-
extracted text
-
DISTRIBUTION OF URBAN HOUSEHOLD
CARBON DIOXIDE EMISSIONS
ALONG A SOCIECONOMIC GRADIENT OF INDIANAPOLIS, IN
by
Kara
Karboski
A
thesis
submitted
in
partial
fulfillment
of
the
requirements
for
the
degree
Master
of
Environmental
Studies
The
Evergreen
State
College
June
2013
©
2013
by
Kara
Karboski.
All
rights
reserved.
This
Thesis
for
the
Master
of
Environmental
Studies
Degree
by
Kara
Karboski
has
been
approved
for
The
Evergreen
State
College
by
___________________________
Erin
Ellis
Member
of
the
Faculty
___________________________
Date
Abstract
Distribution
of
Urban
Household
Carbon
Dioxide
Emissions
along
a
Socioeconomic
Gradient
of
Indianapolis,
IN
Kara
Karboski
As
cities
tackle
climate
change
mitigation,
filling
the
gap
left
by
failures
in
international
agreements,
a
demand
for
more
information
and
data
on
the
dynamics
of
urban
carbon
dioxide
(CO2)
emissions
has
been
created.
Up
to
this
point,
past
research
of
the
drivers
of
anthropogenic
CO2
emissions
have
failed
to
examine
the
dynamics
of
urban
CO2
emissions
at
the
neighborhood
scale
for
an
entire
city.
At
this
scale
the
effects
of
historical,
cultural,
and
structural
forces
that
shape
housing
distribution
of
the
city
become
visible.
This
type
of
information
could
improve
strategies
and
targets
for
local
climate
change
mitigation
policy.
Accordingly,
the
objective
of
this
study
was
to
examine
the
spatial
distribution
and
socioeconomic
drivers
of
urban
household
CO2
emissions.
This
research
used
household
CO2
emission
data,
residence
and
transportation
emissions,
from
the
Hestia
Project
and
the
Center
for
Neighborhood
Technology
to
perform
a
spatial
analysis
of
the
socioeconomic
drivers
of
CO2
emissions
at
the
census
tract
scale
for
the
city
of
Indianapolis,
IN.
A
spatial
lag
regression
was
employed
to
control
for
influences
from
interactions
and
externalities
associated
with
neighboring
census
tracts.
The
results
show
the
model
explained
a
large
portion
of
household
CO2
emissions,
with
spatial
influences
exhibiting
a
strong
influence.
Income
was
found
to
be
a
strong
predictor
of
household
CO2
emissions
(β
=
-‐0.46,
p
<
.001).
Race
and
ethnicity
of
households
for
both
black
households
(β
=
0.11,
p
<
.001)
and
Asian
households
(β
=
-‐0.11,
p
<
.001),
while
significant,
were
found
to
be
weak
predictors
of
emissions.
This
study
concludes
that
there
is
significant
variability
in
household
CO2
emissions
across
the
urban
space
due
in
large
part
to
the
variability
and
distribution
of
socioeconomic
factors.
This
type
of
information
should
be
integrated
into
local
climate
change
policy
to
improve
strategies
to
mitigation.
Table of Contents
List of Tables -------------------------------------------- vi
List of Figures -------------------------------------------- vii
Acknowledgements ------------------------------------- viii
Chapter 1: Introduction ----------------------------------- 1
1.1
Introduction
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
1
1.2
Background
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
3
1.3
Structure
of
Thesis
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
5
Chapter 2: Literature Review ------------------------------ 7
2.1
Introduction
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
7
2.2
Climate
Change
and
the
Carbon
Cycle
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
9
2.3
City
Policy
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
10
2.4
Anthropogenic
Emissions
and
Greenhouse
Gas
Inventories
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
12
2.5
Physical
Drivers
of
Carbon
Emissions
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
17
2.6
Socioeconomic
Drivers
of
Carbon
Emissions
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
20
2.5.1
Income
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
20
2.5.2
Race
and
ethnicity
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
23
2.6
Conclusion
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
26
Chapter 3: Methods ------------------------------------- 28
3.1
Aim
and
Objectives
of
Research
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
28
3.2
Study
Area
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
29
3.3
Data
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
31
3.3.1
Variables
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
31
3.3.1.1
Residence
CO2
Emission
Data
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
31
3.3.1.2
Transportation
CO2
Emissions
Data
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
32
3.3.1.3
Spatial
Data
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
33
3.3.2
Data
Preparations
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
33
3.4
Limitations
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
34
3.5
Statistical
Analysis
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
36
Chapter 4: Results -------------------------------------- 40
4.1
Descriptive
Statistics
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
40
4.2
Statistical
Analysis
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
45
Chapter 5: Discussion ----------------------------------- 52
5.1
Spatial
Variables
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
53
5.2
Income
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
55
5.3
Race
and
ethnicity
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
57
iv
5.4
Other
variables
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
59
5.5
Policy
implications
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
61
Chapter 6: Conclusion ---------------------------------- 64
6.1
Conclusion
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
64
6.2
Interdisciplinary
Statement
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
66
6.4
Recommendations
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
67
Bibliography -------------------------------------------- 69
Appendix A --------------------------------------------- 75
Appendix B ---------------------------------------------- 77
Appendix C --------------------------------------------- 80
v
List of Tables
Table
1:
Five
number
summary
of
independent
and
dependent
variables.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
40
Table
2:
Moran's
I
test
for
spatial
dependency
among
the
residuals
of
the
regression
models.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
46
Table
3:
OLS
and
spatial
lag
model
regression
results.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
51
Table
4:
Moran's
I
test
for
spatial
dependency
among
CO2
variables.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
75
vi
List of Figures
Figure
1:
Inset
map
of
Marion
County
–
Indianapolis
in
the
state
of
Indiana.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
30
Figure
2:
Average
household
transportation
CO2
per
census
tract
in
Indianapolis,
IN.
-‐-‐-‐-‐-‐
41
Figure
3:
Average
household
residence
CO2
per
census
tract
in
Indianapolis,
IN.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
41
Figure
4:
Average
total
household
CO2
per
census
tract
in
Indianapolis,
IN
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
42
Figure
5:
Percentage
of
Asian
households
per
census
tract
in
Indianapolis,
IN.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
44
Figure
6:
Percentage
of
black
households
per
census
tract
in
Indianapolis,
IN.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
44
Figure
7:
Median
household
income
per
census
tract
of
Indianapolis,
IN.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
45
Figure
8:
Local
Moran
on
residuals
of
the
total
household
CO2
OLS
regression
model.
-‐-‐-‐-‐-‐
47
Figure
9:
Local
Moran
on
residuals
of
total
household
CO2
spatial
lag
regression
model.
-‐-‐
47
Figure
10:
The
regression
plot
of
total
household
CO2
emissions
of
the
predicted
model
by
the
observed
emissions
for
the
spatial
lag
regression.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
49
Figure
11:
Local
Moran’s
I
on
total
household
CO2
emissions.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
75
Figure
12:
Local
Moran’s
I
on
transportation
CO2
emissions.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
76
Figure
13:
Local
Moran’s
I
on
residence
CO2
emissions.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
76
Figure
14:
Local
Moran’s
I
on
residuals
of
residence
CO2
OLS
regression
model.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
77
Figure
15:
Local
Moran’s
I
on
residuals
of
residence
CO2
spatial
lag
regression
model.
-‐-‐-‐-‐
78
Figure
16:
Local
Moran’s
I
on
residuals
of
transportation
CO2
OLS
regression
model.
-‐-‐-‐-‐-‐-‐
78
Figure
17:
Local
Moran’s
I
on
residuals
of
transportation
CO2
spatial
lag
regression
model..
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
79
Figure
18:
The
regression
plot
of
transportation
CO2
emissions
of
the
predicted
model
by
the
observed
emissions
for
the
spatial
lag
regression.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
80
Figure
19:
The
regression
plot
of
residence
CO2
emissions
of
the
predicted
model
by
the
observed
emissions
for
the
spatial
lag
regression.
-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐-‐
80
vii
Acknowledgements
This
work
I
set
out
on
extends
far
beyond
the
reaches
of
a
single
institution,
far
beyond
the
relatively
short
time
I’ve
spent
in
the
MES
program.
Likewise,
this
thesis
marks
a
culmination
of
knowledge
and
experience
from
different
places
and
people,
past
and
present.
To
all
of
you
who
helped
me
get
here,
thank
you.
A
giant
thank
you
to
my
thesis
reader
and
MES
faculty
member
Dr.
Erin
Ellis,
a
source
of
encouragement
and
support.
Your
high
expectations,
but
also
belief,
pushed
me
to
produce
my
best
work.
I
would
also
like
to
thank
the
rest
of
the
MES
faculty,
especially
those
I’ve
had
the
great
pleasure
of
learning
and
working
with.
I
would
like
to
acknowledge
the
Hestia
Project
based
at
Arizona
State
University
and
the
Center
for
Neighborhood
technology
for
providing
me
with
the
data
to
make
this
thesis
a
reality.
Finally,
many
deserved
and
inexpressible
thanks
to
my
family,
both
by
blood
and
by
cohort.
viii
Chapter 1: Introduction
1.1 Introduction
The
urban
space
has
become
the
new
battleground
for
climate
change
mitigation.
Undeterred
by
the
lack
of
agreement
and
response
at
the
international
and
national
level,
progressive
municipalities
across
the
United
States
are
drafting
climate
change
policy,
focusing
their
attention
on
the
reduction
of
carbon
dioxide
(CO2)
emissions,
the
largest
anthropogenic
contributor
to
climate
change.
Because
cities
are
committing
the
time
and
financial
resources
to
these
types
of
policies,
an
accurate
understanding
of
the
patterns
and
drivers
of
CO2
emissions
within
cities
is
essential.
Urban
form
is
not
only
restricted
to
the
physical
and
technical
aspects
of
the
city;
important
structural,
cultural,
and
historical
factors
at
work
influence
where
CO2
is
generated
within
the
city.
Consequently,
recognizing
and
analyzing
these
spatial
patterns
is
important
for
potential
integration
into
city
climate
policy.
Information
of
this
type
can
be
used
to
inform
and
improve
policy
targeted
at
reducing
emissions
as
well
as
contribute
to
the
knowledge
of
the
urban
drivers
of
climate
change.
Previous
research
analyzing
the
relationship
between
CO2
emissions
and
cities
has
consistently
found
socioeconomic
factors
as
drivers
of
CO2,
such
as
increasing
CO2
emissions
with
increasing
income.
Most
of
these
studies
have
focused
on
city-‐level
emissions,
carrying
out
their
analysis
by
comparing
CO2
emissions
between
cities.
While
this
reveals
important
information,
the
dynamics
of
sub-‐city-‐level
CO2
emissions
deserve
attention
as
well.
This
gap
can
be
contributed
to
data
constraints;
neighborhood-‐level
1
analysis
requires
fine-‐scale
data
that
is
difficult
and
costly
to
obtain.
Those
few
studies
that
have
examined
CO2
emissions
at
these
smaller
scales
have
had
limited
success
because
their
level
of
analysis
was
too
large
to
perceive
important
spatial
patterns
that
exist
at
the
sub-‐city
level.
This
thesis
represents
another
step
in
examining
fine-‐scale
urban
CO2
emissions
and
seeks
to
bridge
the
gap
in
current
research
by
utilizing
small
scale
CO2
data
at
the
census
tract
level
to
appropriately
model
neighborhood
spatial
dynamics.
Using
data
for
the
city
of
Indianapolis,
IN,
the
research
presented
in
this
thesis
analyzes
the
distribution
of
household
CO2
emissions
across
the
urban
space
as
they
relate
to
socioeconomic
factors
that
influence
where
people
live
by
utilizing
spatial
analyses.
By
specifically
accounting
for
the
influence
of
neighbors,
that
is
the
influence
of
connections
across
space,
this
analysis
attempts
to
develop
an
accurate
and
predictable
model
of
urban
CO2
emissions.
Socioeconomic
factors,
most
importantly
income
and
race
for
this
analysis,
can
drive
particular
populations
into
certain
parts
of
the
city.
As
the
city
is
not
a
homogenous
space,
it
is
predicted
that
the
underlying
structure
of
the
city,
both
physically
and
socially,
can
be
taken
into
account
to
accurately
identify
relationships
between
CO2
emissions
and
demographic
factors.
It
is
hypothesized
that
CO2
emissions
will
be
positively
associated
with
household
income,
that
there
will
be
a
significant
association
with
household
race,
and
variables
of
space
will
explain
significant
part
of
the
creation
of
household
CO2
emissions.
2
1.2 Background
Modern
global
climate
change
refers
to
the
change
in
global
temperatures
caused
by
increasing
concentrations
of
greenhouse
gases,
including
carbon
dioxide,
methane,
nitrous
oxide,
and
fluorinated
gases,
in
Earth’s
atmosphere
as
well
as
by
other
anthropogenic
forcings
such
as
changes
in
land
use.
These
greenhouse
gases
play
a
vital
role
in
regulating
a
habitable
temperature
on
Earth
by
trapping
radiation
energy
sent
by
the
Sun
between
Earth’s
surface
and
this
layer
of
greenhouse
gases
in
the
atmosphere
(IPCC,
2007).
Quantitatively,
the
most
important
anthropogenic
greenhouse
gas
is
carbon
dioxide
(CO2).
Over
the
last
few
centuries
we
have
primarily
derived
our
energy
for
work
through
fossil
fuel
sources
such
as
coal,
oil,
and
natural
gas.
These
energy
dense
substances
are
composed
of
carbon
and
when
combusted
release
CO2.
Research
has
shown
that
since
humans
began
combusting
these
materials
in
increasingly
larger
amounts
beginning
at
the
industrial
revolution,
the
CO2
concentrations
in
the
atmosphere
have
increased
as
well,
intensifying
the
power
of
the
greenhouse
effect
and
increasing
global
temperatures
(IPCC,
2007).
An
increase
in
global
temperatures
would
have,
and
is
currently
having,
a
profound
affect
on
the
natural
and
built
environments
humans
have
adapted
to.
Effects
include
shifting
habitats,
species
extinction,
increased
drought,
changing
precipitation
patterns,
and
rising
sea
levels
(IPCC,
2007).
All
these
biophysical
changes
will
have
extensive
impacts
on
human
systems.
There
is
little
incentive
for
individual
actors
to
take
action
on
expensive
climate
change
mitigation
when
their
own
efforts,
by
themselves,
will
do
little
to
actually
mitigate
the
impacts
of
climate
change.
The
burden
of
the
costs
of
mitigation
rests
fully
on
those
3
who
take
mitigation
actions,
while
the
benefits
accrue
to
everyone,
even
those
who
do
not
act,
thanks
to
the
diffuse
nature
of
CO2
emissions.
Thus
the
benefits
become
diluted
for
those
who
took
mitigation
actions,
and
accordingly,
costs
of
mitigation
action
exceed
the
benefits
of
these
actions.
This
is
what
economics
terms
the
tragedy
of
the
commons.
Unless
there
is
agreement
among
all
parties
involved,
agreement
that
isn’t
presently
occurring,
then
this
economic
conundrum
will
continue.
This
is
why
climate
change
mitigation
has
long
been
thought
to
be
the
realm
of
international
discussion
and
negotiation,
as
these
are
large
enough
actors
to
cover
the
entire
global
population.
However,
cities
have
been
taking
action
regardless
of
these
perceived
barriers.
Local
climate
policy
has
been
of
interest
to
researchers
since
its
inception.
While
these
policies
may
be
successful
at
meeting
goals
at
the
city
level,
they
are
ultimately
hindered
in
realizing
substantial
reduction
in
global
CO2
emissions.
Several
important
factors
keep
local
reductions
from
providing
substantial
reductions
in
global
CO2
emissions.
First,
local
climate
change
mitigation
policy
is
still
in
its
infancy
and
the
coverage
attributed
to
these
policies
is
limited,
and
thus
emission
reductions
are
on
an
extremely
limited
scale.
Although
several
major
cities
have
enacted
CO2
policies,
it
is
unclear
how
many
people
are
falling
under
these
policy
umbrellas.
Second,
these
policies
may
not
call
for
large
cuts
in
emissions,
or
there
may
not
be
hard
incentives
to
reduce
emissions
in
the
first
place,
limiting
their
efficacy.
Finally,
as
mentioned
before,
there
seems
to
be,
at
first
glance,
no
to
little
economic
incentive
for
cities
to
partake
in
climate
change
mitigation
policies.
This
is
not
to
fault
cities.
Indeed
their
decision
to
take
up
any
type
of
climate
change
policy
is
surprising,
but
important.
Perhaps
the
initiative
being
taken
can
4
spur
the
global
scale
adoption
of
local
policies
or
perhaps
help
push
an
international
agreement
For
cities
to
draft
effective
climate
change
mitigation
policy,
they
require
data
on
the
greenhouse
gases
emitted
from
their
jurisdiction.
Additional
information
on
the
nature
of
these
emissions,
the
local
factors
that
drive
these
emissions,
could
help
aid
policy
development.
Furthermore,
modeling
the
variables
that
influence
urban
CO2
emissions
will
contribute
to
our
general
knowledge
of
the
drivers
of
climate
change,
and
we
may
be
better
able
to
project
future
emissions
scenarios.
Thus,
it
is
important
to
understand
the
specific
dynamics
of
urban
CO2
emissions.
1.3 Structure of Thesis
There
are
six
chapters
presented
in
this
thesis:
Chapter
1:
Introduction
Chapter
2:
Literature
Review
Outlines
the
relevant
and
previous
research
relevant
to
this
thesis
beginning
with
local
climate
change
policy,
and
moving
on
to
household
determinants
of
energy
production
and
the
factors
that
drive
CO2
emissions.
Special
focus
is
given
to
socioeconomic
factors
of
CO2.
Chapter
3:
Methods
5
Establishes
the
main
research
goals
and
the
primary
methodology
and
statistical
analysis
employed
within
this
study.
Chapter
4:
Results
The
results
from
the
analysis
are
presented.
Chapter
5:
Discussion
Discusses
the
findings
from
the
results
and
the
implications
of
these
results
on
climate
change
research
and
climate
change
mitigation
policy.
Chapter
6:
Conclusion
Final
conclusions
are
presented.
An
interdisciplinary
statement
discusses
the
importance
of
multiple
disciplines
to
the
analysis
conducted
in
this
paper.
This
paper
ends
with
further
recommendations
for
study.
6
Chapter 2: Literature Review
2.1 Introduction
Climate
change
is
a
global
issue
encompassing
complex
natural
and
human
systems
and
has
thus
necessitated
a
large
body
of
research
outside
of
the
physical
sciences.
Unifying
under
the
single
issue
of
climate
change
has
led
to
research
across
and
between
a
breadth
of
disciplines
ranging
from
economics
and
political
science,
to
conservation
biology
and
international
development.
Interest
has
grown
over
the
last
few
years
regarding
the
relationship,
one
of
considerable
give
and
take,
between
cities
and
climate
change.
Extensive
literature
is
devoted
to
examining
these
particular
dynamics,
where
cities
act
as
substantive
contributors
to
climate
in
addition
to
being
areas
of
particular
concern
for
climate
change
impacts.
Of
immediate
interest
for
this
study
is
the
literature
focused
on
cities’
contributions
to
climate
change
and
the
drivers
of
climate
change,
with
an
emphasis
on
the
socioeconomic
drivers
of
household
CO2
emissions,
as
well
as
spatial
analysis
of
CO2
emissions
between
and
within
cities.
The
spotlight
on
cities
has
magnified
in
part
because
of
continuing
urbanization:
projections
indicate
approximately
67%
of
the
global
population
will
reside
in
urban
areas
by
2050,
up
from
52%
in
2010,
with
an
already
heavily
urbanized
United
States
seeing
a
relatively
smaller
change
from
82%
to
89%
in
that
same
time
period
(United
Nations,
2012).
This
is
of
particular
interest
given
the
amount
of
anthropogenic
CO2
emissions
that
currently
come
from
cities,
a
number
that
is,
unfortunately,
difficult
to
pin
down.
Exactly
how
much
of
current
global
CO2
emissions
can
be
allocated
to
urban
areas
is
still
debated,
7
the
technical
aspects
of
which
will
be
covered
further
in
this
review.
Perhaps
30-‐40%
of
global
greenhouse
gas
emissions
(of
which
CO2
is
only
one
such
gas)
can
be
attributed
to
cities
(Hoornweg,
Sugar,
&
Gómez,
2011),
while
estimates
of
energy
related
CO2
emissions
assert
that
approximately
71%
come
from
cities,
a
discrepancy
due
to
the
inordinate
amount
of
land
change
related
emissions
that
occur
outside
of
cities
(International
Energy
Agency,
2008).
Owing
to
the
larger
proportion
of
urbanization
in
the
U.S.
over
global
averages
(United
Nations
Department
of
Economic
and
Social
Affairs,
Population
Divison,
2012),
it
can
be
reasonably
assumed
that
the
amount
of
CO2
emissions
from
U.S.
cities
is
higher
than
the
global
estimate.
This
leads
to
a
substantial
amount
of
anthropogenic
CO2
emissions
falling
under
local
jurisdictions.
As
urban
populations
increase,
the
importance
of
cities
as
concentrations
of
economy,
policy,
and
culture
is
maintained.
These
trends
place
cities
in
the
unique
position
of
addressing
climate
change
impacts
and
adaptations
at
the
local
level.
Indeed,
due
to
inaction
by
national
governments,
many
cities
have
taken
the
initiative
to
address
climate
change
issues
by
implementing
policies
to
lower
greenhouse
gas
emissions
in
their
jurisdictions.
Thus
the
relationship
between
cities,
carbon
dioxide
(CO2)
emissions,
and
climate
change
has
become
even
more
essential
to
addressing
this
complex
issue.
This
literature
review
will
first
briefly
discuss
the
current
knowledge
of
climate
change
and
the
global
carbon
cycle.
An
examination
of
cities’
responses
to
climate
change
through
mitigation
policies
will
then
follow,
specifically
looking
at
greenhouse
gas
inventories
and
other
CO2
data
and
its
integration
into
the
decision-‐making
process.
I
will
then
examine
the
research
on
anthropogenic
carbon
emissions
and
its
related
energy
8
literature.
In
particular,
there
will
be
focus
on
the
household
sector,
made
up
of
residences
and
their
associated
transportation,
and
the
physical
characteristics
and
environmental
factors
that
drive
emissions
in
these
areas.
Finally,
there
will
be
a
discussion
of
the
human
drivers
of
household
CO2
emissions
with
an
emphasis
on
income,
race,
and
ethnicity.
This
review
will
show
that
there
are
significant
gaps
in
the
analysis
of
the
socioeconomic
drivers
of
CO2,
both
in
content
and
scale.
A
spatial
analysis
of
a
single
urban
system
could
potentially
take
into
account
the
sociological
and
economic
variables
of
urban
form,
and
could
better
inform
policy,
climate
scenarios,
and
our
understanding
of
urban
dynamics.
2.2 Climate Change and the Carbon Cycle
Climate
change
in
this
context
refers
to
the
modern
change
in
Earth’s
climate
brought
on
by
increasing
concentrations
of
atmospheric
gases,
termed
greenhouse
gases,
responsible
for
regulating
Earth’s
habitable
temperature.
As
the
main
driver
of
anthropogenic
global
climate
change,
CO2
emissions,
and
by
relation
carbon
itself,
have
received
significant
attention.
Carbon
cycling
research
examines
the
movement,
fluxes,
and
sinks
of
carbon
through
Earth’s
reservoirs:
the
atmosphere,
ocean,
and
land.
Anthropogenic
carbon,
from
fossil
fuel
combustion
and
land
change,
is
the
third
largest
carbon
flux
to
the
atmosphere.
Current
estimates
put
this
flux
at
9.9±.0.9
Pg
of
carbon
per
year
(1
petagram,
or
Pg,
equals
109
metric
tonnes),
8.7±0.5
Pg
of
which
can
be
attributed
to
fossil
fuel
combustion,
with
the
remainder
due
to
land-‐use
change
(Le
Quéré
et
al.,
2009).
This
is
a
significant
alteration
to
the
global
carbon
cycle.
Furthermore,
9
although
carbon
sinks,
such
as
the
oceans
and
land,
have
absorbed
a
significant
amount
of
this
carbon,
it
is
speculated
that
their
ability
to
do
so
will
weaken
in
the
future
(Canadell
et
al.,
2007).
Recognizing
urban
areas
as
important
sources
of
carbon
has
led
recent
in
integrating
urban
dynamics
–
the
political,
social,
and
physical
complexities
of
cities
–
into
carbon
cycle
modeling
(Churkina,
2008;
Pataki
et
al.,
2006).
This
requires
further
understanding
of
the
urban
drivers
of
anthropogenic
CO2
emissions
as
well
as
more
consistent,
detailed,
and
available
CO2
data.
2.3 City Policy
Reductions
in
CO2
emissions
at
the
individual
city
level
are
too
small
to
affect
global
CO2
concentrations,
so
without
higher
level
policy
from
national
governments
there
seems
to
be
little
incentive
for
local
action
(e.g.
Betsill
&
Bulkeley,
2007;
Betsill,
2001;
Engel
&
Orbach,
2008).
However,
cities
are
taking
the
initiative
on
climate
change
mitigation
regardless,
connecting
a
global
issue
with
local
issues.
A
large
body
of
research
has
developed
around
local
motivations,
analyzing
why
cities
are
taking
action,
the
barriers
and
processes
to
these
actions,
and
analysis
of
the
actions
themselves
(Betsill
&
Bulkeley,
2007).
Motivational
studies
have
found
government
officials
and
planners
very
often
link
climate
issues
to
cost-‐saving
ventures
such
as
energy
and
building
efficiency
programs,
promote
the
environmental
benefits
of
these
policies,
and
in
general
are
responsive
to
both
10
larger-‐scale
national
and
state
pressures,
as
well
as
from
local
citizenry,
groups,
and
businesses
(Engel
&
Orbach,
2008;
Kousky
&
Schneider,
2003;
Sharp
et
al.,
2011;
Sippel
&
Jenssen,
2009).
Very
often,
climate
change
mitigation
benefits
are
linked
to
other
benefits
such
as
these.
Cities
wanting
to
explicitly
address
climate
change
mitigation
often
draft
climate
change
plans,
a
consolidation
of
actionable
policies
organized
around
the
goal
of
reducing
local
CO2
emissions
(Boswell
et
al.,
2010;
Krause,
2011b;
Tang
et
al.,
2010).
These
plans
are
often
explicitly
linked
to
local
greenhouse
gas
inventories.
There
is
a
large
range
of
actions
that
cities
have
taken
to
address
climate
change;
Krause
(2011b)
identifies
several
types:
policies
based
on
enabling
as
through
positive
incentives,
policies
based
on
authority
as
through
regulation
or
negative
incentives,
by
providing
services
that
influence
wanted
behavior,
and
finally
policies
directed
toward
municipal
operations.
Examples
of
the
types
of
policies
that
can
be
tied
to
climate
change
include
changing
building
efficiency
codes,
changing
their
solid
waste
programs,
and
creating
energy
efficient
infrastructure
changes
to
city
property
and
projects
(Betsill,
2001;
Krause,
2011a).
While
many
cities
have
formalized
climate
action
plans,
and
have
taken
actions
in
line
with
overarching
policy,
Krause
(2011a)
notes
that
many
municipalities
are
taking
actions
that
reduce
greenhouse
gases
without
being
implicitly
involved
in
a
larger
climate
action
plan.
Without
a
formalized
policy
framework
however,
and
the
reliable
and
accurate
information
to
drive
that
policy,
it
would
be
ineffective
for
cities
to
direct
policies
and
assess
benefits
and
progress
tied
to
climate
change.
Because
of
this,
many
local,
state,
and
national
governments,
businesses,
and
individuals
track
their
greenhouse
gas
emissions,
11
often
utilizing
the
help
of
already
established
frameworks,
such
as
the
framework
established
by
Local
Governments
for
Sustainability
(to
be
discussed
below)
(Eggleston
et
al.,
2006;
ICLEI,
2012).
2.4 Anthropogenic Emissions and Greenhouse Gas Inventories
A
greenhouse
gas
inventory
tracks
the
flux
of
greenhouse
gases,
most
notably
CO2,
but
also
methane,
nitrous
oxide,
and
fluorinated
gases
(reported
in
their
CO2
equivalents),
for
a
determined
geographic
area,
such
as
a
city
or
state,
over
a
particular
time
period,
often
a
year.
Both
sources
and
sinks
of
CO2
can
be
inventoried.
Emissions
may
be
reported
by
the
particular
sector
that
they
originated
from,
such
as
transportation,
waste,
residential,
industrial,
or
municipal
operations.
Data
from
these
inventories
can
be
used
for
target
setting,
evaluation
of
policies,
projection
of
future
emissions,
and
other
analyses.
Greenhouse
gas
inventories
have
been
used
in
international
agreements
such
as
the
Kyoto
Protocol
and
the
Copenhagen
Accord,
in
international
and
national
agreements
between
cities,
such
as
Local
Governments
for
Sustainability
(ICLEI)
and
the
Mayors
Climate
Protection
Agreement
(MCPA),
and
by
individual
nations,
states,
cities
and
researchers.
The
United
States,
in
accordance
with
the
United
Nations
Framework
Convention
on
Climate
Change,
has
been
conducting
yearly
national
greenhouse
gas
inventories
since
1990.
Emissions
are
tracked
by
type
of
greenhouse
gas,
sector
of
emission,
and
includes
sources
as
well
as
sinks
(US
EPA,
n.d.).
Energy
related
CO2
emissions
are
the
primary
source
of
greenhouse
gases
in
the
U.S.,
accounting
for
81%
of
emissions
in
2007
(U.S.
12
Department
of
Commerce
Economics
and
Statistics
Administration,
2010).
Of
these
energy
related
emissions,
household
emissions,
those
CO2
emissions
from
household
transportation
and
residence,
accounted
for
the
largest,
approximately
30%.
Household
emissions
are
defined
by
the
U.S.
Department
of
Commerce
as
those
energy
emissions
related
to
residential
fuel
and
electricity
use,
as
well
as
light-‐duty
vehicle
related
emissions,
approximated
to
be
transportation
associated
with
households.
Consequently,
research
has
focused
on
the
drivers
of
urban
and
household
CO2
emissions.
Local-‐level
municipalities
are
also
conducting
greenhouse
gas
inventories.
ICLEI
provides
support
and
resources
to
municipalities
who
join,
including
help
with
greenhouse
gas
inventories.
Both
the
MCPA
and
ICLEI
codify
reductions
in
emissions
for
those
municipalities
involved.
As
of
2010,
5%
of
U.S.
cities
were
part
of
either
the
MCPA
or
ICLEI,
covering
approximately
30%
of
the
U.S.
population
under
formal,
and
local,
climate
policy
(Rachel
M.
Krause,
2011).
Inventories
are
utilized
as
benchmarks
to
gauge
progress
and
effectiveness
of
greenhouse
gas
reducing
policies.
Extensive
amounts
of
data
are
required
to
understand
the
climate
system
and
its
interaction
with
human
and
natural
systems.
Greenhouse
gas
inventories
are
just
one
tool,
but
they
have
been
used
in
scientific
research
to
better
understand
both
the
biophysical
and
socioeconomic
driving
factors
of
climate
change,
as
well
as
to
project
future
emissions
and
inform
policy.
This
type
of
research
is
heavily
rooted
in
the
energy
and
energy
metabolism
literature,
in
engineering
and
in
economics.
The
connection
between
energy
consumption
and
CO2
emissions
is
well
established,
generating
interest
in
how
to
reduce
energy
consumption,
and
thus
CO2,
by
improving
the
efficiency
of
infrastructure.
In
this
13
vein,
research
has
examined
the
drivers
of
CO2
emissions
and
energy
consumption,
analyzing
the
physical
characteristics
of
the
built
and
natural
environment,
as
well
as
demographic
and
socioeconomic
factors.
The
initial
steps
in
these
analyses,
and
in
city
climate
action
plans,
is
an
accounting
of
all
the
emissions
associated
with
a
particular
place.
Greenhouse
gas
inventories
can
take
place
on
different
horizontal
and
vertical
scales.
The
first
greenhouse
gas
inventories
were
done
on
the
national
level.
Inventories
at
smaller
scales,
such
as
done
by
municipalities,
followed
shortly
after.
The
Intergovernmental
Panel
on
Climate
Change
(IPCC)
released
their
first
methodology
for
national
greenhouse
gas
inventories
in
1994.
Individual
states
have
also
conducted
greenhouse
gas
inventories
and
ICLEI
has
recently
released
their
inventory
procedures
for
cities
and
other
community
groups
and
organizations
(ICLEI,
2012).
Methodologies
vary
in
the
approach
to
measuring
CO2
emissions.
Some
methodologies
have
directly
measured
the
fluxes
of
CO2
as
they
vary
spatially
and
temporally
within
a
city
using
direct
measurements
of
CO2
(Wentz
et
al.,
2002).
Greenhouse
gas
inventories
however,
utilize
energy
and
consumption
data
as
proxy
measurements
of
CO2.
When
the
amount
of
fuel
used
is
known
we
can
then
get
a
measurement
of
CO2
emitted
for
that
particular
fuel.
The
amount
of
CO2
utilized
by
each
sector
–
generally,
industrial,
commercial,
residential,
and
transportation,
but
varies
by
inventory
needs
–
can
then
be
determined
(Glaeser
&
Kahn,
2010;
Golley
&
Meng,
2012;
Gurney
et
al.,
2012;
Kennedy
et
al.,
2010;
Ramaswami
et
al.,
2008;
VandeWeghe
&
Kennedy,
2007).
Some
approaches
also
try
to
integrate
the
emissions
associated
with
goods
14
produced
outside
the
inventory
boundary
area,
but
consumed
within
in
an
attempt
to
correctly
assign
those
emissions
to
those
who
are
creating
the
demand.
These
types
of
analyses
typically
use
the
household
as
the
level
of
analysis.
However,
this
is
a
very
data
intensive
process
and
so
complete
inventorying
utilizing
this
approach
has
been
limited.
Recent
developments
have
focused
on
several
key
areas
where
data
is
missing
or
is
insufficient
in
inventories.
County
and
city
inventories
are
relatively
recent
tools,
but
progress
in
the
past
few
years
has
centered
on
making
these
inventories
more
consistent
and
rigorous.
These
recent
guidelines
for
inventory
have
been
in
response
to
innovative,
but
inconsistent
community
level
greenhouse
gas
inventories
(S.
Kennedy
&
Sgouridis,
2011).
There
are
several
ways
to
account
for
greenhouse
gas
emissions,
and
determining
the
best
inventory
procedures
has
been
debated
extensively
in
the
literature.
Issues
arise
in
inventories
when
considering
scale,
vertically
or
horizontally,
when
determining
the
sectors
to
include,
and
how
to
assign
the
point
of
measurement
of
emissions
(C.
Kennedy
et
al.,
2009).
One
of
the
more
important
discussions
of
greenhouse
gas
inventories
concerns
the
latter
point,
where
CO2
emissions
are
counted:
the
source
of
their
production
(i.e.
the
point
of
combustion
of
fossil),
or
should
it
be
counted
at
consumption
(i.e.
at
the
final
source
of
the
demand).
The
choice
between
a
production
or
consumption
based
inventory
has
significant
impacts
on
finals
results
of
inventories
or
analyses.
Depending
which
is
used,
the
inventory
will
be
more
or
less
weighted
towards
municipalities,
states,
or
nations
that
have
large
discrepancies
in
trade
of
goods
or
electricity
(Dodman,
2009;
Larsen
&
Hertwich,
2009;
Peters
&
Hertwich,
2008;
Satterthwaite,
2008).
15
There
have
been
few
studies
attempting
to
spatially
aggregate
CO2
emissions
at
smaller
scales
than
the
city-‐scale.
Several
different
methods
of
modeling
emissions
have
been
developed.
Top-‐down
approaches
utilize
data
from
larger
scales
that,
through
various
modeling
techniques
and
established
variable
relationships,
is
projected
onto
smaller
scales.
Bottom-‐up
approaches
utilize
information
from
very
small-‐scales,
say
through
characteristics
of
known
building
types
or
through
actual
consumption
information
of
households,
to
extrapolate
those
values
to
larger
scales
(Swan
&
Ugursal,
2009).
The
Hestia
Project,
which
modeled
county
level
emission
data
generated
from
the
Vulcan
Project
to
the
building
level
to
create
spatial
and
temporal
visualizations
of
CO2
emissions
for
the
city
of
Indianapolis,
IN,
is
an
example
of
a
top-‐down
method,
utilizing
utility
data
to
verify
the
downscaled-‐model
(Gurney
et
al.,
2012).
In
a
different
study,
(VandeWeghe
&
Kennedy,
2007)
spatially
analyzed
the
distribution
of
CO2
by
census
tracts
for
the
Toronto
Census
Metropolitan
Area
as
a
way
to
assess
the
influence
of
urban
form
on
emissions.
They
utilized
energy
data
reported
at
the
census
tract
level
for
their
analysis
examining
variability
in
CO2
emissions.
However,
because
of
their
immense
and
intensive
data
requirements,
as
well
as
difficulties
getting
smaller-‐scale
data
from
electricity
providers,
city
greenhouse
gas
inventories
have
traditionally
been
scaled
to
the
entire
city.
Smaller
scale
data,
at
the
neighborhood
level
for
example,
could
provide
further
levels
of
analysis
and
elucidate
spatial
trends
and
patterns
across
the
city.
Paired
with
demographic
data
(see
Hillmer-‐
Pegram
et
al.
(2012)),
these
types
of
small-‐scale
inventories
could
potentially
help
tease
apart
important
information
and
help
guide
policy
decisions,
just
as
inventories
broken
down
by
sectors
do.
16
Once
completed,
greenhouse
gas
inventory
data
can
be
used
in
conjunction
with
other
types
of
data
to
analyze
trends
and
drivers
of
emissions.
Within
these
types
of
analyses,
there
are
several
key
areas
that
constitute
the
determinants
of
household
CO2
emissions:
the
physical
characteristics
of
buildings
and
urban
structure,
the
biophysical
characteristics
of
the
surrounding
environment,
family
or
household
structure,
and
socioeconomic
factors.
Many
studies
utilize
multiple
factor
types,
both
the
biophysical
and
socioeconomic
factors,
as
a
way
to
predict
emissions
and
create
comprehensive
models.
2.5 Physical Drivers of Carbon Emissions
A
substantial
portion
of
household
CO2
emissions
research
has
been
concerned
with
the
physical
factors
that
can
contribute
to
increased
energy
demand,
and
by
extension
CO2
emissions.
Specifically,
this
research
has
identified
factors
that
exist
in
the
physical
structure
and
construction
of
residences,
within
urban
form,
as
well
as
in
biophysical
aspects
of
climate.
Using
U.S.
urbanization
rates
we
could
predict
that
roughly
25%
of
energy-‐derived
CO2
emissions
in
the
U.S.
come
from
households
within
cities,
and
this
substantial
amount
of
CO2
has
driven
research
into
understanding
how
they
are
created
with
urban
areas.
The
research
on
urban
form
on
household
CO2
emissions
generally
examines
the
impacts
of
density.
While
some
explicitly
look
at
population
density
(Glaeser
&
Kahn,
2010;
Kaza,
2010)
others
examine
how
urban
form
impacts
transportation,
and
thus
CO2
emissions
(Brownstone
&
Golob,
2009;
VandeWeghe
&
Kennedy,
2007).
Density
is
also
17
examined
through
analyses
of
differences
between
rural
and
urban
households,
although
there
is
no
consensus
in
the
literature
as
to
the
relative
emissions
between
urban
and
rural
areas.
While
some
support
conclusions
that
cities
actually
produce
less
CO2
emissions
per
capita
than
in
other
areas
(Dodman,
2009;
Satterthwaite,
2008;
Wier
et
al.,
2001),
others
find
that
rural
areas
produce
less
CO2
per
capita
than
cities
(Heinonen
&
Junnila,
2011).
Possibilities
for
these
discrepancies
may
lie
in
differing
characteristics
between
countries
and
cities,
in
the
inherent
heterogeneity
that
exists
over
large
spaces
and
across
cultures,
factors
more
related
to
climate
or
the
socioeconomic
characteristics
of
a
particular
place
than
to
urban
form.
Another
possibility
for
the
discrepancies
may
be
found
in
how
the
CO2
emissions
are
determined:
consumption
versus
production
inventories.
(Heinonen
&
Junnila,
2011)
specifically
mention
this
problem
and
thus
utilize
a
consumption-‐based
approach.
The
differences
between
rural
and
urban
emissions
in
their
study
are
most
likely
due
to
income
differences
between
rural
and
urban
areas,
with
higher
levels
of
income
in
urban
areas
leading
to
higher
levels
of
consumption,
and
thus
energy
demand.
A
production-‐based
inventory
would
have
underreported
the
amount
of
emissions
for
high
consuming
urban
residents.
Generally,
density
and
urban
form
are
found
to
be
significant
contributors
to
household
level
CO2
emissions
from
both
residences
and
transportation
(Norman
et
al.,
2006)
Physical
housing
characteristics
are
also
important:
several
studies
examine
such
factors
as
age
of
the
residence,
housing
type
(single
family
versus
multi-‐family
units),
and
building
size
(Gurney
et
al.,
2012;
Kaza,
2010;
Min
et
al.,
2010;
Wier
et
al.,
2001).
These
characteristics
are
extensively
utilized
in
the
modeling
of
energy
demand
and
CO2
emissions
(Swan
&
Ugursal,
2009).
The
relationship
between
these
variables
and
CO2
18
emissions
is
complex.
For
example,
housing
size
generally
contributes
to
higher
CO2
emissions
as
there
is
more
space
to
regulate
temperature
and
this
requires
more
energy.
Newer
houses
tend
to
be
made
of
materials
that
are
better
at
regulating
internal
temperatures.
Add
in
other
housing
characteristics
and
it
quickly
becomes
complex
when
attempting
to
understand
how
the
interactions
between
these
drivers
affect
CO2
emissions.
At
larger
scales
there
is
interest
in
the
influence
of
different
fuel
types
and
climates
on
emissions.
Generally,
at
the
urban
scale
these
factors
play
a
significantly
smaller
role,
but
at
the
regional
or
country
level
this
can
explain
much
of
the
variation
in
emission
intensities.
Disparities
in
fuel
types
occur
because
of
differential
access
to
natural
resources
(Glaeser
&
Kahn,
2010;
C.
Kennedy
et
al.,
2009).
For
example,
the
Northwest
United
States
derives
much
of
its
electricity
from
hydropower,
which
is
generally
CO2
neutral,
versus
the
Midwest
which
is
more
reliant
on
coal
and
natural
gas,
very
CO2
intensive
fuels.
The
difference
between
comfortable
residence
temperatures
and
outside
temperatures
contributes
to
the
influence
of
weather
and
climate
directly
(Kriström,
2008).
Research
and
modeling
of
energy
demand
and
CO2
more
specifically
utilizes
both
the
amount
of
heating
or
cooling
required
for
a
household,
and
thus
energy
demand
at
both
the
larger
scale
of
cities
and
the
individual
household
level
(Glaeser
&
Kahn,
2010;
C.
Kennedy
et
al.,
2009;
Min
et
al.,
2010).
Overall
the
CO2
emission
and
energy
literature
has
created
various
methods
for
creating
CO2
data
at
multiple
scales
and
has
extensively
analyzed
these
data
sources
for
the
particular
factors
that
may
drive
emissions.
However,
there
is
not
always
consistency
in
the
19
analyses
of
these
drivers;
the
complexity
of
cities
and
household
energy
demand
contributes
to
much
heterogeneity
in
results.
Household
level
drivers
may
consist
of
more
than
physical
or
natural
factors.
There
are
particular
socioeconomic
factors
that
have
been
established
as
important
factors,
and
so
are
examined
in
addition
to
the
biophysical
drivers.
2.6 Socioeconomic Drivers of Carbon Emissions
The
socioeconomic
drivers
of
CO2
emissions
has
been
unevenly
studied,
although
historically
there
has
been
considerable
literature
devoted
to
residential
energy
demand
and
socioeonomics
(Kriström,
2008;
Lutzenhiser
&
Hackett,
1993).
As
briefly
mentioned
above,
the
relationship
between
income
and
CO2
emissions
has
been
fairly
well
established
on
multiple
scales.
However,
race
and
ethnicity
are
particular
social
factors
of
CO2
emissions
that
have
been
studied
only
in
limited
ways
(Estiri,
2013;
Min
et
al.,
2010).
2.5.1 Income
The
relationship
between
income
and
CO2
emissions
has
been
best
characterized
from
an
economic
perspective,
understandably
given
the
direct
connections
between
energy
demand,
CO2
emissions,
and
economic
development.
Considerable
analysis
has
centered
on
the
Environmental
Kuznets
Curve
(EKC),
a
hypothetical
relationship
between
economic
growth
and
environmental
degradation.
This
relationship
follows
an
inverse
U-‐
shape
whereby
economic
growth
stimulates
an
increase
in
total
environmental
degradation.
However,
due
to
decreasing
marginal
environmental
degradation,
the
curve
20
ultimately
reaches
a
peak.
At
this
point
as
an
economy
grows
total
environmental
degradation
begins
to
decrease.
The
argument
follows
that
as
countries
gain
affluence
they
put
a
higher
value
on
protecting
the
environment
and
thus
reach
the
point
where
action
will
be
taken
to
reduce
environmental
degradation.
The
hypothesis
presented
by
the
EKC
has
been
analyzed
in
terms
of
income
(as
economic
growth)
and
carbon
dioxide
emissions
(as
environmental
degradation).
It’s
hypothesized
that
as
income
increases
CO2
emissions
will
increase
up
to
a
point,
and
then
additional
income
will
induce
an
overall
decrease
in
total
CO2
emissions
as
there
is
a
recognition
among
the
populace,
or
rather
priorities
may
shift,
that
CO2
is
connected
to
climate
change
and
those
associated
environmental
impacts.
Several
analyses
have
controlled
for
potential
spatial
and
temporal
variation
complications.
However
the
relationship
between
income
and
CO2
emissions
is
not
consistent
and
other
variables
seem
to
have
significant
influence
on
CO2
emissions,
such
as
climate
and
fuel
type
(Aldy,
2005;
Burnett
&
Bergstrom,
2010).
Multiple
analyses
across
a
variety
of
countries
have
found
differing
results
on
the
applicability
of
CO2
emissions
to
EKC
(Aldy,
2005),
although
the
relationship
between
CO2
emissions
and
income
itself
have
been
consistently
positive.
Although
the
economics
literature
has
not
found
evidence
for
the
EKC
hypothesis
in
connection
with
CO2
emissions,
the
positive
relationship
between
CO2
and
income
is
fairly
well
established.
Energy
demand,
and
by
direct
connection
CO2
emissions,
is
derived
primarily
as
a
means
of
economic
development.
Rising
incomes
result
in
higher
demand
for
goods
and
services
and
the
energy
needed
to
provide
them.
21
At
the
global
level,
analyses
of
CO2
emissions
and
gross
domestic
product
(GDP)
reveal
a
positive
relationship
(Ramanathan,
2006;
Tucker,
1995).
Large
scale
climate
modeling
utilize
scenarios
as
key
inputs,
where,
these
projections
about
future
demographic
and
economic
trends
specifically
take
into
account
the
influence
of
GDP
as
a
measurement
of
economic
activity,
and
thus
a
driver
of
energy
demand
and
CO2
emissions.
The
IPCC
directly
addresses
the
relationship
between
GDP
and
CO2
emissions
in
their
Special
Report
on
Emissions
Scenarios
and
in
their
Assessment
Reports
(Metz,
2007;
Nakicenovic
et
al.,
2000).
In
analyzing
cities
from
multiple
countries,
(C.
Kennedy
et
al.,
2009)
found
that
income
was
positively
correlated
to
emissions.
Measures
of
income
are
consistently
included
in
analyses
of
household
CO2
emissions
(Estiri,
2013;
Golley
&
Meng,
2012;
Kaza,
2010;
Kerkhof
et
al.,
2009;
Min
et
al.,
2010).
This
empirically
makes
sense;
as
household
incomes
rises
the
demand
for
larger
residences
increase.
Additionally,
there
is
a
strong
correlation
between
increasing
incomes
and
movement
away
from
traditional
city
centers;
these
suburbs
increase
commute
times
and
thus
transportation
costs
and
associated
CO2
emissions
(Kahn,
2000).
Furthermore,
results
from
the
2001
National
Household
Travel
Survey
indicate
that
income
is
positively
associated
not
only
with
increased
number
of
trips
and
travel
distance,
but
also
in
the
number
of
vehicles
owned
(Pucher
&
Renne,
2003).
Finally,
the
poor
are
more
likely
to
utilize
public
transportation
lowering
their
emissions
(Glaeser
&
Kahn,
2010).
While
much
of
the
literature
on
household
consumption
and
CO2
emissions
does
take
into
consideration
socioeconomic,
demographic,
and
physical
housing
variables
as
mentioned
previously,
there
is
a
distinct
lack
of
analysis
across
the
spatial
area
of
a
city.
22
Spatial
dynamics
in
urban
structure
and
in
socioeconomic
and
institutional
forces
can
have
significant
influence
on
where
and
how
people
live.
2.5.2 Race and ethnicity
Sociology
and
urban
studies
have
already
devoted
extensive
resources
to
understanding
housing
and
transportation
inequality.
Three
determinants
of
housing
are
generally
considered:
housing
choice,
economic
decision-‐making,
and
institutional
factors.
Of
these
however,
more
weight
has
been
given
to
these
historical
and
modern
institutional
forces
that
have
segregated
particular
classes
and
races
of
people
in
American
cities.
While
outright
discrimination
is
outlawed
by
the
Fair
Housing
Act,
other
avenues
of
discrimination
that
are
more
difficult
to
uncover
and
address
still
exist
(Charles,
2003;
Roscigno,
Karafin,
&
Tester,
2009).
Situations
of
housing
discrimination
require
the
direct
action
of
the
affected
party,
and,
when
considering
the
often
marginalized
populations
involved,
many
of
whom
do
not
have
access
to
the
resources
or
knowledge
required
to
undertake
such
a
task,
cases
of
housing
discrimination
are
likely
underreported
(Roscigno
et
al.,
2009).
Particular
populations
are
more
likely
to
face
discrimination
than
others.
Of
the
most
affected
groups,
poor
blacks
have
been
subject
to
significant
and
extreme
historical
segregation
perpetuated
by
outright
discrimination
by
the
housing
sector.
From
this
historical
and
modern
spatialization
of
race
across
the
urban
space
there
emerges
the
possibility
of
a
link
between
race
and
ethnicity
and
household
CO2.
For
example,
the
movement
of
higher
income
and
predominantly
white
families
(Charles,
2003)
to
the
suburbs
is
associated
with
higher
levels
of
CO2
emissions
(Kahn,
2000).
In
analyses
of
residential
energy
demand,
the
literature
has
looked
at
energy
as
being
driven
23
by
consumption
of
goods
and
services,
ultimately
examining
behavior
as
the
driving
force.
Household
consumption
may
vary
by
income,
as
the
previous
section
can
attest,
but
also
by
class,
and
race
and
ethnicity
(Lutzenhiser,
1997).
Transportation
inequity,
closely
linked
to
housing
inequity,
has
different
primary
institutional
actors,
and
its
actions
are
much
less
direct.
For
instance,
the
flight
of
the
upper
classes,
predominantly
white,
to
the
suburbs
away
from
the
core
of
the
urban
areas
has
also
shifted
transportation
dollars
(Garrett
&
Taylor,
1999).
Funding
is
increasingly
being
spent
on
commuter
rails
and
other
types
of
transport
to
connect
suburbs
and
outlying
areas
to
jobs.
The
dollars
spent
per
person
is
much
higher
in
these
cases
than
for
inner-‐city
transit.
For
marginalized
populations
that
rely
heavily
on
public
transportation,
this
transfer
of
funding
can
have
important
and
lasting
affects,
essentially
isolating
these
communities
further
(Garrett
&
Taylor,
1999;
Pucher
&
Renne,
2003).
Predominately
poor
as
well,
access
to
personal
vehicles
may
not
be
possible.
(Kahn,
2000)
makes
a
connection
between
movement
from
inner
city
to
suburbs
as
driving
up
household
transportation
costs,
time,
and
distance,
and
thus,
CO2
emissions.
The
historically
established
movement
of
white
households
from
the
inner
cities
to
the
suburbs
is
tied
to
a
possible
relationship
between
race,
ethnicity,
and
household
transportation
CO2.
While
private
vehicle
use
does
not
vary
significantly
between
race
and
ethnic
groups,
minorities
are
more
likely
to
utilize
public
transit,
and
thus
decrease
their
transportation
CO2
emissions
(Glaeser
et
al.,
2008;
Pucher
&
Renne,
2003).
It
has
also
been
found
that
minorities
have
longer
commute
times
(Doyle
&
Taylor,
2000;
Shen,
2000).
These
two
factors
are
confounding
each
other,
making
it
difficult
to
synthesize
conclusions
related
to
24
increased
CO2
emissions
from
those
results;
public
transportation
both
increases
commute
times
and
reduces
CO2
emissions.
Research
that
disentangles
commute
times
and
public
transportation
usage
would
be
better
able
to
address
the
impacts
on
CO2.
In
a
similar
vein,
(Antipova,
Wang,
&
Wilmot,
2011)
analyzed
land
uses
and
socioeconomic
variables
as
they
relate
to
commuting
distances
and
times.
Minority
households
were
found
to
be
significantly
related
to
commuting
times,
but
no
significant
relationships
were
found
between
minorities
and
commuting
distance.
If
this
difference
in
time
is
caused
by
using
public
transportation,
than
this
would
indicate
a
negative
relationship
between
minority
households
and
CO2
emissions.
(Brownstone
&
Golob,
2009)
found
negative
relationships
between
race
and
ethnicity
variables
and
vehicle
fuel
consumption,
leading
support
to
the
hypothesis
that
race
and
ethnicity
leads
to
decreased
transportation
CO2
emissions.
Association,
positive
or
negative,
with
transportation
CO2
emissions
and
race
or
ethnicity
is
difficult
to
synthesize,
and
there
is
little
research
that
explicitly
examines
these
connections.
The
literature
on
transportation
inequity,
explicitly
tied
to
housing
inequity,
and
its
resulting
conclusions
of
less
vehicles
owned,
higher
usage
of
public
transportation,
and
less
fuel
usage
lends
support
to
differing
transportation
CO2
emissions
among
minorities.
Ultimately
transportation
is
connected
back
to
housing,
where
people
are
located
in
the
space
of
the
city.
While
studies
have
implicitly
understood
this
impact
of
space
by
understanding
the
affects
of
differing
energy
prices,
variable
climate,
and
general
housing
structure,
very
few
studies
have
looked
at
this
at
the
small
spatial
scale.
Without
any
direct
analysis
of
these
socioeconomic
factors,
income
(or
class)
and
race/ethnicity,
and
enacting
we
may
not
be
able
to
synthesize
a
complete
understanding
of
urban
CO2.
25
The
question
then
becomes,
how
does
income
and
housing
segregation,
or
the
spatial
consequences
of
these
variables,
affect
household
CO2
emissions?
The
relationships
between
these
variables
are
complex
and
difficult
to
tease
out.
On
one
hand,
lack
of
access
to
a
personal
automobile
or
increased
usage
of
public
transportation
would
result
in
less
CO2
emissions
related
to
transportation.
On
the
other,
housing
next
to
accessible
public
transportation
can
increase
property
values,
causing
gentrification
and
pushing
disadvantaged
populations
out.
With
no
comprehensive
analysis
of
the
interactions
between
these
socioeconomic
variables
and
household
CO2
we
are
left
in
the
dark.
2.6 Conclusion
The
literature
examining
the
factors
that
influence
energy
driven
CO2
emissions
is
deep;
research
abounds
studying
the
influence
of
physical
characteristics
of
housing
and
urban
structure
on
household
CO2
emissions.
However,
a
distinct
lack
of
research
on
smaller
scales
means
urban
dynamics
haven’t
been
thoroughly
examined.
Recent
advances
in
small-‐scale
CO2
emission
modeling
has
opened
up
more
possibilities
for
analysis
at
scales
smaller
than
a
city.
This
allows
the
spatial
distribution
of
CO2
emissions
across
a
city
to
be
analyzed
and
for
more
careful
analysis
of
the
drivers
of
urban
anthropogenic
CO2
emissions.
Additionally,
because
cities
are
both
geographically
located
in
relatively
small
areas,
and
because
cities
are
the
most
local
unit
of
governance,
there
is
less
variability
across
its
footprint
by
some
large
biophysical
drivers.
An
analysis
could
better
control
for
differences
in
climate
and
policy
and
perhaps
other
exogenous
variables.
Given
the
implications
of
climate
change
and
the
continuing
efforts
of
local
action
on
that
issues,
a
26
better
and
more
nuanced
understanding
of
urban
carbon
dioxide
emissions
could
be
beneficial
in
informing
policy.
Accordingly,
this
research
proses
to
examine
the
socioeconomic
drivers
of
CO2
emissions
within
the
boundary
of
a
major
metropolitan
area.
27
Chapter 3: Methods
3.1 Aim and Objectives of Research
The
main
objective
of
this
study
is
to
analyze
and
address
the
influence
of
household
socioeconomic
variables
on
household
CO2
emissions
at
the
sub-‐city-‐level
by
incorporating
and
identifying
the
influence
of
space;
that
is,
the
interactions
of
these
variables
and
other
unmeasured
factors
between
neighborhoods.
This
study
has
three
main
objectives.
First,
the
influence
of
spatiality
in
the
analysis
of
neighborhood-‐level
CO2
emissions
across
a
city
will
be
specifically
controlled
for
and
examined.
Second,
this
study
will
analyze,
and
attempt
to
find
further
support
for,
the
influence
of
household
income
on
household
CO2
emissions.
Third,
this
research
will
attempt
to
analyze
the
influence
of
race
and
ethnicity
of
households
on
household
CO2
emissions.
Finally,
all
of
these
analyses
will
be
performed
separately
for
total
household
CO2
emissions
(which
consists
of
residence
household
CO2
emissions
and
transportation
CO2
emissions),
as
well
as
for
residence
household
CO2
emissions
and
transportation
CO2
emissions,
in
order
to
examine
the
differing
relationships
among
these
household
sectors.
These
objectives
can
be
broken
down
into
three
general
hypotheses,
and
further
defined
by
the
three
CO2
variables,
total,
transportation
and
residence
household
CO2
emissions,
creating
9
total
hypotheses:
1)
Income
Household
Income
has
a
positive
relationship
with
total
household
CO2
emissions,
transportation
emissions,
and
residence
emissions,
respectively.
28
2)
Race
and
ethnicity
The
race
and
ethnicity
of
a
household
has
a
significant
relationship
with
total
household
CO2
emissions,
transportation
emissions,
and
residence
emissions,
respectively.
3)
Spatial
variables
For
the
total
household
CO2,
transportation,
and
residence
statistical
models,
the
spatial
lag
regression
will
produce
a
better
overall
fit
than
the
OLS
regression.
To
test
these
hypotheses,
data
was
gathered
from
independent
outside
sources
and
organized
at
the
census
tract
level.
OLS
regression,
exploratory
spatial
data
analyses
and
spatial
lag
regression
analyses
were
performed
on
the
resultant
dataset.
3.2 Study Area
Data
availability
determined
that
the
analyses
would
be
carried
out
for
the
city
of
Indianapolis,
IN.
Indianapolis
is
the
capitol
of
Indiana
and
the
county
seat
of
Marion
County.
Indianapolis
and
Marion
County
form
a
consolidated
city-‐county
–
they
have
merged
into
one
single
jurisdiction,
termed
Unigov.
During
this
merger,
some
previously
incorporated
cities
within
Marion
County
elected
to
retain
their
autonomy
from
Unigov.
Similarly,
some
towns,
although
now
included
within
Unigov,
retain
some
independent
government
functions.
The
term
balance
–
as
used
in
Indianapolis
(balance)
–
is
a
census
term
used
to
designate
the
area
of
Indianapolis-‐Marion
County
that
excludes
these
particular
cities
and
towns.
As
of
2010
the
population
of
Indianapolis
(balance)
was
820,445,
the
12th
largest
city
in
the
United
States.
This
analysis
however,
utilizes
data
from
29
2000
when
the
population
of
Indianapolis
(balance)
was
781,870.
The
population
of
Marion
County
in
2000
was
860,454.
The
county
is
largely
urbanized,
with
99%
of
the
population
residing
with
census
designated
urban
areas.
Indianapolis
is
much
less
dense
than
many
similarly
populated
cities
such
as
San
Francisco,
CA,
but
is
similar
to
Phoenix,
AZ
in
density.
There
are
a
total
of
212
census
tracts
in
all
of
Marion
County,
however
only
210
are
employed
in
the
analysis
because
of
insufficient
data
for
two
of
the
tracts
(Figure
1).
Because
of
the
close
geographic
and
demographic
similarities
between
Indianapolis
(balance)
and
the
entirety
of
Marion
County,
this
analysis
will
be
conducted
on
the
entire
county.
Figure
1:
Inset
map
of
Marion
County
–
Indianapolis
in
the
state
of
Indiana.
There
are
212
census
tracts
in
the
county,
however
only
210
are
used
in
this
analysis.
Unused
census
tracts
are
shown
with
white
hatching.
30
3.3 Data
3.3.1 Variables
3.3.1.1
Residence
CO2
Emission
Data
Household
CO2
data
was
gathered
for
two
separate
sources:
emissions
associated
with
the
residence
and
emissions
associated
with
household
transportation.
First,
household
residence
CO2
was
obtained
from
the
Hestia
Project.
They
used
downscale
modeling
techniques
to
determine
building-‐level
CO2
emissions
of
the
residential,
commercial,
and
industrial
sectors
for
the
entirety
of
Marion
County
in
2002
from
county
estimates
of
CO2
emissions
(see
discussion
below).
Residences
were
identified
using
building
parcel
data
from
the
Marion
County
Assessor’s
Office
(Gurney
et
al.,
2012).
The
downscaling
process
utilized
in
the
Hestia
Project
defined
general
physical
factors
of
each
residential
building
such
as
age
and
type
(e.g.
apartments
versus
single-‐family
detached
home)
as
a
function
of
energy
use
intensity
(EUI),
that
is,
the
energy
used
per
unit
floor
area.
Each
residential
building
was
placed
into
eight
EUI
categories
and
its
total
energy
use
was
found
through
its
EUI
and
its
size
(Zhou
&
Gurney,
2010).
In
general,
this
modeled
function
corresponded
to
higher
EUI
and
thus
higher
CO2
emissions
with
older
and
larger
buildings.
This
downscaling
process
does
not
consider
particular
unique
characteristics
that
are
specific
to
each
individual
building
itself,
such
as
remodels
and
retrofits
or
the
quality
of
building
that
could
potentially
impact
CO2
emissions.
For
its
CO2
input,
the
Hestia
Project
used
county-‐level
CO2
emission
sector
estimates
from
the
Vulcan
Project,
another
downscaling
endeavor
which
itself
utilized
projections
from
Environmental
Projection
Agency
and
state-‐level
reporting
on
energy
utilities
31
(Gurney
et
al.,
2012;
Zhou
&
Gurney,
2010).
Mapping
CO2
emissions
at
a
scale
smaller
than
the
city
scale
is
relatively
new
and
thus
very
little
to
no
data
exists
at
small-‐scale
levels.
CO2
data
from
the
Hestia
Project
was
the
limiting
factor
that
necessitated
analysis
of
Indianapolis,
IN
because
it
was
the
only
source
from
which
small-‐scale
residential
CO2
emissions
could
be
obtained.
3.3.1.2
Transportation
CO2
Emissions
Data
Second,
data
for
CO2
emissions
attributed
to
household
transportation
was
acquired
with
permission
from
the
H
+
T
Affordability
Index
created
by
the
Center
for
Neighborhood
Technology.
The
2009
H
+
T
Affordability
Index
uses
data
from
the
Census
Bureau’s
2009
American
Community
Survey
(ACS)
5-‐year
(2005-‐2009)
yearly
estimates
as
its
primary
dataset
to
model
housing
and
transportation
costs
at
the
census
block
group
level.
Transportation
costs
were
modeled
as
part
of
a
houses
location
based
on
auto
ownership,
auto
use,
and
transit
use.
Vehicle
miles
traveled
were
calculated
as
part
of
auto
use,
and
CO2
emissions
associated
with
transportation
per
household
were
thus
calculated
from
these
models.
Socioeconomic
data
was
procured
from
the
U.S.
Census
Bureau
for
the
2000
Census
at
the
census
tract
level
for
Marion
County.
The
census
is
designed
to
obtain
counts
of
the
entire
U.S.
population
as
well
as
additional
demographic
characteristics
such
as
race
and
ethnicity.
Demographic
data
is
thus
accurately
available
for
the
entire
United
States
at
extremely
small
scales.
ACS
data
was
deemed
unfit
for
race/ethnicity
data
because
of
large
margins
of
error
at
the
small
spatial
scales
required
for
this
analysis.
Independent
variables
32
utilized
for
this
analysis
include
median
income,
race
and
ethnicity
(Asian
and
black
households),
median
age,
and
household
size
all
received
from
the
U.S.
Census
Bureau.
3.3.1.3
Spatial
Data
The
level
of
analysis
is
the
census
tract,
a
geographic
area,
designated
to
be
between
2,500
to
8,000
people,
by
local
census
statistical
areas
committees
using
guidelines
established
by
the
U.S.
Census
Bureau
(United
States
Census
Bureau,
2000).
Marion
County
has
212
census
tracts
in
total,
but
only
210
were
utilized
in
this
analysis
because
of
missing
data.
One
census
tract
was
omitted
because
of
missing
census
data
(as
shown
in
both
the
2000
Census
and
2009
ACS),
the
other
because
of
missing
residence
CO2
data.
Shapefiles
for
Marion
County
census
tracts
were
procured
from
ESRI
for
the
2000
Census.
These
files
were
transformed
into
GIS
shapefiles
by
ESRI
using
the
U.S.
Census
Bureau’s
TIGER
database.
3.3.2 Data Preparations
Variables
were
compiled
into
a
spatial
database
using
ArcMap
10.
While
socioeconomic
data
from
the
U.S.
Census
Bureau
was
obtained
at
the
census
tract
scale,
both
CO2
emissions
for
residence
and
transportation
were
reported
at
smaller
scales.
Residential
CO2
data
as
gathered
from
the
Hestia
Project
was
organized
in
a
GIS
shapefile
with
individual
polygons
for
each
building.
A
spatial
join
was
performed
on
this
data
with
the
Marion
County
census
tract
shapefile
(obtained
from
the
ESRI
Census
2000
TIGER/Line
Shapefiles
database)
using
the
center
of
each
building
polygon
as
a
bias-‐free
criterion
to
aggregate
the
buildings
to
the
census
tract.
As
part
of
this
operation
CO2
emissions
were
summed
from
each
building
polygon
to
create
census
tract
totals
of
residential
CO2,
which
33
were
were
then
divided
by
the
number
of
households
for
each
tract,
obtaining
average
household
residence
CO2
emissions.
Transportation
CO2
data
was
organized
at
the
census
block
group
level.
Block
groups
are
the
statistical
spatial
groups
organized
within
census
tracts.
Data
was
transferred
into
Excel
where
the
block
groups
and
their
associated
attributes
were
aggregated
through
summation
to
their
respective
census
tracts.
Subsequently,
the
transportation
CO2
data
was
moved
into
ArcMap
10
and
joined
to
the
census
tract
shapefile.
Residence
CO2
emissions
per
household
and
transportation
CO2
emissions
per
household
data
were
summed
to
create
total
CO2
emissions
per
household
per
census
tract.
All
CO2
emissions
are
reported
in
average
CO2
emissions
per
household
per
census
tract.
The
polygon
census
tract
feature
class
thus
has
associated
with
it
that
tract’s
average
household
residential
CO2,
average
household
transportation
CO2,
average
total
household
CO2,
median
household
income,
distribution
of
race
and
ethnicity
(percentage
Asian
and
black
households),
average
household
size,
and
average
age.
Average
household
CO2
data
is
reported
in
metric
tons
(t)
of
CO2
emitted
per
year.
3.4 Limitations
There
were
several
limitations
to
this
study.
First,
data
requirements
have
thus
far
held
back
extensive
study
at
the
neighborhood
scale.
The
city
of
Indianapolis,
IN
was
chosen
for
this
study
because
the
type
of
fine-‐scale
data
required
was
only
freely
and
readily
available
for
that
city.
If
further
analysis
is
to
be
done
on
other
cities,
fine-‐scale
data
will
have
to
be
produced
first,
which
is
an
expensive,
extensive,
and
difficult
task.
Hopefully,
with
continuing
advances
in
modeling,
that
will
become
possible.
This
lack
of
34
accessibility
has
implications
for
this
current
analysis,
however.
Due
to
these
difficulties
with
fine-‐scale
data,
several
additional
limitations
were
imposed
on
the
analysis
as
discussed
over
the
next
paragraphs.
Second,
the
type
of
data
utilized
in
this
study
cannot
account
for
behavioral
differences.
Household
emissions
for
transportation
CO2
and
residence
CO2
were
derived
from
different
modeling
processes
and
from
different
sources.
These
models
did
not
necessarily
take
into
account
behavior
factors.
For
example,
the
residential
CO2
data
derived
from
The
Hestia
Project
was
created
using
downscaled
CO2
emissions
at
the
building
scale
for
the
commercial,
residential,
and
industrial
sectors
of
Indianapolis,
IN.
The
nature
of
the
downscale
modeling
process,
utilizing
large-‐scale
characteristics
such
as
year
and
square
footage,
meant
individual
changes
to
buildings
could
not
be
factors
of
that
model.
Homes
that
had
upgraded
to
energy
efficient
windows
or
insulation
for
example,
would
not
see
this
reflected
in
their
CO2
emissions.
Third,
the
data
used
in
this
analysis
does
not
cover
the
totality
of
emissions
associated
with
households.
Consumption
is
not
captured
in
this
modeling
process
at
all.
There
is
an
entire
section
of
economics
research
devoted
to
this
issue,
that
will
not
be
discussed
here,
but
this
is
a
significant
amount
of
CO2
emissions
to
not
include
in
the
modeling.
Additionally,
the
residence
CO2
data
was
derived
from
point
emissions
only.
This
means
natural
gas
emissions
were
modeled
to
the
households
as
would
be
used
in
central
heating,
for
example.
But
electricity
data,
of
which
the
point
source
of
fuel
combustion
is
a
power
plant,
was
attributed
to
that
power
plant
in
the
Hestia
Project
and
is
not
attributed
to
the
household.
However,
because
CO2
emissions
were
attributed
to
all
households,
35
regardless
of
their
actual
consumption
of
natural
gas
for
heating,
this
analysis
may
still
pick
up
on
overall
trends
and
patterns
in
the
relationships
between
variables.
Finally,
the
data
obtained
for
analysis
was
from
different
years.
Explanatory
socioeconomic
variables
were
obtained
from
the
2000
Census.
Household
transportation
CO2
was
modeled
using
data
from
the
2009
ACS
5-‐year
(2005-‐2009)
estimates,
although
the
data
was
organized
using
2000
Census
boundaries.
Household
residence
CO2
was
modeled
using
emissions
estimates
for
2002.
These
differences
are
insurmountable
for
this
particular
study;
analysis
needs
necessitated
these
datasets.
The
difference
between
these
two
sources
and
the
2009
ACS
and
resultant
transportation
CO2
emissions
may
be
slightly
difficult
to
justify,
but
since
that
dataset
utilized
2000
Census
boundaries,
not
to
discount
the
intensive
time
demands
in
any
attempt
to
project
onto
2010
Census
boundaries,
it
was
deemed
tolerable.
The
difference
between
the
2000
Census
data
and
the
2002
residence
CO2
emissions
is
more
acceptable.
3.5 Statistical Analysis
Neighborhoods
are
not
disconnected
units.
Spatial
dependency
exists
between
contiguous
neighborhoods,
and
interactions
between
them
must
be
taken
into
consideration
when
performing
statistical
analyses
or
the
reliability
of
results
may
be
overestimated.
According
to
(Ward
&
Gleditsch,
2008),
“ignoring
spatial
dependence
will
tend
to
underestimate
the
real
variance
in
the
data.”
To
assess
whether
spatiality
exists
within
the
variables,
a
spatial
weights
matrix
is
constructed
as
a
descriptor
of
the
spatial
36
relationships
between
individuals,
in
this
case
individuals
being
census
tracts.
The
spatial
weights
matrix
can
be
structured
in
several
different
ways,
either
utilizing
distances
between
individuals
or
through
assessing
shared
boundaries
between
neighbors.
The
shapefile
containing
the
Marion
County
census
tracts
and
associated
variables
was
transferred
to
OpenGeoDa
1.2.0
(2012)
for
exploratory
spatial
analysis
and
to
construct
a
spatial
weights
matrix
to
employ
within
these
analyses.
For
this
study
a
boundary
based
matrix
was
chosen,
specifically
a
queen
contiguity
based
spatial
weights
matrix
was
chosen
over
the
rook
contiguity
based
spatial
weights
matrix
as
a
better
overall
fit.
Exploratory
analysis
showed
minimal
differences
in
final
regression
fit
between
the
queen-‐based
contiguity
matrix
and
rook-‐based
contiguity
matrix,
but
the
distribution
of
connections
was
found
to
be
more
normal
in
the
queen
matrix
than
the
rook.
This
along
with
considerations
of
possible
connection
between
census
tracts,
pointed
towards
selecting
the
queen
matrix.
This
matrix
was
constructed
using
GeoDa.
A
typical
regression
can
be
expressed
as
𝑦 = 𝑥𝛽 + 𝜀,
(
1
)
where
y
is
the
dependent
variable,
x
is
the
independent
variable,
β
is
the
slope
of
the
regression
equation,
representing
the
relationship
between
the
x
and
y
variables,
and
ε
represents
the
error
of
the
equation.
The
spatial
lag
regression
model
incorporates
the
spatial
autoregressive
term,
𝜌𝑊𝑦,
on
the
right
side,
accounting
for
the
influence
of
the
neighbor
on
each
census
tract,
an
influence
that
is
actually
a
weighted
average
of
the
surrounding
neighbors.
Thus,
following
Anselin
(1988),
37
𝑌 = 𝜌𝑊𝑦 + 𝑋𝛽 + 𝜀.
(
2
)
In
this
spatial
lag
model
ρ
is
the
autoregressive
parameter
and
W
is
the
constructed
spatial
weights
matrix,
which
in
the
case
of
this
study
is
the
queens
contiguity
spatial
weights
matrix
as
discussed
earlier,
whereby
both
parameters
work
on
a
right
side
y
variable
to
complete
the
spatial
autoregressive
term.
Both
the
Y
and
X
are
in
matrix
form,
as
consistent
with
the
rest
of
the
model
(Ward
&
Gleditsch,
2008).
Three
regression
models
were
created:
one
each
for
average
total
household
CO2
(total
CO2),
average
household
transportation
CO2
(transportation
CO2),
and
average
household
residence
CO2
(residence
CO2).
Ordinary
Least
Squares
(OLS)
regressions
were
carried
out
on
the
three
models.
The
resultant
regression
residuals
were
tested
for
spatial
autocorrelation
using
Moran’s
I
test.
Positive
spatial
autocorrelation
in
the
residuals
revealed
the
presence
of
spatial
dependence
in
the
variables.
According
to
Lagrange
Multiplier
tests
ran
on
the
OLS
regressions
(Equation
1),
and
through
theoretical
considerations,
the
spatial
lag
model
(Equation
2)
was
selected
over
the
spatial
error
model,
another
type
of
spatial
regression
where
the
influence
of
space
is
seen
to
be
an
error
and
thus
treated
as
such.
The
data
and
spatial
matrix
were
transferred
to
Stata/SE
12.0
(2011)
for
use
in
the
Stata
module
SPMLREG
(Jeanty,
2010).
Spatial
lag
regressions
were
performed
in
Stata
on
the
three
CO2
models
using
the
queen
contiguity
spatial
weights
matrix
to
construct
a
spatially
lagged
CO2
variable
to
account
for
spatial
dependencies.
Persistant
heteroskedasticity
among
the
residuals,
even
after
the
log
transformation
of
median
income
lent
support
for
robust
standard
errors
over
regular
standard
errors.
38
Huber-‐white
estimators
(robust
standard
errors)
can
be
use
to
compensate
for
the
possible
effects
of
heteroskedasticity
and
non-‐normality
of
residuals
in
overestimating
standard
errors,
and
thus,
p-‐values,
leading
to
skewed
inferences.
The
inclusion
of
these
robust
standard
errors
does
not
affect
coefficients.
Tests
were
carried
on
the
residuals
of
all
the
models
to
assess
for
spatial
dependency.
All
regression
results
were
transferred
into
GeoDa
to
perform
Local
Moran’s
I
on
the
residuals.
39
Chapter 4: Results
4.1 Descriptive Statistics
The
dependent
and
independent
variables
exhibit
clear
variability
and
spatial
clustering
throughout
Marion
County.
Table
1
provides
quick
quartile
descriptions
of
each
variable.
The
dependent
variable
total
household
CO2
is
a
summation
of
household
transportation
and
residence
CO2
emissions.
Evident
from
Table
1
is
the
discrepancy
between
transportation
and
residence
CO2
emissions:
at
the
low-‐end
residence
CO2
barely
registers
in
at
least
one
census
tract.
This
is
likely
due
to
to
issues
in
the
downscale
modeling
process.
Dependent
Variables
Total
CO2
(t)
Transportation
CO2
(t)
Residence
CO2
(t)
Independent
Variables
Median
income
($)
Median
age
Average
household
size
%
Asian
%
Black
Five
Number
Summary
from
Across
Census
Tracts
Minimum
Q1
Median
Q3
Maximum
4.564
8.238
8.717
9.481
13.067
4.438
7.215
7.817
8.243
10.976
0.003
0.750
1.081
1.435
5.287
13,125
29,798
35,547.5
49,794.5
21.9
30.7
34.0
37.6
1.39
2.18
2.44
2.63
0.00
0.27
0.65
1.44
0.09
3.03
12.13
39.74
133,479
51.8
3.20
7.79
98.17
Table
1:
Five
number
summary
of
independent
and
dependent
variables.
Presented
are
the
minimum,
first
quartile,
median,
third
quartile,
and
maximum
value
across
all
census
tracts.
40
Figure
2:
Average
household
transportation
CO2
per
census
tract
in
Indianapolis,
IN.
Figure
3:
Average
household
residence
CO2
per
census
tract
in
Indianapolis,
IN.
41
Figure
4:
Average
total
household
CO2
per
census
tract
in
Indianapolis,
IN
The
difference
at
the
median
between
transportation
CO2
and
residence
CO2
is
approximately
7
t.
Variability
within
the
transportation
and
residence
CO2
variable
is
evident,
the
range
for
each
being
approximately
5t
and
6t
respectively,
while
total
household
CO2
sees
a
range
of
around
8
t.
Examination
of
the
dependent
variable
across
their
respective
maps
reveals
evident
visual
clustering.
Transportation
CO2
increases
as
one
moves
away
from
the
center
of
Marion
County
(Figure
2).
Residence
CO2
exhibits
less
easily
identifiable
clustering,
but
several
areas
of
clustering
do
exist
at
several
places
(Figure
3).
Total
CO2,
an
amalgamation
of
both
variables,
has
notable
clustering
around
the
outskirts
of
Marion
County,
especially
near
the
southern
edge,
while
the
center
also
exhibits
apparent
high
levels
of
low-‐value
CO2
emission
clustering
(Figure
4).
42
Race
and
ethnicity
ranges
and
clusters
noticeably
across
the
county.
Asian
households,
absent
entirely
from
some
census
tracts,
are
not
present
in
considerable
numbers
across
Marion
County.
For
example,
the
median
percentage
of
Asian
households
per
census
tract
is
0.65%,
with
a
maximum
being
approximately
7.8%
of
Asian
households
per
tract.
Those
few
census
tracts
with
significant
Asian
populations
are
located
predominantly
in
two
clusters,
one
just
northwest
of
the
center
of
Indianapolis
and
then
an
additional
sector
that
is
directly
north
of
the
center,
at
the
boundary
of
the
county
(Figure
5).
A
clear
contrary
to
Asian
households,
black
households
range
from
approximately
0.08%
to
just
over
98%.
The
map
of
black
households
(Figure
6)
indicates
an
area
of
considerable
clustering
just
north
of
the
center
of
Marion
County,
that
does
not
quite
reach
the
county
line,
and
where
concentrations
of
black
households
are
around
50%.
Income
also
reveals
significant
variation
and
clustering
throughout
Marion
County.
Especially
stark
is
the
range
of
the
median
income
of
households
per
census
tract
across
census
tracts,
from
$13,125
to
$133,479.
Median
household
income
for
Marion
County
was
$40,421
in
2000,
close
to
the
2000
national
median
income
of
$41,994.
Median
household
income
across
census
tracts
for
Marion
County
is
$35,547.50,
a
discrepancy
due
to
differences
in
the
scale
of
analysis:
each
census
tract
has
varying
number
of
households.
Examining
the
map
for
income
(Figure
7)
the
clustering
is
especially
evident
along
the
border
of
the
county,
where
high
levels
of
income
per
household
are
in
clear
contrast
with
the
center
of
Marion
County
where
income
per
household
is
considerably
lower.
43
Figure
5:
Percentage
of
Asian
households
per
census
tract
in
Indianapolis,
IN.
Figure
6:
Percentage
of
black
households
per
census
tract
in
Indianapolis,
IN.
44
Figure
7:
Median
household
income
per
census
tract
of
Indianapolis,
IN.
4.2 Statistical Analysis
Visual
inspection
of
the
maps
illustrating
the
average
CO2
–
total,
transportation,
and
residence
–
of
households
per
census
tract
indicate
spatial
clustering
in
fairly
distinct
patterns.
Moran’s
I
statistical
tests
can
reveal
the
extent
and
significance
of
these
spatial
patterns,
statistically
determining
when
clustering
or
dispersal
of
variables
is
present.
The
Moran’s
I
was
ran
on
each
of
the
CO2
variables,
and
indicated
significant
spatial
clustering
among
all
three,
supporting
the
use
of
spatial
lag
models
to
control
for
that
spatial
autocorrelation
among
residuals
(see
Appendix
A
for
results).
45
Moran's
I
OLS
Total
CO2
0.267
*
Transportation
CO2
0.484
*
Residence
CO2
0.278
*
Spatial
Lag
Total
CO2
Transportation
CO2
Residence
CO2
0.022
-‐0.029
0.096
*
Table
2:
Moran's
I
test
for
spatial
dependency
among
the
residuals
of
the
regression
models.
P-‐values
are
based
on
replications
where
significant
p-‐values
denote
spatial
dependency
among
the
residuals.
*
denotes
significance
at
p<.01
Additionally,
Moran’s
I
tests
were
ran
on
the
residuals
of
the
three
OLS
regression
models.
Spatial
dependency
was
found
in
the
residuals
of
all
OLS
models
with
particularly
evident
clustering
found
in
household
transportation
CO2
model.
Table
2
compares
Moran’s
I
tests
from
the
OLS
and
spatial
lag
models.
Improvements
were
made
in
all
spatial
lag
models.
Additionally,
neither
the
Moran’s
on
the
residuals
of
the
spatial
lag
model
for
total
CO2
and
transportation
CO2
are
significant
at
all,
indicating
that
we
cannot
reject
the
null
hypothesis
that
the
data
is
randomly
distributed
(not
clustered
or
dispersed).
Figure
8
illustrates
the
output
of
a
Local
Moran’s
I
on
the
residuals
of
the
total
CO2
OLS
regression.
Clustering
of
the
residuals
is
evident.
Figure
9
contains
the
results
of
the
Local
Moran’s
of
the
total
CO2
spatial
lag
regression.
Although
a
clustering
of
values
is
still
present
in
the
spatial
lag
residuals,
comparison
between
the
two
maps
indicates
a
reduction
in
overall
clustering.
Local
Moran’s
I
maps
for
the
transportation
and
residence
CO2
OLS
and
spatial
lag
models
follow
similar
patterns
and
trends.
46
Figure
8:
Local
Moran
on
residuals
of
total
household
CO2
OLS
regression
model.
Significant
clustering
is
represented
by
the
Moran
clustering
typology.
High-‐high
and
low-‐low
designate
clustering
of
positive
spatial
autocorrelation,
while
high-‐low
and
low-‐high
designate
spatial
outliers
or
negative
spatial
autocorrelation.
Typology
designates
the
core
of
the
spatial
autocorrelation.
Significant
replication
at
p<0.05.
Figure
9:
Local
Moran
on
residuals
of
total
household
CO2
spatial
lag
regression
model.
Significant
replication
at
p<0.05.
The
legend
is
as
described
in
Figure
8.
47
The
spatial
lag
regression
models
for
total
CO2
and
transportation
CO2
made
particular
improvements,
approximately
eliminating
spatial
dependency
among
residuals
and
testing
for
non-‐significant
Moran’s
I.
The
regression
model
for
residence
CO2
saw
a
decrease
in
Moran’s
I,
however
it
still
tests
significantly
for
spatial
dependency
(Table
2),
and
thus
this
particular
model
does
not
adequately
control
for
all
spatiality,
although
it
is
reduced.
This
is
sufficient
support
that
spatial
regression
techniques
should
be
utilized
to
reduce
spatial
autocorrelation
among
residuals
and
improve
the
model
(see
Appendix
B
for
Local
Moran’s
of
other
models).
Results
from
the
spatial
lag
model
regressions
are
displayed
alongside
their
OLS
counterparts
in
Table
3
to
demonstrate
improvement
in
overall
fit
of
models
after
the
inclusion
of
the
spatial
variable.
Because
R2
values
do
not
hold
when
spatial
dependence
is
present,
the
log
likelihood
can
be
used
to
assess
goodness
of
fit
(Anselin,
1988).
Among
each
modeled
dependent
variables
of
CO2,
the
log
likelihood
increases
with
the
introduction
of
the
spatial
elements
of
the
regression
model
over
their
OLS
counterparts.
For
example,
the
total
CO2
OLS
model
has
a
log
likelihood
(df
=
204)
of
-‐181.425.
The
total
CO2
spatial
lag
model
has
a
log
likelihood
(df
=
203)
of
-‐141.636,
an
increase
that
indicates
an
improvement
in
the
fit
of
the
model.
Figure
10
illustrates,
through
a
regression
scatterplot,
the
values
of
the
modeled
total
CO2
spatial
lag
regression
against
the
observed
values
of
total
CO2
per
census
tract.
The
regression
fits
well
until
higher
levels
of
CO2,
and
then
the
variability
between
the
modeled
values
and
observed
values
increases
slightly
(see
Appendix
C
for
scatter
plots
of
the
other
spatial
lag
models).
Using
Wald
Tests
to
assess
the
hypotheses
of
the
spatial
lag
models
(again,
because
R2
cannot
be
used
in
spatial
48
lag
regression)
we
find
that
all
models,
total
CO2
(χ2203,
210
=
314.9,
p
<
.001),
transportation
CO2
(χ2203,
210
=
136.1,
p
<
.001),
and
residence
CO2
(χ2203,
210
=
117.5,
p
<
.001)
are
significant
(Anselin,
1988).
The
direction
of
coefficients
is
consistent
across
all
spatial
lag
models
for
each
explanatory
variable,
however
there
is
considerable
derivation
in
the
explanatory
power
of
variables.
Standardized
beta
coefficients
can
be
used
to
compare
coefficient
strength.
Median
income
has
the
most
influence
in
the
total
CO2
and
transportation
CO2
models,
although
household
size
also
has
a
strong
influence,
and
in
some
cases
(i.e.
transportation
CO2),
this
coefficient
is
only
slightly
smaller
than
the
coefficient
for
median
income.
The
median
age
also
explains
a
significant
amount
of
this
model
as
well.
In
the
residence
CO2
model
however,
both
household
size
and
age,
first
and
second
most
influential
respectively,
have
more
influence
than
income.
P-‐values
for
these
three
variables
are
significant
in
all
spatial
models
to
at
least
the
0.001
level.
Total
CO2
Spatial
Lag
Regression
Actual
CO2
Emissions
15
t
CO2
per
household
13
11
9
7
5
3
3
5
7
9
11
13
15
Predicted
CO2
Emissions
Figure
10:
The
regression
plot
of
total
household
CO2
emissions
of
the
predicted
model
by
the
observed
emissions
for
the
spatial
lag
regression.
49
Variation
exists
in
the
direction
and
significance
between
the
race
and
ethnicity
variables
and
the
CO2
dependent
variables.
Black
households
exhibit
a
positive
association
with
CO2
in
all
models,
although
both
the
coefficients
and
standardized
(beta)
coefficients
are
small
and
thus
have
weak
influence.
By
contrast,
Asian
households
have
a
negative
association
with
CO2,
although
the
coefficients
are
similarly
small
and
have
weak
influence.
Asian
households
do
not
test
significantly
in
the
transportation
CO2
model
and
is
only
tenuously
significant
in
the
residence
CO2
model
with
a
p-‐value
of
0.059.
Black
households
attain
significance
at
the
0.05
level
in
the
residence
CO2
model
with
a
p-‐value
of
0.045.
The
lagged
spatial
variable
has
a
consistently
strong
and
significant
influence
in
all
three
spatial
models.
The
influence
of
the
weighted
CO2
variable
is
strongest
within
the
transportation
CO2
model.
Here,
the
weighted
spatial
variable
has
the
highest
standardized
coefficient
(Beta
=
0.675)
between
all
other
variables
and
also
exerts
considerable
change
on
the
coefficients
of
the
variable
when
introduced
into
the
model.
For
example
the
coefficient
of
the
log
of
median
income
in
the
OLS
transportation
CO2
model
diminishes
from
1.605
to
0.525
in
the
spatial
lag
model.
A
similar
change
is
noticed
in
the
total
CO2
model
where
the
weighted
variable
(Beta
=
0.337)
leads
to
a
decrease
in
the
log
of
median
income
from
2.145
in
the
OLS
model
to
1.418
in
the
spatial
model.
The
weighted
spatial
variable
exerts
less
influence
in
the
residence
CO2
model
(Beta
=
0.193)
compared
to
the
others,
although
there
is
observable
influence
from
this
variable
on
Asian
households
and
median
income,
reducing
the
coefficients
of
both
in
the
spatial
lag
model.
50
TOTAL
HOUSEHOLD
CO2
OLS
coefficient
p-‐value
Median
Age
0.050
0.000
Black
Households
0.005
0.001
Asian
Households
-‐0.168
0.000
1.402
0.000
Average
Household
Size
Log
Median
Income
Constant
2.145
0.000
-‐18.783
0.000
beta
0.204
coefficient
0.695
OLS
Log
likelihood
=
-‐144.636
0.133
-‐0.166
0.402
Spatial
Variable
Median
Age
coefficient
p-‐value
0.014
0.204
Black
Households
0.002
0.196
Asian
Households
-‐0.062
0.168
Average
Household
Size
0.878
0.000
Log
Median
Income
1.605
0.000
-‐11.759
0.000
Constant
OLS
Median
Age
0.035
0.000
Black
Households
0.003
0.020
Asian
Households
-‐0.106
0.002
Average
Household
Size
0.524
0.000
Log
Median
Income
0.540
0.000
-‐7.023
0.000
Constant
-‐0.108
1.244
0.000
***
0.357
1.418
0.000
***
0.459
-‐15.328
0.000
***
0.153
-‐0.210
0.302
0.352
0.000
***
0.337
p-‐value
0.020
0.000
***
0.002
0.012
**
-‐0.023
beta
0.104
0.067
0.360
-‐0.028
0.592
0.000
***
0.215
0.525
0.000
***
0.215
-‐6.395
0.000
***
0.000
***
0.675
0.834
beta
0.294
Spatial
Variable
0.017
**
Log
likelihood
=
-‐134.459
p-‐value
-‐0.110
coefficient
0.318
0.657
coefficient
0.114
Log
likelihood
=
-‐77.637
0.071
-‐0.077
0.219
0.001
***
Spatial
Lag
beta
0.074
0.000
***
0.005
beta
0.053
Spatial
Variable
p-‐value
0.498
Log
likelihood
=
-‐195.916
RESIDENCE
CO2
Spatial
Lag
Log
likelihood
=
-‐181.425
TRANSPORTATION
CO2
Spatial
Lag
Log
likelihood
=
-‐125.132
coefficient
p-‐value
beta
0.036
0.000
***
0.295
0.003
0.045
**
0.126
-‐0.078
0.059
*
-‐0.155
0.534
0.000
***
0.309
0.474
0.010
***
0.309
-‐6.819
0.000
***
0.399
0.000
***
0.193
Table
3:
OLS
and
spatial
lag
model
regression
results.
For
spatial
lag
regression
*
indicates
weakly
significant
results
at
p
<
.10,
**
indicates
significant
results
at
p
<
.05,
and
***
indicates
significant
results
at
p
<
.01.
Comparison
of
p-‐values
between
the
OLS
models
and
spatial
lag
models
is
not
recommended
as
the
spatial
lag
models
use
robust
standard
errors
at
the
OLS
models
do
not.
51
Chapter 5: Discussion
The
urban
scale
provides
a
unique
analysis
opportunity.
Its
relatively
small
scale,
as
compared
to
state
or
national
levels,
renders
several
common
variables
that
are
seen
to
be
drivers
of
CO2
at
larger
scales,
such
as
climate
or
fuel
type,
negligible.
At
the
same
time,
the
diversity
and
variability
of
populations
across
the
urban
gradient
allows
for
a
careful
analysis
of
all
these
groups.
Household
CO2
emissions
vary
considerably
across
Indianapolis,
disaplying
visually
apparent
and
statistically
significant
spatial
patterning.
The
variation
in
these
emissions
within
and
across
Indianapolis,
lends
support
to
an
analysis
that
addresses
the
factors
that
drive
these
emissions.
In
conjunction
with
proposed
socioeconomic
factors,
our
understanding
of
household
emissions
will
be
improved.
It
is
by
first
examining
the
factors
that
drive
the
variability
in
CO2
emissions
that
we
can
efficiently
and
effectively
improve
strategies
and
policy
directed
at
combating
CO2
emissions
within
cities.
Results
from
the
spatial
lag
regression
indicate
the
importance
of
including
spatial
analysis
in
modeling
at
this
scale.
There
is
clear
support
for
the
spatial
lag
regression
when
considering
the
improved
modeled
fits
via
log
likelihood
measurements
from
the
OLS
models
to
the
spatial
lag
models.
The
reduction
in
spatial
autocorrelation
among
residuals
(as
measured
by
Moran’s
I)
between
OLS
and
spatial
lag
models
further
reinforces
the
importance
of
the
spatial
model.
Importantly,
although
there
is,
generally,
static
or
declining
explanatory
power
among
coefficients
with
the
integration
of
the
spatial
variables,
the
explanatory
power
of
almost
all
the
variables
remain
significant
or
become
52
significant
with
the
introduction
of
spatiality.
These
results
support
the
hypotheses
presented
in
this
paper
that
socioeconomic
variables
are
strong
predictors
of
urban
CO2,
and
that
space
is
an
important
concept
that
should
be
integrated
into
analyses
at
this
scale,
potentially
providing
useful
information
to
policy
makers
on
local
climate
change
policy
and
helping
expand
the
current
knowledge
on
climate
change
drivers.
5.1 Spatial Variables
It
was
important
to
include
and
consider
the
impact
of
space
and
spatial
interactions
in
this
analysis.
Besides
issues
with
autocorrelation
among
error
terms,
there
are
theoretical
considerations
as
well.
Census
tracts
are
devised
along
particular
guidelines,
attempting
to
create
fairly
homogenous
areas
taking
demographic
and
economic
characteristics
into
consideration.
However,
tracts
are
not
individual
in
the
same
way
a
person
or
household
might
be;
tracts
do
not
have
the
same
standard
and
clear
demarcations
between
units.
Instead,
the
guidelines
for
census
tracts,
although
set
for
the
particular
census,
are
fluid
and
determined
by
local
census
committees
(United
States
Census
Bureau,
2000)
and
thus
are
open
to
interpretation,
movement,
and
connection.
Interactions
between
census
tracts
are
inevitable.
Similarly,
the
spatial
lag
model
can
account
for
externalities
or
spatial
variables
not
necessarily
captured
in
the
regression
model.
That
becomes
especially
important
for
this
analysis
given
factors
of
urban
form
and
space,
factors
that
may
be
difficult
to
calculate
or
quantify
and
arise
from
similarities
in
urban
form
that
can
exist
between
census
tracts
which
occupy
similar
space.
At
the
same
53
time,
this
level
of
analysis,
while
large
in
comparison
to
a
single
household,
is
small
enough
to
assess
for
variations
across
the
urban
and
institutional
space.
The
spatial
analysis
supports
these
theoretical
considerations.
The
spatial
variables
are
significant
across
all
spatial
lag
models.
The
transportation
CO2
model
exhibits
the
highest
correlation
between
the
spatially
lagged
CO2
variable
and
the
dependent
CO2
variable
across
all
models
with
a
regression
correlation
coefficient
of
approximately
0.83.
Interpreted,
as
the
averaged
CO2
of
all
of
a
census
tract’s
neighbors
increases
by
1
metric
tonne
(t),
that
census
tract’s
CO2
can
be
said
to
increase
by
.83
t.
This
is
perhaps
expected
when
transportation
CO2
associated
with
the
household
is
highly
dependent
on
the
length
and
time
of
a
commute.
This
in
turn
is
highly
dependent
on
space,
or,
the
distance
from
the
household
to
the
job.
Patterning
associated
with
this
distance
is
clear
when
examining
a
map
of
transportation
CO2
(Figure
2).
Although
the
correlation
coefficients
for
the
other
spatial
lag
models
are
not
as
strong,
they
are
also
statistically
significant
and
have
substantial
explanatory
power
in
their
respective
regression
models.
There
is
spatial
clustering
evident
in
the
residence
CO2
model,
with
particularly
high
CO2
per
household
census
tracts
clustering
around
the
edges
and
in
the
center
of
the
study
area
(Figure
2).
These
clusters
could
be
caused
by
several
factors,
but
there
are
several
likely
reasons
that
we
can
consider.
For
example,
this
clustering
could
be
due
to
higher
levels
of
larger
households
than
one
would
associate
with
the
suburbs
and
with
older
housing
that
one
associates
with
city
centers
and
initial
early
developments,
both
of
which
are
going
to
increase
CO2
per
household
in
those
tracts.
Analysis
of
the
Hestia
Project
database
confirms
both
of
these
suspicions.
54
Total
household
CO2
(Figure
3)
is
an
amalgamation
of
transportation
CO2
and
residence
CO2
and
thus
contains
reference
to
patterns
visible
in
both.
Clear
spatiality
in
these
models
is
confirmation
that
spatial
analysis
is
integral
to
this
type
of
research.
Examination
of
households’
CO2
emissions
without
these
considerations
may
result
in
inadequate
models
and
overestimate
the
coefficients
of
the
other
variables.
However,
limitations
of
the
data
used
in
this
study,
that
is
the
relatively
larger
contribution
of
transportation
emissions
over
residence
emissions
to
household
emissions,
may
contribute
to
an
overemphasis
of
transportation
CO2
in
the
analysis
of
household
CO2,.
Consequently,
the
drivers
of
total
CO2
may
be
largely
driven
by
transportation
CO2.
The
direction
and
relative
power
of
coefficients
(betas)
within
each
spatial
lag
model
may
lend
some
support
to
reliability
to
the
overall
influence
of
individual
socioeconomic
factors
when
considering
that
(excluding
the
spatial
variable)
income
is
consistently
the
biggest
predictor
of
emissions,
followed
by
average
household
size,
then
median
income,
and
finally
the
race
and
ethnicity
variables.
This
holds
for
all
spatial
models,
so
while
transportation
CO2
is
quantitatively
more
significant
in
the
amount
of
metric
tonnes
per
household
it
contributes
to
CO2,
the
relationships
between
variables
seem
consistent.
5.2 Income
Unsurprisingly,
income
is
a
significant
predictor
of
CO2
across
all
models.
This
is
backed
by
substantial
literature,
including
research
that
has
examined
the
relationship
between
CO2
and
income
controlling
for
spatiality
among
variables,
albeit
at
larger
scales
than
in
the
present
analysis
(Burnett
&
Bergstrom,
2010;
Glaeser,
2012;
Golley
&
Meng,
55
2012;
Min
et
al.,
2010).
The
mechanisms
surrounding
these
trends,
which
exists
among
all
models,
is
clear-‐cut:
higher
levels
of
income
per
household
have
associated
higher
levels
of
spending.
This
increased
spending
power
manifests
itself
through
increased
CO2,
potentially
through
increased
housing
size,
number
of
vehicles,
and
vehicle
travel
distance.
Additionally,
through
a
mechanism
strongly
associated
with
race,
but
also
cheap
land,
upper-‐class
and
white
flight
from
inner
cities
has
moved
higher
income
households
farther
out
to
the
suburbs
increasing
commute
times
(Charles,
2003;
Garrett
&
Taylor,
1999).
Figure
4
illustrates
the
spatial
clustering
associated
with
median
income
in
Indianapolis.
Higher
income
households
are
intensely
located
at
the
edges
of
the
boundary
of
the
city.
But
even
controlling
for
spatiality
and
other
explanatory
variables,
income
is
a
significant
predictor
of
household
CO2.
Taking
into
account
the
log
transformation
of
the
median
income
variables
and
the
resultant
regression
coefficient
for
log
of
median
income,
increasing
income
by
100%,
say,
for
example,
from
the
50th
percentile
of
census
tracts’
household
median
income
to
approximately
the
95th
percentile,
would
increase
total
household
CO2
by
approximately
0.98t.
A
50%
increase
in
income
results
in
an
approximate
0.57t
increase
in
total
CO2.
The
log
relationship
between
CO2
and
income
has
been
well
established
in
the
literature
and
implicates
a
decreasing
marginal
relationship
of
the
two
variables
(Glaeser
&
Kahn,
2010;
Limpert,
Stahel,
&
Abbt,
2001).
In
other
words,
each
successive
additional
dollar
in
household
income
has
a
decreasing
impact
of
CO2,
and
thus
at
lower
levels
of
income,
increases
in
actual
income
(not
percent
increases)
have
greater
impacts
on
CO2
than
at
larger
household
incomes.
56
The
explanatory
power
of
median
household
income
is
reduced
with
the
addition
of
the
spatial
variable
across
all
models.
However,
median
income
in
the
residence
CO2
model
is
influenced
substantially
less
than
its
counterparts
with
reduction
in
explanatory
power
of
approximately
15%
between
models.
The
total
CO2
and
the
transportation
CO2
models
see
considerable
reduction
however,
approximately
34%
and
67%
respectively.
Despite
this
impact,
median
income
remains
one
of
the
most
important
explanatory
variables
across
the
models,
and
the
theoretical
and
research
background
that
supports
this
assertion
stands
up
to
the
addition
of
spatial
variables,
further
reinforcing
its
significance
as
a
factor
in
household
CO2
emissions.
5.3 Race and ethnicity
Due
to
a
variety
of
institution,
cultural,
and
historical
factors,
the
concentration
of
minority
households
across
the
urban
space
is
not
homogenous.
This
is
well
illustrated
by
Figure
5
and
Figure
6,
the
percentage
of
black
households
and
Asian
households
respectively.
The
two
race
variables
vary
considerably
in
their
explanatory
power
across
the
models.
Both
black
and
Asian
households
test
significantly
in
the
total
CO2
model,
although
Asian
households
fail
to
attain
significance
in
the
transportation
CO2
model
and
are
only
significant
at
the
0.1
level
in
the
residence
CO2
model,
narrowly
missing
significance
with
a
p-‐vale
of
0.059.
However,
the
most
interesting
conclusion
from
this
analysis,
and
contrary
to
the
few
analyses
that
have
taken
household
race
into
consideration
(Estiri,
2013;
Min
et
al.,
2010)
is
the
differing,
and
consistent,
direction
of
influence
that
the
respective
household
race
variables
have.
57
Previous
research,
although
limited,
has
found
a
negative
association
with
minorities
and
CO2.
(Estiri,
2013)
first
examined
race
and
ethnicity
as
a
singular
minority
variable
for
households
across
the
U.S.
using
data
from
the
Residential
Energy
Consumption
Survey
created
by
the
U.S.
Energy
Information
Agency.
The
minority
variable
was
further
broken
down
into
individual
race
and
ethnicities,
but
a
consistent
negative
relationship
was
established
between
all.
(Min
et
al.,
2010)
employ
a
somewhat
similar
model
as
used
in
this
present
analysis
where
several
variables
representing
a
portion
of
minorities
is
used.
Negative
relationships
between
race
and
ethnicity
and
CO2
were
uncovered
there
as
well.
However,
neither
paper
attempts
to
draw
conclusions
or
provide
explanations
for
the
direction
or
strength
of
influence
of
minority
households
on
CO2
emissions.
The
differing
results
found
in
this
study
may
provide
a
clue
at
the
heterogeneity
that
exists
between
cities.
The
previously
mentioned
studies
employed
national
databases
while
this
analysis
examined
one
particular
city.
The
spatial
pattern
of
household
race
and
ethnicity
depends
on
a
range
of
forces
that
are
place
dependent,
and,
given
the
uniqueness
of
these
forces
for
each
city,
it
is
perhaps
unsurprising
to
find
the
results
of
this
small-‐scale
study
do
not
follow
that
of
national
averages.
In
the
present
analysis,
this
previously
seen
negative
relationship
between
race
and
ethnicity
and
household
CO2
holds
for
Asian
households,
which
have
consistently
negative
coefficients
across
the
models.
Black
households
however,
have
consistently
positive
coefficients
across
all
three
models,
holding
all
other
variables
constant,
including
spatiality.
At
the
institutional
level,
these
results
imply
that
black
households
are
consistently
and
structurally
living
in
places
that
have
associated
higher
CO2
emissions,
either
through
residence
or
transportation
associated
mechanisms.
Although
the
58
significance
for
residence
CO2
and
black
households
is
tenuous,
this
relationship
could
be
due,
for
example,
to
black
households
living
in
older
neighborhoods
where
heating
and
cooling
demands
may
be
increased
due
to
inefficiencies
in
residence
structure.
It
is
useful
to
point
out
that
both
black
households
and
Asian
households
are
overall
weak
predictors
of
CO2,
however
these
results,
even
controlling
for
the
strong
influence
of
spatiality
in
these
models,
concur
with
previous
research
that
race
is
at
least
a
slight
factor
in
household
CO2
emissions.
5.4 Other variables
Median
age
and
average
household
size
are
utilized
in
this
model
as
controlling
factors.
But,
these
variables
are
also
strongly
linked
to
important
socioeconomic
variables.
Indeed,
issues
of
severe
multicollinearity
required
that
certain
variables,
such
as
number
of
children,
had
to
be
excluded
from
the
model.
The
number
of
children
per
household
is
strongly
multicollinear
with
household
size.
As
families
grow
and
more
space
is
required,
they
may
potentially
seek
larger
and
cheaper
housing
away
from
the
center
of
the
city,
a
dual
effect
of
raising
CO2
by
both
transportation
and
residence.
In
the
transportation
CO2
and
residence
CO2
spatial
lag
models,
average
household
size
and
income
have
relatively
the
same
explanatory
power.
In
the
total
CO2
model,
income
becomes
a
stronger
predictor
than
average
household
size,
perhaps
suggesting
that
there
is
some
overlap
in
predictive
power
of
average
household
size
between
the
transportation
and
residence
models,
whereas
the
predictive
power
of
income
across
those
two
models
may
be
picking
up
on
different
patterns.
59
Similarly,
age
of
individuals
living
in
a
household
is
a
significant
influencing
factor
across
all
spatial
models
in
the
present
analysis.
The
influence
of
age
on
household
CO2
emissions
is
not
consistently
agreed
upon
in
the
literature
(Kriström,
2008).
A
positive
significant
relationship
between
age
and
household
CO2
emissions
was
found
in
national
models,
although
the
influence
of
this
variable
was
consistently
weak
(Estiri,
2013;
Glaeser,
2012;
Golley
&
Meng,
2012;
Min
et
al.,
2010).
(Lutzenhiser
&
Hackett,
1993)
found
that
among
older
households
there
were
higher
per
capita
household
square
footage,
ostensibly
associated
with
children
leaving
while
aging
parents
stay
in
the
same
house.
Higher
square
footages
are
associated
with
higher
emissions.
This
is
backed
up
by
(Estiri,
2013)
analysis
that
found
age
to
be
positively
associated
with
per
capita
emissions
and
emissions
per
square
foot.
These
two
studies
support
the
conclusion
that
age
is
positively
associated
with
household
CO2
emissions.
On
the
other
end,
it
can
be
reasonably
agreed
that
at
some
point
age
negatively
influences
transportation
CO2
emissions
since
the
elderly
drive
much
less
(Okada,
2012).
This
type
of
age
influence,
while
a
plausible
mechanism,
is
more
than
likely
not
captured
the
particular
analysis
employed
in
this
paper.
The
lowest
median
household
age
in
Marion
County
census
tracts
is
around
22,
while
the
oldest
is
52.
In
actuality
the
model
presented
in
this
study
has
more
potential
in
capturing
the
change
as
young
adults
begin
buying
houses.
If
this
analysis
included
wealth,
or
total
assets
per
household,
the
results
might
be
slightly
different,
as
housing
as
a
part
of
wealth
would
account
for
some
of
this
activity.
60
5.5 Policy implications
Research
into
the
creation
of
household
CO2
has
increasingly
looked
past
the
technical
aspects
of
households
and
started
looking
more
closely
at
the
behavioral
and
socioeconomic
structural
factors
that
contribute
to
CO2
emissions.
This
analysis
attempted
to
analyze
the
contribution
of
socioeconomic
factors
to
household
CO2
emissions
within
a
particular
urban
setting,
Indianapolis,
IN,
while
specifically
considering
the
influence
of
space
and
the
interactions
between
neighborhoods.
The
results
of
this
study
may
not
be
generalizable
to
other
urban
areas,
and
indeed,
one
would
expect
that
depending
on
the
urban,
cultural,
and
historical
structure
of
the
city
that
the
analysis
would
indeed
change
between
cities.
This
restriction
was
known
at
the
start
of
this
research.
If
it
is
difficult
to
find
consistency
and
agreement
among
studies
on
household
variables
and
CO2
emissions
at
broader
scales
(Kriström,
2008),
it
is
fair
to
assume
that
variation
will
be
visible
at
smaller
scales
as
well.
Still,
the
present
study
does
represent
a
further
step
in
understanding
the
drivers
of
CO2
emissions,
even
as
the
generalizability
of
these
exact
results
are
questionable.
Small-‐scale
analyses
of
household
CO2
emissions
have
not
been
done
so
this
research
offers
insight
and
first
steps
to
gaining
a
clearer
picture
of
the
factors
that
drive
emissions.
As
such,
there
are
interesting
and
important
conclusions
to
be
derived
from
these
results.
First,
the
continuing
importance
of
particular
socioeconomic
factors,
especially
income,
as
supported
by
previous
research,
but
confirmed
at
the
neighborhood
scale
of
a
city
in
this
analysis.
Second,
the
inclusion
of
spatiality
in
modeling
household
CO2
61
emissions
is
unique
at
this
scale.
This
study
demonstrates
the
need
for
the
integration
of
spatiality
into
the
understanding
of
small-‐scale
CO2
emissions.
This
analysis
focused
on
household
emissions
and
not
per
capita
emissions
because
policy
undertaken
at
the
municipal
level
may
be
more
focused
on
household
level
policies,
instead
of
targeting
the
individual.
The
distinction
between
per
capita
and
household
level
analyses
is
important
because
the
conclusions
reached
can
be
dramatically
different.
Previous
research
indicates
that
household
size
has
a
negative
relationship
with
per
capita
CO2
emissions
(Kriström,
2008).
Analyses
of
residential
energy
and
CO2
emissions
has
shifted
to
households
because
these
components
are
shared
among
all
household
members
(de
Sherbinin
et
al.,
2007).
This
makes
sense
instinctually;
more
people
living
in
one
residence
reduces
the
CO2
emissions
per
person.
Household
level
policies
may
include
efficiencies
to
residence
or
subsidies
for
electric
cars
for
example.
This
entire
analysis
is
not
to
say
that
we
should
advocate
reducing
household
incomes,
for
example,
as
a
deterrence
to
increased
CO2
emissions.
This
analysis
indicates
how
policy
could
potentially
be
targeted,
or
may
be
best
targeted,
to
reduce
emissions.
That
is,
if
income
is
a
significant
explanatory
variable
of
total
household
CO2
emissions
then
perhaps
policy
makers
should
be
directing
their
climate
change
mitigation
policies
at
this
particular
group.
Tax
cuts
for
household
energy
efficiency
upgrades
is
one
potential
example.
The
recent
sustainable
and
environmental
urban
planning
community
has
advocated
ideas
such
as
walkable
cities.
By
integrating
residential
and
commercial
areas
more
fluidly,
household
transportation
needs
are
reduce.
By
advocating
for
denser
living,
62
residence
sizes
are
also
substantially
reduced.
Overall
CO2
emissions
associated
with
the
household
may
be
reduced
(Bulkeley
et
al.,
2010).
Race
and
ethnicity
has
a
much
weaker
influence
on
household
CO2
emissions
than
income,
but
it
is
useful
to
see
how
policy
makers
may
utilize
this
type
of
information.
The
clustering
of
black
households
in
a
certain
area
of
Indianapolis
may
be
a
prime
target
for
neighborhood
level
policies,
especially
considering
the
positive
association
between
black
households
and
CO2
emissions.
Policy
directed
here
may
first
want
to
gather
additional
information
to
obtain
a
clearer
understanding
of
the
factors
behind
this
relationship.
Perhaps
public
transportation
needs
to
be
improved
in
the
area
to
reduce
associated
household
transportation
CO2.
Perhaps
this
area
of
the
city
is
associated
with
older
homes,
and
thus
household
energy
efficiency
policies,
and
associated
education
outreach
of
said
policies,
may
be
the
best
direction
for
policy
makers.
The
type
of
information
to
come
out
of
an
analysis
such
as
presented
in
this
paper
should
be
paired
with
other
types
of
knowledge
as
well.
These
policy
recommendations
are
not
actual
prescriptions
to
action;
instead,
they
are
examples
of
how
this
type
of
information
may
be
valuable
in
the
policy
making
process.
In
general,
there
has
been
a
focus
on
the
technical
and
physical
aspects
of
CO2
emissions
and
this
has
been
reflected
in
policy
as
well.
The
results
of
this
analysis
add
credence
to
the
recent
literature
that
demonstrates
the
influence
of
socioeconomics
factors
on
household
CO2,
as
well
as
demonstrating
the
way
in
which
these
socioeconomic
variables
can
be
modeled
at
small
spatial
scales,
a
scale
that
would
be
useful
to
policymakers.
63
Chapter 6: Conclusion
6.1 Conclusion
Climate
change
is
perhaps
the
single
most
important
issue
facing
the
world
community,
and
the
Earth
at
large,
this
century.
The
causes
and
impacts
are
global,
crossing
international
boundaries
through
trade
and
the
diffuse
nature
of
CO2
emissions.
Solutions
at
the
international
level
have
predominately
failed
to
bring
about
changes
in
emissions
required
to
see
significant
reduction
in
the
expected
impacts.
However,
these
failures
have
spurred
action
at
lower
levels
of
government.
In
the
U.S.
this
has
meant
states
and
local
jurisdictions
are
crafting
policy
that
addresses
climate
change.
This
analysis
was
not
performed
as
a
thought
experiment
over
potential
local
climate
policy.
Local
jurisdictions
are
taking
actions
now,
and
thus
there
is
a
need
for
accurate
information
on
the
nature
of
CO2
emissions.
While
the
city
examined
in
this
study,
Indianapolis,
does
not
currently
have
comprehensive
climate
change
policy,
this
is
the
type
of
information
that
may
be
utilized
by
local
jurisdictions.
Thus,
this
study
was
a
first
attempt
to
create
this
type
of
analysis,
to
demonstrate
how
it
could
be
done,
the
types
of
questions
that
could
be
answered,
and
the
potential
uses
of
this
data.
Because
of
the
specific
and
non-‐randomized
nature
of
this
study,
the
results
of
the
analysis
are
not
necessarily
generalizable
to
other
cities
or
jurisdictions.
However,
it
is
in
the
author’s
opinion
that
although
there
may
be
changes
among
the
direction
and
significance
of
variables
between
cities,
as
physical
and
social
structures
will
vary,
patterns
of
spatiality
will
still
exist.
Spatiality
is
a
concept
that
recognizes
not
just
the
differences
in
64
socioeconomic
groups
across
the
city
space,
but
also
in
the
biophysical
structure
as
well,
and
thus
this
study
demonstrates
the
need
to
include
that
influence
in
analysis
of
household
CO2.
These
findings
recognize
that
particular
places
in
the
city
may
be
more
prone
to
higher
or
lower
CO2
emissions
either
due
to
biophysical
or
socioeconomic
factors,
and
should
be
considered
in
policy
decisions.
Furthermore,
although
it
cannot
be
said
for
certain
that
the
patterns
for
race
and
ethnicity
that
were
observed
for
this
study
will
hold
for
other
cities,
the
concept
of
housing
discrimination
is
well
documented
and
has
been
observed
across
the
country.
It
is
thus
not
a
stretch
to
say
that
race
and
ethnicity
will
be
a
significant
factor
in
other
cities
as
well,
and
should
be
taken
into
consideration
in
other
models.
The
other
variables
analyzed
in
this
study
have
much
more
previous
research
to
back
up
their
inclusion
in
future
work.
Positive
and
significant
relationships
between
total
household
CO2
emissions
were
observed
between
two
of
the
variables
of
main
interest
to
this
study
as
a
main
components
of
socioeconomic
status,
income
and
black
households.
Although
a
controlling
variable,
average
household
size
is
also
strongly
associated
with
total
CO2.
Income
has
a
long
established
positive
relationship
and
is
backed
by
this
analysis.
Black
households
revealed
a
positive
relationship
with
CO2
that
is
contrary
to
previous
results.
Coinciding
with
previous
research,
Asian
households
have
a
negative
relationship
with
household
CO2.
This
emphasizes
the
individual
nature
of
urban
space
and
individual
cities,
as
well
as
the
individual
nature
in
relationships
between
different
races
and
emissions.
The
historical
policies
and
structure
of
the
city
has
meant
certain
populations
have
clustered
in
either
areas
that
are
associated
with
either
higher
or
lower
CO2
Overall
this
study
confirmed
the
65
initial
research
questions,
identifying
the
importance
of
spatiality,
and
the
importance
of
socioeconomic
indicators
such
as
income
and
race
and
ethnicity
as
factors
of
household
CO2
emissions.
6.2 Interdisciplinary Statement
An
issue
of
climate
change’s
complexity
and
breadth
requires
research
not
just
on
the
biophysical
causes
and
impacts,
but
from
the
socioeconomic,
cultural,
and
political
causes
and
impacts
as
well.
The
integration
of
all
these
disciplines
is
difficult,
but
necessary
for
accurate
climate
research
and
policy.
Disciplinary
intersection
can
be
seen
in
the
IPCC,
a
group
of
scientists
from
around
the
world
who
use
complex
biophysical
models
of
the
Earth
to
predict
and
understand
future
impacts,
while
integrating
multiple
socioeconomic
and
political
scenarios
into
these
models.
There
is
tacit
recognition
of
the
need
for
all
types
of
science
and
policy
within
climate
change
research.
It
is
in
this
spirit
that
this
analysis
has
attempted
to
merge
CO2
and
socioeconomic
data.
The
importance
of
geography
in
this
research
is
not
just
in
ensuring
an
accurate
and
robust
regression
analysis,
but
as
a
tacit
acknowledgement
that
spatial
influence
exists
within
the
urban
form
and
is
an
important
variable
to
consider
in
policy.
The
analysis
also
borrows
from
the
economic
and
sociology
literature,
in
the
examination
of
CO2
and
income
and
the
relationship
between
race,
ethnicity,
housing,
and
transportation.
The
data
that
was
utilized
for
this
analysis
was
created
by
downscaling,
a
complex
process
that
utilized
GIS
and
energy
modeling.
Finally,
part
of
the
intent
of
this
research
to
create
a
method
to
66
extract
relevant
information
from
CO2
emission
data
that
could
potentially
inform
climate
change
mitigation
policy
making
on
the
local
level.
The
integration
of
all
these
disciplines
was
paramount
to
the
success
of
this
analysis.
6.4 Recommendations
Several
opportunities
for
improvement
on
this
study
for
future
consideration
come
immediately
to
mind.
Specifically,
as
discussed
in
section
3.4,
there
are
limitations
to
the
data,
limitations
that,
if
managed
correctly,
could
improve
the
robustness
and
accuracy
of
analysis.
Additionally,
time-‐series
analyses
could
help
identify
trends
in
changing
socioeconomic
and
physical
structures
of
cities,
considering
especially
the
movement
of
populations
between
and
away
from
neighborhoods.
Both
these
recommendations
require
improvements
in
data
gathering
or
modeling
efforts
as
well
in
complex
analysis,
and
would,
in
general,
require
larger
amounts
of
data.
It
is
also
recommended
that
further
studies
consider
per
capita
emissions,
as
well
as
household
level
emissions
in
the
analyses.
There
is
a
well-‐established
relationship
wherein
the
marginal
increase
in
household
CO2
emissions
decreases
with
each
additional
member
of
the
household.
Overall
emissions
are
higher
in
this
case,
but
per
capita
emissions
are
actually
lower.
This
research
controlled
for
household
size,
and
so
the
relationships
among
all
other
variables
still
stand,
but
it
would
be
of
interest
to
see
how
per
capita
CO2
emissions
vary
over
the
urban
space.
While
this
paper
focused
solely
on
households
as
the
unite
of
analysis
in
consideration
of
household-‐level
focused
policy
decisions,
relevant
67
information
could
be
derived
from
per
capita
analyses
as
well,
and
both
should
be
carried
out
side-‐by-‐side
in
future
work.
68
Bibliography
Aldy,
J.
E.
(2005).
An
Environmental
Kuznets
Curve
Analysis
of
U.S.
State-‐Level
Carbon
Dioxide
Emissions.
The
Journal
of
Environment
&
Development,
14(1),
48–72.
doi:10.1177/1070496504273514
Anselin,
L.
(1988).
Spatial
Econometrics:
Methods
and
Models.
Springer.
Antipova,
A.,
Wang,
F.,
&
Wilmot,
C.
(2011).
Urban
land
uses,
socio-‐demographic
attributes
and
commuting:
A
multilevel
modeling
approach.
Applied
Geography,
31(3),
1010–
1018.
doi:10.1016/j.apgeog.2011.02.001
Betsill,
M.
(2001).
Mitigating
climate
change
in
US
cities:
opportunities
and
obstacles.
Local
environment,
6(4),
393–406.
Betsill,
M.,
&
Bulkeley,
H.
(2007).
Looking
Back
and
Thinking
Ahead:
A
Decade
of
Cities
and
Climate
Change
Research.
Local
Environment,
12(5),
447–456.
doi:10.1080/13549830701659683
Boswell,
M.
R.,
Greve,
A.
I.,
&
Seale,
T.
L.
(2010).
An
Assessment
of
the
Link
Between
Greenhouse
Gas
Emissions
Inventories
and
Climate
Action
Plans.
Journal
of
the
American
Planning
Association,
76(4),
451–462.
doi:10.1080/01944363.2010.503313
Brownstone,
D.,
&
Golob,
T.
F.
(2009).
The
impact
of
residential
density
on
vehicle
usage
and
energy
consumption.
Journal
of
Urban
Economics,
65(1),
91–98.
doi:10.1016/j.jue.2008.09.002
Bulkeley,
H.,
Broto,
V.
C.,
Hodson,
M.,
&
Marvin,
S.
(2010).
Cities
and
Low
Carbon
Transitions.
Taylor
&
Francis.
Burnett,
J.
W.,
&
Bergstrom,
J.
C.
(2010).
U.S.
State-‐Level
Carbon
Dioxide
Emissions:
A
Spatial-‐
Temporal
Econometric
Approach
of
the
Environmental
Kuznets
Curve
(Faculty
Series
No.
96031).
University
of
Georgia,
Department
of
Agricultural
and
Applied
Economics.
Retrieved
from
http://econpapers.repec.org/paper/agsugeofs/96031.htm
Canadell,
J.
G.,
Quéré,
C.
L.,
Raupach,
M.
R.,
Field,
C.
B.,
Buitenhuis,
E.
T.,
Ciais,
P.,
…
Marland,
G.
(2007).
Contributions
to
accelerating
atmospheric
CO2
growth
from
economic
activity,
carbon
intensity,
and
efficiency
of
natural
sinks.
Proceedings
of
the
National
Academy
of
Sciences,
104(47),
18866–18870.
doi:10.1073/pnas.0702737104
Charles,
C.
Z.
(2003).
The
dynamics
of
racial
residential
segregation.
Annual
Review
of
Sociology,
167–207.
69
Churkina,
G.
(2008).
Modeling
the
carbon
cycle
of
urban
systems.
Ecological
Modelling,
216(2),
107–113.
doi:10.1016/j.ecolmodel.2008.03.006
De
Sherbinin,
A.,
Carr,
D.,
Cassels,
S.,
&
Jiang,
L.
(2007).
Population
and
Environment.
Annual
Review
of
Environment
and
Resources,
32,
345–373.
doi:10.1146/annurev.energy.32.041306.100243
Dodman,
D.
(2009).
Blaming
cities
for
climate
change?
An
analysis
of
urban
greenhouse
gas
emissions
inventories.
Environment
and
Urbanization,
21(1),
185–201.
Doyle,
D.
G.,
&
Taylor,
B.
D.
(2000).
Variation
in
Metropolitan
Travel
Behavior
by
Sex
and
Ethnicity.
Retrieved
from
http://trid.trb.org/view.aspx?id=899770
Eggleston,
S.,
Buendia,
L.,
Miwa,
K.,
Ngara,
T.,
&
Tanabe,
K.
(2006).
IPCC
guidelines
for
national
greenhouse
gas
inventories.
Institute
for
Global
Environmental
Strategies,
Hayama,
Japan.
Retrieved
from
http://library.wur.nl/WebQuery/clc/1885455
Engel,
K.
H.,
&
Orbach,
B.
Y.
(2008).
Micro-‐Motives
and
State
and
Local
Climate
Change
Initiatives.
Harvard
Law
&
Policy
Review,
2,
119.
Estiri,
H.
(2013).
21
Percent:
The
Role
of
Socioeconomics
and
Housing
Characteristics
on
CO2
Emissions
from
the
U.S.
Residential
Sector
(SSRN
Scholarly
Paper
No.
ID
2196984).
Rochester,
NY:
Social
Science
Research
Network.
Retrieved
from
http://papers.ssrn.com/abstract=2196984
Garrett,
M.,
&
Taylor,
B.
(1999).
Reconsidering
social
equity
in
public
transit.
Berkeley
Planning
Journal,
13,
6–27.
Glaeser,
E.
L.
(2012).
The
challenge
of
urban
policy.
Journal
of
Policy
Analysis
&
Management,
31(1),
111–122.
doi:10.1002/pam.20631
Glaeser,
E.
L.,
&
Kahn,
M.
E.
(2010).
The
greenness
of
cities:
Carbon
dioxide
emissions
and
urban
development.
Journal
of
Urban
Economics,
67(3),
404–418.
doi:10.1016/j.jue.2009.11.006
Golley,
J.,
&
Meng,
X.
(2012).
Income
inequality
and
carbon
dioxide
emissions:
The
case
of
Chinese
urban
households.
Energy
Economics,
34(6),
1864–1872.
doi:10.1016/j.eneco.2012.07.025
Gurney,
K.
R.,
Razlivanov,
I.,
Song,
Y.,
Zhou,
Y.,
Benes,
B.,
&
Abdul-‐Massih,
M.
(2012).
Quantification
of
fossil
fuel
CO2
emissions
at
the
building/street
scale
for
a
large
US
city.
Environmental
Science
&
Technology.
Retrieved
from
http://pubs.acs.org/doi/abs/10.1021/es3011282
Heinonen,
J.,
&
Junnila,
S.
(2011).
A
Carbon
Consumption
Comparison
of
Rural
and
Urban
Lifestyles.
Sustainability,
3(8),
1234–1249.
doi:10.3390/su3081234
70
Hillmer-‐Pegram,
K.
C.,
Howe,
P.
D.,
Greenberg,
H.,
&
Yarnal,
B.
(2012).
A
geographic
approach
to
facilitating
local
climate
governance:
From
emissions
inventories
to
mitigation
planning.
Applied
Geography,
34,
76–85.
doi:10.1016/j.apgeog.2011.11.001
Hoornweg,
D.,
Sugar,
L.,
&
Gómez,
C.
L.
T.
(2011).
Cities
and
greenhouse
gas
emissions:
moving
forward.
Environment
and
Urbanization,
23(1),
207–227.
doi:10.1177/0956247810392270
ICLEI.
(2012,
October).
U.S.
Community
Protocol
for
Accounting
and
Reporting
Greenhouse
Gas
Emissions.
ICLEI
-‐
Local
Governments
for
Sustainability
USA.
International
Energy
Agency.
(2008).
World
Energy
Outlook
2008
(p.
569).
Paris:
IEA.
Retrieved
from
http://www.worldenergyoutlook.org/media/weowebsite/2008-‐
1994/WEO2008.pdf
IPCC.
(2007).
Climate
Change
2007:
Synthesis
Report.
Contribution
of
Working
Groups
I,
II
and
III
to
the
Fourth
Assessment
Report
of
the
Intergovernmental
Panel
on
Climate
Change.
IPCC.
Jeanty,
P.
W.
(2010).
SPMLREG:
Stata
module
to
estimate
the
spatial
lag,
the
spatial
error,
the
spatial
durbin,
and
the
general
spatial
models
by
maximum
likelihood.
Boston
College
Department
of
Economics.
Kahn,
M.
E.
(2000).
The
Environmental
Impact
of
Suburbanization.
Journal
of
Policy
Analysis
&
Management,
19(4),
569–586.
Kaza,
N.
(2010).
Understanding
the
spectrum
of
residential
energy
consumption:
A
quantile
regression
approach.
Energy
Policy,
38(11),
6574–6585.
doi:10.1016/j.enpol.2010.06.028
Kennedy,
C.,
Steinberger,
J.,
Gasson,
B.,
Hansen,
Y.,
Hillman,
T.,
Havránek,
M.,
…
Mendez,
G.
V.
(2009).
Greenhouse
Gas
Emissions
from
Global
Cities.
Environmental
Science
&
Technology,
43(19),
7297–7302.
doi:10.1021/es900213p
Kennedy,
C.,
Steinberger,
J.,
Gasson,
B.,
Hansen,
Y.,
Hillman,
T.,
Havránek,
M.,
…
Mendez,
G.
V.
(2010).
Methodology
for
inventorying
greenhouse
gas
emissions
from
global
cities.
Energy
Policy,
38(9),
4828–4837.
doi:10.1016/j.enpol.2009.08.050
Kennedy,
S.,
&
Sgouridis,
S.
(2011).
Rigorous
classification
and
carbon
accounting
principles
for
low
and
Zero
Carbon
Cities.
Energy
Policy,
39(9),
5259–5268.
doi:10.1016/j.enpol.2011.05.038
Kerkhof,
A.
C.,
Benders,
R.
M.
J.,
&
Moll,
H.
C.
(2009).
Determinants
of
variation
in
household
CO2
emissions
between
and
within
countries.
Energy
Policy,
37(4),
1509–1517.
doi:10.1016/j.enpol.2008.12.013
71
Kousky,
C.,
&
Schneider,
S.
H.
(2003).
Global
climate
policy:
will
cities
lead
the
way?
Climate
Policy,
3(4),
359–372.
Krause,
Rachel
M.
(2011).
Policy
Innovation,
Intergovernmental
Relations,
and
the
Adoption
of
Climate
Protection
Initiatives
by
U.S.
Cities.
Journal
of
Urban
Affairs,
33(1),
45–60.
doi:10.1111/j.1467-‐9906.2010.00510.x
Krause,
Rachel
Marie.
(2011).
An
assessment
of
the
greenhouse
gas
reducing
activities
being
implemented
in
US
cities.
Local
Environment,
16(2),
193–211.
Kriström,
B.
(2008).
Residential
Energy
Demand.
OECD
Journal:
General
Papers,
2008(2),
95–115.
Larsen,
H.
N.,
&
Hertwich,
E.
G.
(2009).
The
case
for
consumption-‐based
accounting
of
greenhouse
gas
emissions
to
promote
local
climate
action.
Environmental
Science
&
Policy,
12(7),
791–798.
doi:10.1016/j.envsci.2009.07.010
Le
Quéré,
C.,
Raupach,
M.
R.,
Canadell,
J.
G.,
Al,
G.
M.
et,
Al,
C.
L.
Q.
et,
Al,
C.
L.
Q.
et,
…
Bopp,
L.
(2009).
Trends
in
the
sources
and
sinks
of
carbon
dioxide.
Nature
Geoscience,
2(12),
831–836.
doi:10.1038/ngeo689
Limpert,
E.,
Stahel,
W.
A.,
&
Abbt,
M.
(2001).
Log-‐normal
Distributions
across
the
Sciences:
Keys
and
Clues.
BioScience,
51(5),
341–352.
Lutzenhiser,
L.
(1997).
Social
Structure,
Culture,
and
Technology:
Modeling
the
Driving
Forces
of
Household
Energy
Consumption.
In
Environmentally
Significant
Consumption:
Research
Directions.
National
Academies
Press.
Lutzenhiser,
L.,
&
Hackett,
B.
(1993).
Social
Stratification
and
Environmental
Degradation:
Understanding
Household
CO2
Production.
Social
Problems,
40(1),
50–73.
doi:10.2307/3097026
Metz,
B.
(2007).
Climate
Change
2007
-‐
Mitigation
of
Climate
Change:
Working
Group
III
Contribution
to
the
Fourth
Assessment
Report
of
the
IPCC.
Cambridge
University
Press.
Min,
J.,
Hausfather,
Z.,
&
Lin,
Q.
F.
(2010).
A
High-‐Resolution
Statistical
Model
of
Residential
Energy
End
Use
Characteristics
for
the
United
States.
Journal
of
Industrial
Ecology,
14(5),
791–807.
doi:10.1111/j.1530-‐9290.2010.00279.x
Nakicenovic,
N.,
Alcamo,
J.,
Davis,
G.,
de
Vries,
B.,
Fenhann,
J.,
Gaffin,
S.,
…
Dadi,
Z.
(2000).
Special
Report
on
Emissions
Scenarios:
A
Special
Report
of
Working
Group
III
of
the
Intergovernmental
Panel
on
Climate
Change
(No.
PNNL-‐SA-‐39650).
Pacific
Northwest
National
Laboratory,
Richland,
WA
(US),
Environmental
Molecular
Sciences
Laboratory
(US).
Retrieved
from
http://www.osti.gov/energycitations/product.biblio.jsp?osti_id=15009867
72
Norman,
J.,
MacLean,
H.
L.,
&
Kennedy,
C.
A.
(2006).
Comparing
High
and
Low
Residential
Density:
Life-‐Cycle
Analysis
of
Energy
Use
and
Greenhouse
Gas
Emissions.
Journal
of
Urban
Planning
&
Development,
132(1),
10–21.
doi:10.1061/(ASCE)0733-‐
9488(2006)132:1(10)
Okada,
A.
(2012).
Is
an
increased
elderly
population
related
to
decreased
CO2
emissions
from
road
transportation?
Energy
Policy,
45,
286–292.
doi:10.1016/j.enpol.2012.02.033
Pataki,
D.
E.,
Alig,
R.
J.,
Fung,
A.
S.,
Golubiewski,
N.
E.,
Kennedy,
C.
A.,
McPherson,
E.
G.,
…
Romero
Lankao,
P.
(2006).
Urban
ecosystems
and
the
North
American
carbon
cycle.
Global
Change
Biology,
12(11),
2092–2102.
Peters,
G.,
&
Hertwich,
E.
(2008).
Post-‐Kyoto
greenhouse
gas
inventories:
production
versus
consumption.
Climatic
Change,
86(1),
51–66.
doi:10.1007/s10584-‐007-‐9280-‐
1
Pucher,
J.,
&
Renne,
J.
L.
(2003).
Socioeconomics
of
urban
travel:evidence
from
the
2001
NHTS.
Transportation
Quarterly,
57(3).
Retrieved
from
http://trid.trb.org/view.aspx?id=662423
Ramanathan,
R.
(2006).
A
multi-‐factor
efficiency
perspective
to
the
relationships
among
world
GDP,
energy
consumption
and
carbon
dioxide
emissions.
Technological
Forecasting
and
Social
Change,
73(5),
483–494.
doi:10.1016/j.techfore.2005.06.012
Ramaswami,
A.,
Hillman,
T.,
Janson,
B.,
Reiner,
M.,
&
Thomas,
G.
(2008).
A
Demand-‐
Centered,
Hybrid
Life-‐Cycle
Methodology
for
City-‐Scale
Greenhouse
Gas
Inventories.
Environmental
Science
&
Technology,
42(17),
6455–6461.
doi:10.1021/es702992q
Roscigno,
V.
J.,
Karafin,
D.
L.,
&
Tester,
G.
(2009).
The
Complexities
and
Processes
of
Racial
Housing
Discrimination.
Social
Problems,
56(1),
49–69.
doi:10.1525/sp.2009.56.1.49
Satterthwaite,
D.
(2008).
Cities’
contribution
to
global
warming:
notes
on
the
allocation
of
greenhouse
gas
emissions.
Environment
and
Urbanization,
20(2),
539–549.
doi:10.1177/0956247808096127
Sharp,
E.
B.,
Daley,
D.
M.,
&
Lynch,
M.
S.
(2011).
Understanding
local
adoption
and
implementation
of
climate
change
mitigation
policy.
Urban
Affairs
Review,
47(3),
433–457.
Shen,
Q.
(2000).
Spatial
and
Social
Dimensions
of
Commuting.
Journal
of
the
American
Planning
Association,
66(1),
68–82.
Sippel,
M.,
&
Jenssen,
T.
(2009).
What
About
Local
Climate
Governance?
A
Review
of
Promise
and
Problems.
SSRN
eLibrary.
Retrieved
from
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1514334
73
Swan,
L.
G.,
&
Ugursal,
V.
I.
(2009).
Modeling
of
end-‐use
energy
consumption
in
the
residential
sector:
A
review
of
modeling
techniques.
Renewable
and
Sustainable
Energy
Reviews,
13(8),
1819–1835.
doi:10.1016/j.rser.2008.09.033
Tang,
Z.,
Brody,
S.
D.,
Quinn,
C.,
Chang,
L.,
&
Wei,
T.
(2010).
Moving
from
agenda
to
action:
evaluating
local
climate
change
action
plans.
Journal
of
Environmental
Planning
and
Management,
53(1),
41–62.
doi:10.1080/09640560903399772
Tucker,
M.
(1995).
Carbon
dioxide
emissions
and
global
GDP.
Ecological
Economics,
15(3),
215–223.
doi:10.1016/0921-‐8009(95)00045-‐3
U.S.
Department
of
Commerce
Economics
and
Statistics
Administration.
(2010).
U.S.
Carbon
Dioxide
Emissions
and
Intensities
Over
Time:
A
Detailed
Accounting
of
Industries,
Government
and
Households
|
Economics
and
Statistics
Administration.
Retrieved
from
http://www.esa.doc.gov/Reports/u.s.-‐carbon-‐dioxide
United
Nations
Department
of
Economic
and
Social
Affairs,
Population
Divison.
(2012).
World
Urbanization
Prospects,
the
2011
Revisions.
Retrieved
from
http://esa.un.org/unpd/wup/pdf/WUP2011_Highlights.pdf
United
States
Census
Bureau.
(2000,
April
19).
Census
Tracts
and
Block
Numbering
Areas.
Retrieved
May
12,
2013,
from
http://www.census.gov/geo/www/cen_tract.html
US
EPA,
C.
C.
D.
(n.d.).
U.S.
Greenhouse
Gas
Inventory
Report
(Reports
&
Assessments,).
Retrieved
from
http://www.epa.gov/climatechange/ghgemissions/usinventoryreport.html
VandeWeghe,
J.
R.,
&
Kennedy,
C.
(2007).
A
Spatial
Analysis
of
Residential
Greenhouse
Gas
Emissions
in
the
Toronto
Census
Metropolitan
Area.
Journal
of
Industrial
Ecology,
11(2),
133–144.
doi:10.1162/jie.2007.1220
Ward,
M.
D.,
&
Gleditsch,
K.
S.
(2008).
Spatial
Regression
Models.
SAGE.
Wentz,
E.
A.,
Gober,
P.,
Balling,
R.
C.,
&
Day,
T.
A.
(2002).
Spatial
Patterns
and
Determinants
of
Winter
Atmospheric
Carbon
Dioxide
Concentrations
in
an
Urban
Environment.
Annals
of
the
Association
of
American
Geographers,
92(1),
15–28.
Wier,
M.,
Lenzen,
M.,
Munksgaard,
J.,
&
Smed,
S.
(2001).
Effects
of
Household
Consumption
Patterns
on
CO2
Requirements.
Economic
Systems
Research,
13(3),
259–274.
doi:10.1080/09537320120070149
Zhou,
Y.,
&
Gurney,
K.
(2010).
A
new
methodology
for
quantifying
on-‐site
residential
and
commercial
fossil
fuel
CO2
emissions
at
the
building
spatial
scale
and
hourly
time
scale.
Carbon
Management,
1(1),
45–56.
doi:10.4155/cmt.10.7
74
Appendix A
Moran’s
I
and
Local
Moran’s
I
maps
were
analyzed
in
the
exploratory
spatial
analysis
stage
of
this
study.
The
results
are
presented
below,
with
each
CO2
variable
exhibiting
significant
clustering.
Moran's
I
on
Dependent
Variables
Moran's
I
Total
CO2
0.522
*
Transportation
CO2
0.736
*
Residence
CO2
0.199
*
Table
4:
Moran's
I
test
for
spatial
dependency
among
CO2
variables.
P-‐values
are
based
on
replications
where
significant
p-‐values
denote
spatial
dependency
among
variables.
*
denotes
significance
at
p<.01
Figure
11:
Local
Moran’s
I
on
total
household
CO2
emissions.
Significant
clustering
is
represented
by
the
Moran
clustering
typology.
High-‐high
and
low-‐low
designate
clustering
of
positive
spatial
autocorrelation,
while
high-‐low
and
low-‐high
designate
spatial
outliers
or
negative
spatial
autocorrelation.
Typology
designates
the
core
of
the
spatial
autocorrelation.
Significant
replication
at
p<0.05.
75
Figure
12:
Local
Moran’s
I
on
transportation
CO2
emissions.
The
legend
is
as
described
in
Figure
11.
Figure
13:
Local
Moran’s
I
on
residence
CO2
emissions.
The
legend
is
as
described
in
Figure
11.
76
Appendix B
Local
Moran’s
I
maps
on
the
residuals
of
the
residence
CO2
regression
models
and
transportation
CO2
regression
models
are
presented
below.
Both
the
spatial
lag
models
exhibit
decreased
visual
and
overall
clustering,
and
per
their
calculated
Moran’s
I
statistic
available
in
Table
2,
their
spatial
dependency
among
their
residuals
decreases
over
their
respective
OLS
models.
Figure
14:
Local
Moran’s
I
on
residuals
of
residence
CO2
OLS
regression
model.
Significant
replication
at
p<0.05.
The
legend
is
as
described
in
Figure
11.
77
Figure
15:
Local
Moran’s
I
on
residuals
of
residence
CO2
spatial
lag
regression
model.
Significant
replication
at
p<0.05.
The
legend
is
as
described
in
Figure
11.
Figure
16:
Local
Moran’s
I
on
residuals
of
transportation
CO2
OLS
regression
model.
Significant
replication
at
p<0.05.
The
legend
is
as
described
in
Figure
11.
78
Figure
17:
Local
Moran’s
I
on
residuals
of
transportation
CO2
spatial
lag
regression
model.
Significant
replication
at
p<0.05.
The
legend
is
as
described
in
Figure
11.
79
Appendix C
Actual
CO2
Emissions
The
regression
scatterplots
of
the
spatial
lag
of
transportation
CO2
and
residence
CO2
are
presented
below.
12
11
10
9
8
7
6
5
4
3
Transportation
CO2
Spatial
Lag
Regression
t
CO2
per
household
4
5
6
7
8
9
Predicted
CO2
Emissions
10
11
Figure
18:
The
regression
plot
of
transportation
CO2
emissions
of
the
predicted
model
by
the
observed
emissions
for
the
spatial
lag
regression.
Residence
CO2
Spatial
Lag
Regression
Actual
CO2
Emissions
6
-‐1.0
t
CO2
per
household
5
4
3
2
1
0
-‐0.5
-‐1
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Predicted
CO2
Emissions
Figure
19:
The
regression
plot
of
residence
CO2
emissions
of
the
predicted
model
by
the
observed
emissions
for
the
spatial
lag
regression.
80