
4.6 describes the independent and dependent variables of interest,
source and measure, or scale used. Unless noted otherwise,
previously validated measures were scored according to published
protocols and a reliability coefficient reported. Research staff
checked all data at the time of collection for missing values and
out-of-range responses. Research staff compiled the data from
multiple sources into a central database. Staff used standard
research protocols, such as locked hard copy files and password-
protected databases to protect participant identities. SPSS 16.0
was used to compute new variables and scales from raw data. We
collapsed categorical variables due to limited or empty cells, as
appropriate, and conducted preliminary analyses to assure that all
variables met the assumptions of the subsequent analyses.
Table 4.6 Summary of measures for dissertation
INDEPENDENT VARIABLES
Perceived Social capital
Variable Level Method of collection Measure/Scale Used
Safety Individual Telephone Survey NEWS-A
Social cohesion Individual Telephone Survey Neighborhood Cohesion
Civic Participation Individual Telephone Survey SCBS
Health Resources Individual Telephone Survey SCBS
48


Table 4.6 (Cont)
Objective Social Capital
Variable Level Method of Collection Measure/Scale Used
Incivilities1 Block Walking Audit Indicators of incivilities (e.g. graffiti, overall condition of groups, litter, broken windows)
Territoriality1 Block Walking Audit Indicators of territoriality (e.g. proportion w/ high borders; proportion w/ window bars or security signs; proportion with dog signs)
Alleyways1 Block Walking Audit Condition and cleanliness of alleyways
Built Environment
Variable Level Method of Collection Measure/Scale Used
% Single family homes Block Walking Audit Computed from walking audit
% buildings excellent condition Block Walking Audit Computed from walking audit
%Mixed use Block Walking Audit Computed from walking audit
%Bus or Train Stop Block Walking Audit Computed from walking audit
Walkability Block GIS Computed index from GIS data (e.g. block length, intersection density, pop density)
iiMiiiiii
Variable Level Method of Collection Measure/Scale Used
Physical Activity: % meeting rec Individual Telephone Survey IPAQ-S
Obesity Individual Telephone Survey BMI Computed from self reported height & weight
Self-rated health Individual Telephone Survey Single item
49


Table 4.6 (Cont.)
Covariates
Variable Level Method of Collection Measure/Scale Used
Age Individual Telephone Survey Date of Birth
Race/ethnicity Individual Telephone Survey Hispanic/Latino Self-Identified Race
Socioeconomic status Individual Telephone Survey Household income; categorical
Educational Attainment Individual Telephone Survey Highest grade completed; categorical
Marital Status Individual Telephone Survey Are you married; divorced; widowed; separated; never married?
Employment Status Individual Telephone Survey Are you employed for wages; self-employed; out of work; student/homemaker
^aggregated to block-group level for analyses
Independent Variables
Neighborhood (Block Group) Variables
Scales computed from the audit data capture indicators of social
capital including territoriality, incivilities, and condition of alleyways.
Standardized means across items created the following variables
for objective social capital.
Territoriality
Indicators of territoriality send signals of defense against possible
invaders and are typically protective. However, they may
discourage interaction between neighbors by lowering chance
encounters thereby lowering social capital. It may indicate social
disorder and concern for crime, safety, and protection of
property.(IOO) Signs of territoriality collected during the audit
included neighborhood watch signs, high fences or borders,
50


window bars or security doors/gates, security signs (e.g. ADT), and
beware of dog signs.
Incivilities
Similarly, incivilities within a neighborhood are associated with
lower perceptions of social capital. They have also been
associated with lower physical activity and walking behavior.
Variables included graffiti, litter, overall condition of grounds or
landscaping, overall condition of buildings, condition of vacant lots,
and broken windows or abandoned houses.(IOO)
Condition of alleyways
According to our neighborhood coalition, use of and activities
taking place in alleyways was important. Auditors rated the
cleanliness of alleyways and the overall condition rated on a scale
of 1-3 (excellent-poor).
Other built environment variables
Other variables of interest of the built environment came from the
walking audit, or GIS-level data. Auditors collected information on
the type and condition of residential and non-residential buildings
in each block. Availability of public transport was a dichotomous
variable (present/absent) and overall condition of stops were rated
according to shade/shelter, benches, and cleanliness. GIS data
provided intersection density, block length, and population density
to computed walkability index. The walkability index classifies
blocks as low, medium, or high.
Individual level variables
The professional survey unit collected individual demographic,
perceptual, and outcome data. Self-reported socio-demographic
data collected during the telephone survey included age, income,
education, race/ethnicity, and employment status.
51


Social capital Indicators
Four different measures were collected as indicators of social
capital. These indicators fell under the traditional conceptualization
of social capital of trust, shared values, norms of reciprocity and
civic participation.
Perceived Safety
A subscale of the Neighborhood Environment Walkability
Scale Abbreviated (NEWS-A) measure was used to assess
perceptions of neighborhood environment and personal
safety (Saelens, Sallis, Black, et al. 2003). This instrument
has acceptable reliability and validity for telephone surveys.
(Saelens, Sallis & Black, 2003)The subscale is a nine-item
Likert scale with four response categories ranging from,
1=strongly disagree, to 4 =strongly agree. To score,
negatively worded items were reversed and coded, with
scales created by summing across items. The created
scales are M7.AII (all 9 items); M7.crime (5 items focused
on personal safety from crime); M7.traffic (4 items focused
on pedestrian safety from traffic). Reliability coefficients for
the full scale a =0.745 (n=9), were acceptable (>0.70). The
subscale coefficients were safety from traffic a = 0.430
(n=4) and safety from crime a =0.730 (n=5) respectively.
Hypothesis testing uses the nine-item scale.
Social Cohesion
The Neighborhood Cohesion scale consists of 5 items and
is a subscale of a collective efficacy questionnaire.(56) This
subscale includes the proxy measures for social capital
indicators of trust, shared values, and norms of reciprocity.
The cohesion subscale has a reliability of 0.70 (Cagney,
Browning & Wen, 2005; Sampson et. al 1997). Items
include statements such as people within the neighborhood
who can be trusted, people in the neighborhood who share
the same values, and this is a close-knit neighborhood. To
52


score, negatively worded items were reverse coded and
items summed. For this study, the Cronbach alpha was
acceptable a =0.81 (n=5).
Civic Participation
Another indicator of social capital is participation in a civic or
social group. To capture this aspect of social capital within
our study neighborhoods, items from the Social Capital
Benchmark Survey on civic participation were included. This
is a three-item scale with yes/no responses. The items
asked residents about participation in neighborhood groups
(e.g. PTA, neighborhood association), social clubs,
(activities/hobbies) and religious organizations. Each item
is scored as 0 (no) and 1 (yes) and summed across items.
Lower scores indicate negative or no response by
participants. Reliability was acceptable a =0.704 (n=3).
Health Resources
The health resources scale from the Social Capital
Benchmark Survey assessed if these social or civic
organizations are a resource for health information or health
behavior change. This is a two-item scale asking residents if
they use these groups (reference civic groups from previous
question) for information about health and if they have
changed behavior because of this information or group.
Items are scored 0 (no) and 1 (yes) and summed. Again,
the higher score indicates participation in these
organizations as a resource for health information or
behavior change. Reliability for this scale was also
acceptable a =0.70 (n=2).
53


Dependent Variables (Individual Level)
Health status
A single item assessed self-rated health on a scale 1=excellent-
5=poor.(102)
BMI
Participants reported height and weight during the telephone
survey. Calculated BMIs included a continuous and categorical
variable.
Physical Activity
The telephone survey included the International Physical Activity
Questionnaire Short Form (IPAQ-S). The IPAQ-S is a self-
administered or telephone-administered survey intended for
population-based physical activity surveillance and is primarily
used in international research. IPAQ-S consists of items that
assess physical activity in leisure, work, transportation, and
domestic activities. The short form asks about walking and
moderate and vigorous physical activity in each setting. To create
outcome measures, published protocols were followed (Guidelines
to Scoring IPAQ, Nov. 2005).Reported validity and reliability for the
IPAQ-S have both been acceptable.(103) The outcome in this
dissertation is a dichotomous outcome of meeting
recommendations for physical activity.
54


Section 4.4 Specific Aims and Hypotheses
Specific Aim #1: Characterize current levels of social capital in
the study neighborhoods.
Research questions of interest: Are there differences in social
capital between neighborhoods? Is there a relationship between
social capital and self-rated health, BMI, or physical activity?
Hypothesis 1a: Within the block-group, diversity will be
associated with lower perceived social capital.
Hypothesis 1 b: People living in block-groups with high-
perceived social capital will have higher self-rated health,
higher levels of physical activity, and lower BMI.
Specific Aim #2: To determine the potential relationships between
physical activity, built environment, and social capital in these
neighborhoods.
Research questions of interest: Is there a relationship between
perceived social capital, objective social capital indicators, and
physical activity? Does the relationship vary based on objective
versus subjective measures of social capital?
Hypothesis 2a. Participants living in neighborhoods with
higher walkability will have lower BMIs and higher rates of
physical activity.
Hypothesis 2b: Perceived social capital and objective social
capital will be associated with physical activity. Specifically,
perception of safety will have a positive association.
Increases in indicators of territoriality and incivilities will
negatively influence physical activity.
Hypothesis 2c: Perceived social capital will be a stronger
predictor than objective social capital of physical activity.
55


Typical measures of central tendency (mean, median and range)
will provide a description of the variables and characterize study
participants, independent, and dependent variables. For
continuous variables, ANOVA and appropriate post hoc tests
provide descriptive information about differences between
neighborhoods. For categorical variables, chi-square analyses are
used. Simple linear regressions assess the potential relationships
between the independent variables and the dependent variables.
As presented in Figure 4.5 this study conceptualizes a multi-level
framework consisting of individuals (level 1) nested within
households (at level 2) for hypothesis testing. Therefore, the data
are fit using random effects, multilevel linear or logistic regression,
and adjustments for individual-level covariates. Mixed effects
analyses were performed using SAS 9.2. Due to the high number
of models, a conservative significance value of p<.01 will be used.
56


Section 4.5 Analyses and Results Specific Aim #1
Neighborhood and participant characteristics
Table 4.3 shows the general characteristics of the TNHTH study
neighborhoods. For descriptive data, participant characteristics
and independent and dependent variables are grouped by
neighborhood.
Participation Rate
Overall participation rate in the TNHTH telephone survey was
36%.Neighborhoods ranged from 31%-45% participation with total
of 950 individuals completing the telephone survey. (See Appendix
C for full TNHTH recruitment flow-chart).
Table 4.7. Household recruitment rates by neighborhood
Neighborhood
East Montclair Park Hill NEPark Hill Stapleton NW Aurora Total
# contacted 328 793 245 793 767 2,646
# surveyed* 146 258 95 214 237 950
% surveyed* 68 62 71 70 57 64
"Includes only those participants who met eligibility criteria and consented to survey.
Participants
Table 4.8 shows participant characteristics by neighborhood.
Adults who participated in the household survey had a mean age
of 43.33 (14.30) and 66% were female. Forty-eight percent were
Caucasian, 24% Hispanic, 23% African American, and 5% were
other ethnic or racial categories. More than half (59%) were
married or part of a long-term couple. 18% divorced, 5% widowed,
and 18% never married. In general, it was an educated sample
with 46% holding college or professional/graduate degrees.
Another 20% had some college education, 20% a high school
diploma, and 14% had less than a high school education. Sixty-
seven percent were employed full-time, 16% were students or stay
at home mothers, and 9% retired. Only 3% reported being out of
57


work and 4% were unable to work. The median household income
fell in the 35,000-74,999 category, 17% less than $20,000, 14%
$20,000-34,999, 32% £$75,000, and 17% declined to answer the
income question. These characteristics varied greatly by
neighborhood and there were significant differences between
neighborhoods.
Table 4.8. Participant socio-demographic characteristics by
neighborhood. _______________________________________
Neighborhood
East Montclair N=146 Park Hill N=258 NEPark Hill N=95 Stapleton N=214 NW Aurora N=237
Age* Mean (Range) 41.75 (19-82) 47.34 (18-86) 45.95 (20-87) 40.30 (22-85) 41.59 (18-89)
Gender* N<%) N(%) N(%) N<%> N(%)
Male 55(38) 84(33) 10(10) 90(42) 82(35)
Female 91(62) 174(67) 85(90) 124(58) 155(65)
Race/Ettinicity* N<%> N(%) N<%) N<%) N(%)
White 68(47) 149(58) 4(4) 176(82) 58(25)
Hispanic 37(25) 22(9) 19(20) 16(8) 129(54)
Black 35(24) 74(29) 67(71) 6(3) 37(16)
Other 6(4) 12(5) 5(5) 16(7) 13(6)
Education* N(%) N(%) N<%) N <%) N<%)
< than high school 21(14) 6(2) 21(22) 0 89(38)
HS Graduate or GED 18(19) 33(13) 35(37) 9(4) 82(35)
Some college or technical school 49(34) 52(20) 27(28) 19(9) 44(19)
College graduate and/or graduate/ professional degree 48(33) 167(65) 12(13) 186(87) 21(9)
58


Table 4.8 (Cont)
Marital Status* N<%) N(%) N(%) N(%) N(%)
Married/part of unmarried couple 63(44) 145(56) 43(46) 165(78) 138(58)
Divorce/Separated 36(25) 51(20) 20(22) 23(11) 42(18)
Widowed 4(3) 17(7) 9(10) 5(2) 16(7)
Never married 41(29) 45(17) 21(23) 20(9) 41(17)
Employment* m%) N(%) N(%) N(%) N<%)
Emptoyed/self- employed 101(70) 183(71) 42(45) 167(78) 144(61)
Out of work 10(7) 9(4) 5(5) 1(<1) 7(3)
Homemaker or Student 19(13) 19(7) 25(27) 32(15) 55(23)
Retired 9(6) 40(16) 10(11) 11(5) 19(8)
Unable to work 6(4) 7(3) 12(13) 3(1) 12(5)
Household Income* N<%) N(%> N(%) N(%) N(%)
<$20,000 30(21) 23(9) 43(45) 5(2) 60(25)
$20,000-$34,999 37(25) 30(12) 17(18) 5(2) 45(19)
$35,000-$74,999 47(32) 62(24) 15(16) 40(19) 33(14)
$ >75,000 17(12) 117(45) 7(7) 150(70) 10(4)
Declined to answer 15(10) 26(10) 13(14) 14(7) 89(38)
*p < 0.001
Specific Aim #1
Describe social capital within study neighborhoods
The first aim of this research project was to describe the current
levels of social capital within each of the study neighborhoods.
Table 4.9 and 4.10 provide descriptive analyses in the form of
frequencies, means, and standard deviations for these variables,
respectively. In terms of social capital, one-way ANOVAs indicate
significant differences between neighborhoods.
59


Table 4.9 Individual level means and standard deviations for
independent and dependent variables by neighborhood.
Neighborhood
E Montclair N=146 Park Hill N=258 NE Park Hill N=95 Stapleton N=214 NW Aurora N=237
Perceived Social M M M M M
Capital Variables (SD) (SD) _ (SD) (SD) (SD)
Combined Safety* Safety Crime* Safety Traffic* Social Cohesion* Civic Participation* Health Resources3* 2.63 (0.52) 2.74 (0.64) 2.50 (0.64) 3.54 (0.84) 5.32 (0.77) 3.51 (0.74) 3.03 (0.47) 3.14 (0.52) 2.88 (0.64) 4.05 (0.73) 4.81 (0.97) 3.29 (0.84) 2.70 (0.42) 2.79 (0.62) 2.60 (0.57) 3.44 (0.88) 5.05 (0.88) 3.02 (0.93) 3.36 (0.4) 3.56 (0.39) 3.13 (0.56) 4.42 (0.58) 4.79 (0.94) 3.25 (0.82) 2.53 (0.5) 2.51 (0.67) 2.55 (0.59) 3.10 (0.9) 5.33 (0.7) 3.54 (0.77)
Outcome Variables M (SD) M (SD) M (SD) M (SD) M (SD)
Self-rated Health* BMI* Physical Activity (meet rec. yes/no)* 2.62 (1.18) 26.97 (6.33) 46 2.19 (0.94) 26.38 (5.7) 65 2.92 (1.02) 29.40 (5.91) 41 1.80 (0.76) 24.42 (3.62) 66 3.08 (0.99) 28.33 (6.95) 33
*p < 0.001; ANOVA or )C indicates significant differences between
neighborhoods.
increasing scores indicate a higher number of no responses meaning they do not
participate in these groups or use for health resources
Differences in perceived social capital (individual)
Perceived safety operationalized as perceptions of safety from
crime or traffic differed significantly by neighborhood
F(4,945)=107.5 (p<.001). Residents of Stapleton held more
positive perceptions of safety within their neighborhood
(Mean=3.36) and Northwest Aurora residents held the lowest
perceptions of safety (Mean =2.53). Post hoc analyses indicate
60


that Northwest Aurora is significantly different from all other study
neighborhoods (p<.001). East Montclair and Northeast Park Hill
were significantly lower than Stapleton and Park Hill (p<.001).
Park Hill and Stapleton were significantly different from each other
as well (p<.001).
Social capital in the form of trust, shared values, and norms of
reciprocity also differed significantly by neighborhood
F(4,944)=93.16, p<.001). Northwest Aurora was significantly lower
than the other study neighborhoods (p< 001). Again, Northeast
Park Hill and East Montclair were lower than Park Hill and
Stapleton. Stapleton and Park Hill were higher than the other
neighborhoods but still significantly different from each other. It
should be noted that social cohesion was high overall with means
ranging from 3.1-4.4.
Social capital measured as participation in civic groups or activities
differed significantly between neighborhoods F(4,945)=21.69,
p.<001). Stapleton and Park Hill were significantly higher than the
other study neighborhoods (p<001). Generally, all neighborhoods
had a low score on this measure, indicating overall low levels of
reported participation in civic activities or groups.
Using civic resources as a resource for health information also
differed significantly between neighborhoods F(4,609)=5.27,
p.<001). NE Park Hill was significantly lower than East Montclair
and NW Aurora (p<.001). The mean values across all
neighborhoods were low, indicating these places were not a
source for health information or help with behavior change.
Differences in objective social capital (block-group level)
Incivilities per block group differed by neighborhood
F=(4,55)=30.62, p<.001). Post hoc analyses indicate that
Stapleton redevelopment and Park Hill had significantly (p<.001)
61


lower incidents of incivilities than East Montclair, NW Aurora, and
NE Park Hill. Territorialities per block group differed by
neighborhood F(4,55) = 4.55, p<.01). Again, significant differences
between neighborhoods exist. Stapleton had significantly lower
(p< 001) indicators of territoriality than NW Aurora. There were no
significant differences in indicators of territoriality between Park
Hill, East Montclair, and NE Park Hill. The highest numbers of
indicators of territoriality were in NW Aurora. Ratings of cleanliness
and condition of alleyways differed by neighborhood F(4,55)=4.55,
p <.01). Post hoc analyses indicate that Stapleton had significantly
more positive ratings for alleyways (p<.001) than all other
neighborhoods.
Table 4.10. Block group mean and standard deviation for objective
indicators of social capital and built environment by neighborhood.
Neighborhood
E Park Montclair Hill N=4 N=24 NEPark Hill N=7 Stapleton N=4 NW Aurora N=21
M(SD) M(SD) M(SD) M(SD) M(SD)
Incivilities* 0.37 0.28 0.41 0.13 0.43
(.04) (.08) (.04) (.03) (.05)
Territoriality* 0.33 0.35 0.34 0.27 0.40
(.08) (.07) (.06) (.07) (.04)
0.50 0.35 0.37 0.002 0.49
Alleyways* (.06) (-29) (.11) (.005) (.16)
% Single family housing 85% 95% 70% 88% 74%
% buildings excellent condition 28% 60% 18% 100% 9%
% mixed use 27% 12% 9% 23% 4%
% Bus or Train Stop Walkability (% high) 23% 20% 43% 4% 20%
74% 22% 68% 94% 25%
*p < 0.001; ANOVA or x1 indicates significant differences between
neighborhoods.
62


Differences between neighborhoods on dependent variables
Self-Rated Health
The original scoring was reversed so that a low value indicates a
poorer reported self-rated health. The mean value for self-rated
health was 2.46 (SD=1.08) and a range from 1-5. One-way
ANOVAs indicated significant differences by neighborhood
F(4,945)=61.39fp<.001); NWA was the lowest and Stapleton the
highest. Post hoc analyses indicate that NWA and NEPH are
similar as are PH and EMC. Stapleton was different from all other
study neighborhoods (p< 0.001).
BMI
BMI measured as an individual-level continuous variable had a
mean value of 26.71 (SD=5.9) with a minimum reported 15.38 and
high of 61.07. Stapleton had a mean BMI of 24.42. One way
ANOVAs indicated that there were significant differences between
neighborhoods (F (4,879)=17.45,p<.0001). Post hoc analysis
indicates that participants in Stapleton had significantly lower BMI
than the other neighborhoods. Park Hill and EMC are similar to
each other while NWA and NEPH, with the highest mean BMIs,
are different from the other three neighborhoods.
Physical activity
Physical activity was measured as a continuous variable in
MET/MINS and categorized into a dichotomous variable for
meeting current recommendations for moderate to vigorous
physical activity. The mean MET/MIN for the sample was
2,588(SD=3500). Just over half of the sample (52%) was meeting
recommendations for physical activity. Mean neighborhood values
for meeting recommendations ranged from 41-66%. Neighborhood
patterns were similar to BMI. Stapleton and Park Hill had over 60%
report meeting recommendations and these were significantly
different (p<.001) than the other study neighborhoods. EMC was
significantly different from all other neighborhoods. NWA and
63


NEPH were not different from each but significantly lower the other
study neighborhoods.
Summary: Describe current levels of social capital
In general, social capital was moderate to high across
neighborhoods. Significant differences do exist on both objective
and perceived measures of social capital. Different from
perceptions of safety, most residents felt social capital existed
within their immediate block or among close neighbors. Social
capital in the form of trust, norms of reciprocity, and shared values
existed all neighborhoods. The data support a general hypothesis
that diverse neighborhoods tended to have lower perceptions of
social capital in terms of safety, neighborhood resources, and
higher presence of incivilities and indicators of territoriality. NW
Aurora and East Montclair, both diverse and lower socioeconomic
neighborhoods, had lower social capital in terms of incivilities and
indicators of territoriality. However, perceived social capital, as
rated by residents, was moderate.
In general, Stapleton redevelopment had the highest scores for
both objective and perceived social capital. It is not surprising that
objective measures indicated high levels of social capital in
Stapleton. Not only is the neighborhood new; it meets and often
exceeds current building codes that require features such as curb
cuts, wide sidewalks, lighting, and open space. The homes are
new and well maintained. In the traditional neighborhoods, there is
more variety in the age, style, and condition of homes and
surrounding homes or businesses. The objective social capital
indicators come from the walking audit in which trained research
staff, not residents, assessed the neighborhood on these factors.
64


Potential Relationship between social capital and self-rated health.
BMI and physical activity.
A second research question of interest is to examine potential
relationships between perceived social capital and the outcomes of
self-rated health, BMI, and physical activity. Hierarchical, linear, or
logistic regression with purposeful selection was used to build the
models. Initial unconditional regression models for each perceived
social capital variable and each outcome were significant. Next, all
perceived social capital variables were entered simultaneously into
the model with each outcome. Table 4.11 shows the parameter
estimates, standard errors, and p-values. Perceptions of safety,
social cohesion, civic resources, and health resources were
significantly related to self-rated health (F=33.26(4,609) (p<0001).
The perceived social capital variables were significantly related to
BMI and physical activity, (F=11.89(4,564)p<.0001) and
F=14.58(4,609)p<0001) respectively.
Refinement of the preliminary model included the addition of
theoretically or conceptually important covariates to account for
factors that may also influence self-rated health, physical activity,
and BMI. Potential covariates were prescreened using bivariate
correlations and variables that were highly correlated (>.70)
remained in the model at future steps. All possible variables
entered the model. Variables that were significant at p<0.25
remained. The process repeats until only significant covariates
remain, resulting in a preliminary main effects model. To keep a
standard set of covariates across all outcomes, significant
covariates in any one of the models remain in the model. Table
4.12 shows the results for this model.
65


Table 4.11. Unadjusted regression analysis for perceived social
capital variables and outcome variables._________________
Dependent Variables
Self-Rated Health BMI Physical Activity
Individual Level Social Capital WSE) P(SE) 3(SE)
Perceived Safety -0.20 (.09)* -0.14 (.48)* -0.89 (.04)*
Social Cohesion(Trust, Shared -0.30 -0.16 -0.15
Values, Norms of reciprocity) (06)* (-30)* (-03)*
Civic Participation 0.09 (-06)* -0.04 (.34) 0.11 (-03)*
Health Resources 0.03 (-05) -0.09 (-27)* 0.11 (-02)*
Table 4.12 Final model with covariates, social capital and
outcomes
Outcomes
Self-Rated Health BMI Physical Activity
Covariates B(SE)* B(SE)* B(SE)*
Gender (ref=male) .00(08) -0.13(.49) -0.06(.04)
Age (ref=30-49) 18-29 50-64 -.06(12) .12(.09) -0.11 (.69) 0.11 (.53) -0.05(.06) -0.03(.05)
Race/Ethnicity (ref=white) Hispanic Black Other .16(13) 0.09(11) .06(.18) 0.21(77) 0.11 (.62) 0.04(1.06) 0.09(.07) 0.07(.06) -0.02(.09)
Income (ref=middle income) Low High Decline to answer 04(.12) -.09(11) .02(.14) 0.09(.72) -0.06(.60) 0.00(80) -0.01 (.07) -0.02(.06) 0.15(.07)
Educational Attainment (ref=grad/professional degree) HS or less .23(.17) -0.05(1.05) Some college/2yr degree ,03(.13) 0.06(.74) College Degree ,08(.10) 0.09( 63) Unemployed .17(.20) 0.01(1.47) 0.04(09) 0.05(07) 0.06(.06) 0.08(.13)
Employment Status (ref=employed) Homemaker/Student 10(. 11) Retired -.08(.13) Disabled .17(20) 0.05(.65) -0.05(.75) 0.09(1.11) 0.01 (.06) 0.02(.07) 0.12(.10)
66


Specific Aim #1: Hypothesis Testing
Model 1: Diversity and perceived social capital
The hypothesis that higher diversity would be related to lower
social capital was tested with a random effects 2-level regression
model. The diversity index includes both ethnic and economic
diversity and categorizes block groups as low, medium, or high
diversity. Table 4.13 provides the results of unadjusted and
adjusted social capital variables and diversity. The following
discussion on models is restricted to adjusted models only, but
unadjusted data are available in the tables.
There was a significant inverse relationship between diversity and
perceived safety, social cohesion, and civic participation. As
diversity increases, participants reported lower perceived social
capital in the form of safety, cohesion and civic participation.
Health resources were not significant.
Models 2-5: Self-rated health, physical activity. BMI and perceived
social capital.
The next hypothesis under specific aim #1 was to test the
relationship between perceived social capital and outcomes of self-
rated health, physical activity, and BMI. Tables 4.14-4.16 include
unadjusted and adjusted models for each outcome. Perceived
safety and social cohesion were significantly associated with self-
rated health. A significant inverse relationship exists between
perceptions of safety, social cohesion, health resources, and BMI.
As perceptions of social capital decrease, BMI increases.
Perceived social capital indicators of social cohesion, civic
participation, and health resources were positively associated with
meeting recommendations to physical activity. Individuals who
reported higher social cohesion had 65% greater odds of meeting
recommendations for physical activity (OR=1.65).
67


Summary Specific Aim #1
Hypothesis 1 a: Diversity will be associated with lower perceived
social capital within block group.
Block group diversity was associated with social capital in the
hypothesized direction. After controlling for individual-level
covariates, lower block group diversity was related to higher social
capital indicators of perceived safety, social cohesion, and civic
participation. Lower income persons in diverse block groups report
lower perceived safety, social cohesion, and civic participation.
Persons < 29 years of age reported lower civic participation in
diverse block groups.
68


Table 4.13 Model 1: Unadjusted and adjusted multi-level models for perceived social capital and
Unadjusted Adjusted1
Parameter SE P Estimate ICC Parameter SE P ICC
Outcome: Perceived safety2 (Uncond. ICC=0.20)a Diversity -0.52 0.04 <.001 Intercept 4.03 0.10 <.001 .06 -0.49 4.10 0.05 0.13 <.0001 <.0001 .06
Outcome: Social Cohesion2 (Uncond. ICC=0.31)8 Diversity -0.80 0.09 <.0001 Intercept 5.62 0.23 <.0001 .13 -0.63 5.00 0.09 0.23 <.001 <.001 .08
Outcome: Civic Participation3 (Uncond. ICC=0,09)* Diversity -0.40 0.07 <.0001 Intercept 1.87 0.16 <.0001 .03 -0.23 1.17 0.08 0.21 .004 <.001 .03
Outcome: Health Resources3 (Uncond. ICC=0.07)a Diversity -0.19 0.08 <.02 Intercept 1.08 0.18 <.0001 .05 -011 0.77 0.08 0.23 .167 .001 .02
2 higher rating indicates positive perceptions of safety, social capital
3lower score indicates low/no civic participation or use of community groups as a health resources
8 Uncond. ICC=Unconditional Inter-class correlation dependent variable; unconditional ICC provides an indicator of the between
block group variation with dependent variable only in model
69


Table 4.14 Model 2: Unadjusted and adjusted multi-level analyses for self-rated health.8
Unadjusted Adjusted1
Parameter Estimate SE P ICC Parameter Estimate SE P ICC
Perceived Social Capital Variables
Perceived Safety2 .344 .065 <.0001 0.14 .283 .056 <.0001 0
Intercept 2.48 .191 <.0001 3.08 .213 <.0001
Social Cohesion2 .231 .039 <.0001 0.13 .174 .035 <.0001 0.08
Intercept 2.603 .153 <.0001 3.30 .183 <.0001
Civic Participation3 .071 .036 .053 0.19 0.026 .034 0.193 0.005
Intercept 3.38 .076 <.0001 3.90 .132 <.0001
Health Resources3 .070 .049 .157 0.22 .0128 .046 .780 0.007
Intercept 3.40 .084 <.0001 4.07 .171 <.0001
individual level covariates controlled for: gender, age, ethnicity/race, income, education and employment status.
2 higher rating indicates higher perceptions of safety, social capital
3lower score indicates low/no civic participation or use of community groups as a health resources
aUncond ICC=0.20
70


Table 4.15 Model 3: Unadjusted and adjusted multi-level analyses for BMI.a
Unadjusted Adjusted^
Parameter Estimate SE P ICC Parameter Estimate SE P ICC
Perceived Social Capital variables
Perceived Safety2 -1.38 0.38 .0003 0.004 -1.00 0.37 <.006 0
Intercept 31.00 1.11 <.0001 29.15 1.20 <.0001
Social Cohesion2 -0.87 0,22 .0001 0.04 -0.70 0.23 0.003 0.17
Intercept 30.29 0.88 <.0001 29.15 1.20 <.0001
Civic Participation3 -0.13 0.22 .551 0.08 -0.12 0.22 0.573 0.21
Intercept 27.26 0.37 <.0001 26.69 0.85 <.0001
Health Resources3 0.71 0.27 .009 0.15 0.68 0.27 .011 0.21
Intercept 26.43 0.41 <.0001 26.88 1.00 <.0001
individual level covariates controlled for: gender, age, ethnicity/race, income, education and employment status.
2 higher rating indicates higher perceptions of safety, social capital
3lower score indicates low/no civic participation or use of community groups as a health resources
"Uncond ICC=0.08
71


Table 4.16 Model 4: Unadjusted and adjusted multi-level analysis for physical activity.8
Unadjusted Adjusted1
Parameter Estimate SE P ICC OR Parameter Estimate SE P ICC OR
Perceived Variables
Perceived Safety* 0.10 0.03 <.0001 0.06 2.71 0.04 0.03 0.16 0.01 1.49
Intercept 1.77 0.09 .002 0.32 0.12 <.008
Social Cohesion* 0.08 0.02 <.0001 0.05 2.23 0.05 0.02 0.01 0.01 1.65
Intercept 1.80 0.08 <.0001 0.28 0.10 <.007
Civic Participation'3 0.05 0.02 .005 0.07 0.61 0.04 0.02 0.02 0.00 1.49
Intercept 1.55 0.03 <.001 .0.43 0.07 <.0001
Health Resources3 0.09 0.02 .0004 0.150 0.41 0.006 0.03 0.01 0.03 1.06
Intercept 1.54 0.03 <.0001 0.47 0.09 <.0001
individual level covariates controlled for: gender, age, ethnicity/race, income, education and employment status.
2 higher rating indicates higher perceptions of safety, social capital
3lower score indicates low/no civic participation or use of community groups as a health resources
Uncond ICC=0.06
72


Hypothesis 1b: Persons living in block groups with higher
perceived social capital will be associated with higher self-rated
health, higher levels of physical activity, and lower BMIs.
Higher perceptions of safety and social cohesion were associated
with higher self-rated health. This relationship was in the
hypothesized direction. Not surprising, those with higher income
and higher self-rated health had positive perceptions of safety and
social cohesion. Elderly participants (>65 yrs) with lower income,
and lower educational attainment had lower self-rated health and
perceptions of social capital. Civic participation and health
resources were not significant.
The data show an inverse relationship between BMI, perceptions
of safety, and social cohesion. There was a positive association
between health resources and BMI, indicating that individuals with
higher BMIs reported using civic groups or organizations for
information regarding health. Persons with lower perceptions of
safety and lower social cohesion had higher BMIs. Ethnic
minorities, females, and those over 65 years of age had higher
BMIs and lower perceptions of safety and social cohesion.
Discussion
The results partially support the first set of hypotheses. Perceived
safety and social cohesion are lower in more diverse block groups
The scale used for social cohesion included questions about trust,
shared values, and norms. In other disciplines, the link between
increased diversity and decreased social capital is supported.(56)
Similar to several studies, there is support for the relationship
between perceived social capital in the form of perceived safety,
social cohesion, civic participation, and self-rated
health.(21 ;50;50;76;77;83;85;97) (25)
73


While civic participation was significantly related to self-rated
health, it was not significantly related to BMI or meeting physical
activity recommendations. In general, civic participation was low
across all neighborhoods. It may not have much relevance as a
predictor if so few in the study population report these types of
activities. Alternatively, the measure may not capture some of the
ways in which participants may be active in their communities. The
health resources question is tied to civic participation in the
sequence of the survey, thus if few people reported participation in
various community groups, it might be assumed they would not be
using these groups for health information. Measures of
participation in civic organizations may capture another aspect of
social capital (e.g. social networks) that does not correspond to
physical activity.
Interestingly, perceived safety had an inverse relationship with
BMI, but no significant association with physical activity. Other
research has found that higher social capital, including perceptions
of safety was protective against obesity and associated with
increasing level of physical activity.(91;92) Currently, these data
support an association between higher social capital in the form of
perceptions of safety and lower BMI and not an increase in
physical activity. In this dissertation, the measured outcome of
meets the recommendations for physical activity, is based on
moderate to vigorous activity. Therefore, the measure does not
capture leisurely activity or walking, which may be influenced by
negative perceptions of safety. For example, perceived safety
might negatively influence walking in ones neighborhood. Other
causes for increased BMI, related to perceptions of safety, (such
as increased stress) might also influence physical activity.
74


The results show a positive association between increasing BMI
and the use of civic organizations for health information. However,
BMI was not related to civic participation.
This study supports a positive association between physical
activity and perceived social capital indicators such as social
cohesion and health resources. A small number of studies have
examined the relationship between physical activity and social
capital.(91;93; 104-107) These studies found a positive or inverse
relationship with social capital indicators when the outcome was
physical inactivity. (91;93;106) Current results show that those who
report having higher social capital in the form of social cohesion
and health resources are more likely to meet recommendations for
physical activity. This is similar to Wen et al., who found a positive
relationship between social capital (measured as trust and norms
of reciprocity) and physical activity.(108) Lindstrom et al.
conceptualized social capital as social participation and found that
people who were more engaged in their community reported
increased physical activity.(109) However, using a conservative
indicator of significance in this dissertation (p<.01), civic
participation was not significant in the current model. As discussed
above, the results do not support a relationship between perceived
safety and physical activity.
75


Section 4.6 Specific Aim #2
The purpose of Specific Aim #2 was to test the relationships
between social capital, built environment, and physical activity and
BMI in the study neighborhoods.
Research questions of interest: Is there a relationship between
perceived social capital, objective social capital indicators, and
physical activity? Does the relationship vary based on objective
versus subjective measures of social capital? Does the walkability
of a neighborhood influence physical activity?
Models 5-7: Objective social capital and outcomes
Potential relationships between the outcomes and objective
indicators of social capital were tested using random effects, linear,
and logistic models. Tables 4.17-4.19 include results of unadjusted
and adjusted models for all dependent variables. Self-rated health
decreases as incivilities and indicators of territoriality increase.
Increased incivilities were associated with increased BMI. None of
the objective social capital indicators are significantly related to the
outcome of meeting recommendations for physical activity.
Models 8-10: Combined social capital variables and physical
activity
The next step was to determine if a difference existed between
physical activity and perceived versus objective indicators of social
capital, by testing a combined model. I used similar methods as
previously described for model building in this analysis. The
potential predictors and covariates entered the model using a
liberal selection criterion (p < 0.20). Iterations of the model
continue dropping non-significant predictors until a preliminary
main effects model remains. Table 4.20 and 4.21 lists the
parameter estimates, standard error, and p-values for final models.
76


To test the combined model, the same process and criteria
applied. The combined model used the full set of covariates. All
objective and perceived social capital variables entered the model.
For this model, covariates were allowed to drop based on
significance criteria. Table 4.22 shows the final combined model.
For Model 8, only social cohesion (p<.001) and health resources
(p<.01) remain significant. Those reporting higher perceptions of
social cohesion have greater than twice the odds of meeting
recommendations for physical activity. In Model 9, the only
objective social capital indicator that remains in the final model is
incivilities (p<.01). Similar to the earlier model, as incivilities
increase, meeting recommendations for physical activity
decreases. In this model, gender and age remain significant. Males
and younger respondents have more than 2.5 times the odds of
meeting the recommendations for physical activity than females
and older respondents. In the combined model (Model 10), only
perceived social cohesion and incivilities are significant. Those
who report higher perceptions of perceived social cohesion and
lower rated incivilities have 57% greater odds of meeting
recommendations for physical activity.
Neighborhood design and physical activity & BMI
Increased connectivity of blocks, higher density, and low block
length indicated a more walkable neighborhood.(110) For each
block, a walkability score was computed using these criteria. Block
groups were categorized as low, medium, or high.
There were significant differences in meeting physical activity
recommendations between block groups with low walkability
versus high walkability. Fifty-five percent of residents did not meet
the recommendations for physical activity in low walkable areas
versus 45% in medium and 45% high walkable areas (x2 (2)=8.249,
p<.01). Post hoc test indicate a significant difference between low
77


walkability and medium (x2(1) = 4.17,p< 02) and high walkability
(x2(1) = 7.45,p<.004). There was no difference in meeting physical
activity recommendations between medium and high walkable
areas (x2 = -003, NS). There were no significant differences in BMI
and walkability (x2(6) = 6.67, p<.344).
Summary Specific Aim #2
Research questions of interest: Is there a relationship between
perceived social capital, objective social capital indicators, and
physical activity? Does the relationship vary based on objective
versus subjective measures of social capital?
Hypothesis 2a. Participants living in neighborhoods with higher
walkability will have lower BMIs and higher rates of physical
activity.
The data support the hypothesis related to meeting
recommendations for physical activity, but not for BMI.
Neighborhoods with low walkability had fewer persons meeting
physical activity recommendations. However, higher BMIs were not
associated with low walkability. The study population, on average,
had high rates of physical activity compared to the national
average. This may attenuate any relationship the built environment
might have on walkability and BMI. Moderate to vigorous physical
activity, not walking, may be used to control or maintain weight.
The study neighborhoods were urban neighborhoods laid out on a
grid pattern. Therefore, the area is likely to rate higher in
walkability than if the sample had included suburban
neighborhoods.
Hypotheses 2b & c: Perceived social capital and objective social
capital will be associated with physical activity. Specifically,
perceptions of safety will have a positive association with physical
78


activity. Increases in indicators of territoriality and incivilities will
negatively influence physical activity. Perceived social capital will
be a stronger predictor than objective social capital of physical
activity.
Different from other studies, indicators of incivilities and territoriality
did not influence physical activity.(11;56;69;89;100;111) Again, this
may be due to the measurement used for physical activity.
Presently, physical activity is defined by moderate to vigorous
intensity. Much of the literature focuses on walking as an outcome,
which may be more influenced by these indicators than moderate
to vigorous physical activity.
For the objective indicators of social capital, incivilities were a
consistent predictor across self-rated health and BMI. As incivilities
increased, BMI increased and self-rated health decreased.
Incivilities include graffiti, litter, and trash, which affect the
aesthetics of a neighborhood. Other research has found that
aesthetics is an important aspect of the built environment that
encourages physical activity.(11 ;69) Indicators of territoriality
include window bars, security signs, neighborhood watch, and high
borders or fences. To residents of those blocks or neighborhoods,
these indicators may not be negative or even noticed but just part
of the scenery for the neighborhood. An individuals perception of
their safety may be more influential than passive indicators in their
environment. However, in the combined model neither safety nor
territoriality remains significant.
In the combined model, very few covariates remained significant.
In the combined model, territoriality and perceived safety are not
significant. These two predictors measure aspects of safety. First,
the individual residents from the study population rated the degree
to which they felt safe in their neighborhoods. Territoriality was an
indicator of safety as perceived by research staff who
79


independently audited the neighborhood. Neither territoriality nor
perceived safety remains in this model to predict physical activity
participation. In summary, the combined model indicates that
incivilities and perceptions of social cohesion are significant
predictors and associated with meeting recommendations for
physical activity.
80


Table 4.17. Model 5: Unadjusted and adjusted multi-level analysis for self-rated health.3
Unadjusted Adjusted1
Parameter Estimate SE P ICC Parameter Estimate SE P ICC
Objective Social Capital
Incivilities2 -3.85 .33 <.0001 0.03 -1.40 0.31 <.0001 0.0
Intercept 4.75 0.12 <.0001 4.31 0.15 <.0001
Territorality2 -4.18 0.85 <.0001 0.13 -1.45 0.44 0.002 0.00
Intercept 4.93 0.31 <.0001 4.41 0.19 <.0001
Alleyways2 -0.95 0.27 <.0001 0.15 -0.34 0.13 0.013 0.00
Intercept 3.80 0.12 0.007 4.05 0.14 <.0001
11ndividual level covariates controlled for: gender, age, ethnicity/race, income, education and employment status,
2 higher rating indicates higher number of incivilities, higher indicators of territoriality and poorer ratings of alleyways
Uncond ICC=0.20
81


Table 4,18. Model 6: Unadjusted and adjusted multi-level analysis for BMl.a
Unadjusted Adjusted1
Parameter Estimate SE P ICC Parameter Estimate SE P ICC
Objective Social Capital
Incivilities2 13.40 1.52 <.0001 0.00 8.07 2.03 0.0002 0.0
Intercept 22.59 0.50 <.0001 24.40 1.00 <.0001
Territorality2 12.77 3.92 0.002 0.05 5.74 2.84 0.048 0.00
Intercept 22.62 1.40 <.0001 24.78 1.23 <.0001
Alleyways2 3.80 1.13 0.001 0.05 1.74 0.88 0.05 0.21
Intercept 25.74 0.48 <.0001 26.07 0.88 <.0001
individual level covariates controlled for: gender, age, ethnicity/race, income, education and employment status.
2 higher rating indicates higher number of incivilities, higher indicators of territoriality and poorer ratings of alleyways
aUncond ICC=0.20
82


Table 4.19. Model 7: Unadjusted and adjusted multi-level analysis for physical activity.8
Unadjusted Adjusted1
Parameter Estimate SE P ICC OR Parameter Estimate SE P ICC OR
Objective Social Capital
Incivilities* -1.07 0.15 <.0001 0.02 0.00 -0.46 0.18 0.013 0.01 0.01
Intercept 0.86 0.05 <.0001 0.58 0.85 <.0001
Territorality* -1.03 0.34 0.004 0.07 0.00 -0.21 0.26 0.42 0.01 0.12
Intercept 0.87 0.12 <.0001 0.53 0.11 <.0001
Alleyways* 0.32 0.10 0.002 0.06 24.53 -0.10 0.08 0.19 0.01 0.36
Intercept 0.62 0.04 <.0001 0.49 0.07 <.0001
'individual level covariates controlled for: gender, age, ethnicity/race, income, education and employment status.
2 higher rating indicates higher number of incivilities, higher indicators of territoriality and poorer ratings of alleyways
"Uncond ICC=0.06
83


Table 4.20 Model 8: Coefficients for final model perceived social capital and physical activity.
Variable Parameter Estimate SE P OR
Intercept 0.22 0.10 0.04
Race/Ethnicity
Hispanic -0.12 0.06 0.03 0.30
African American -0.14 0.05 <0.001 0.25
Income
decline to answer -0.21 0.06 <0.001 0.12
Age
18-29 0.09 0.06 0.14 2.46
Perc. Social Capital Indicators
Social cohesion1 0.07 0.02 <0.001 2.01
Civic participation2 0.05 0.03 0.07 1.65
Health resources2 0.06 0.02 0.01 1.82
1 higher rating indicates higher perceptions of social cohesion
2lower score indicates low/no civic participation or use of community groups as a health resources
aICC=0.04
84


Table 4.21 Model 9: Coefficients for final model objective social capital and physical activity.
Variable Parameter SE p OR
Estimate __________________
Intercept 0.62 0.07 <.0001
Gender 0.10 0.03 0.00 2.72
Race/Ethnicity
Other 0.10 0.07 0.14 2.72
Income
Low -0.09 0.05 0.04 0.41
Decline to answer -0.30 0.05 <.0001 0.05
Age
18-29 0.11 0.04 0.01 3.00
Educational Attainment
HS diploma or less -0.12 0.06 0.05 0.30
Some college -0.12 0.05 0.02 0.30
College degree -0.11 0.05 0.02 0.33
Employment Status
Unemployed -0.15 0.09 0.08 0.22
Obj. Social Capital Indicators
Incivilities1 -0.49 0.17 0.00 0.01
1 higher rating indicates higher number of incivilities
"Uncond IC00.01
85