University of Southern California Dissertations and Theses

Household carbon footprints: how to encourage adoption of emissions‐reducing behaviors and technologies

(USC Thesis Other)

Household carbon footprints: how to encourage adoption of emissions‐reducing behaviors and technologies

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Household carbon footprints:
how to encourage adoption of emissions-
reducing behaviors and technologies

By Lee V. White

Dissertation submitted for Program Urban Planning and Development
In fulfilment of requirements for Doctor of Philosophy
University of Southern California
May 2018

Dissertation by Lee V. White
2

Acknowledgements

To my parents; for commiserating, for their unwavering support, and for their conviction that
I knew what I was doing.
To my cohort and the Gateway crowd, for providing kind ears and like minds – this wouldn’t
have been nearly as fun and rewarding without you.
To my dissertation committee for always having my back; my chair Dan Mazmanian for his
unfailing encouragement and gentle course corrections; Nicole Sintov for her generous and
enthusiastic mentorship and guidance, and intensive editing of paper drafts; Hilda Blanco for
kind conversations helping me remember the context outside a university; and Adam Rose
for solid advice and a wealth of practical insights.
I would also like to thank Elizabeth Currid-Halkett teaching guidance and inspiration, and TJ
McCarthy for extreme patience and extended help with improving my statistical analysis
skillset. For Chapter 1 I also thank Kurt Newick of the Sierra Club for generous sharing of
several years of city level solar permit fee data. Chapter 2 was partly funded by the US
Department of Energy Grant #DE-OE0000192 and Los Angeles Department of Water and
Power (LADWP), and the authors’ time was also supported by the USC Price School of
Public Policy Schwarzenegger Institute of State and Global Policy. Also for Chapter 2, the
authors’ would like to thank Emil Abdelshehid, Justin Powels, and Adam Chhan at LADWP
for their integral roles in implementing the project; Michael Orosz at USC for guidance and
project oversight; Dan Mazmanian and Ryan Merrill for their input in previous versions of
this draft; and Zachary Manta, Agassi Tran, and Arash Zadeh for assisting in data entry.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
3

Contents
Acknowledgements ................................................................................................................................. 2
Abstract ................................................................................................................................................... 6
1 Introduction .................................................................................................................................... 7
1.1 What this dissertation considers ............................................................................................ 7
1.2 Why I focus on the household and local levels ....................................................................... 8
1.3 Specific household activities selected..................................................................................... 9
1.4 Theoretical contributions and chapter outline ..................................................................... 10
1.4.1 Chapter 1: Drivers of Residential Solar Installations: Do Local Permitting Processes
Have an Effect? ............................................................................................................................. 11
1.4.2 Chapter 2: You are what you drive: Environmentalist and social innovator symbolism
drives electric vehicle adoption intentions (publication co-authored with Nicole D. Sintov) ...... 12
1.4.3 Chapter 3: Controlling household electricity loads: The effect of actual versus
perceived savings on Time-of-Use rate acceptance (co-authored with Nicole D. Sintov) ........... 12
1.5 Conclusion: policy lessons learned ....................................................................................... 13
2 Chapter 1: Drivers of Residential Solar Installations: Do Local Permitting Processes Have an
Effect? ................................................................................................................................................... 16
2.1 Abstract ................................................................................................................................. 16
2.2 Introduction .......................................................................................................................... 16
2.3 Predictors of residential PV adoption ................................................................................... 17
2.4 Method ................................................................................................................................. 18
2.4.1 Empirical approach ....................................................................................................... 19
2.4.2 Streamlined Permitting ................................................................................................. 20
2.4.3 Permitting Fees ............................................................................................................. 21
2.4.4 Potential model issues .................................................................................................. 23
2.4.5 Dependent variable: Rate of PV system installation ..................................................... 24
2.4.6 City characteristics ........................................................................................................ 25
2.4.7 Utility Characteristics .................................................................................................... 25
2.5 Analysis ................................................................................................................................. 26
2.6 Conclusions and Policy Implications ..................................................................................... 29
2.7 Limitations and future research ............................................................................................ 30
3 Chapter 2: You are what you drive: Environmentalist and social innovator symbolism drives
electric vehicle adoption intentions (publication co-authored with Nicole D. Sintov)......................... 31
3.1 Abstract ................................................................................................................................. 31
3.2 Introduction .......................................................................................................................... 31
Dissertation by Lee V. White
4

3.2.1 Contributions ................................................................................................................ 32
3.2.2 Predictors of EV adoption intentions ............................................................................ 33
3.3 Methods ................................................................................................................................ 41
3.3.1 Procedures .................................................................................................................... 41
3.3.2 Participants ................................................................................................................... 41
3.3.3 Measures ....................................................................................................................... 43
3.3.4 Data preparation ........................................................................................................... 45
3.4 Results ................................................................................................................................... 46
3.4.1 Factor analysis of symbolic attributes .......................................................................... 46
3.4.2 Factors impacting EV adoption intentions .................................................................... 47
3.4.3 Concern about climate change: mediation ................................................................... 50
3.4.4 Normative messaging intervention ............................................................................... 52
3.5 Discussion .............................................................................................................................. 53
3.5.1 Specification of symbolic attributes reflecting aspects of self-identity ........................ 53
3.5.2 The predictive strength of symbolic attributes reflecting self-identity ........................ 53
3.5.3 Concern about climate change: mediation ................................................................... 54
3.5.4 Effects of messaging...................................................................................................... 55
3.6 Conclusion ............................................................................................................................. 55
3.7 Limitations and Future Directions ......................................................................................... 56
4 Chapter 3: Controlling household electricity loads: The effect of actual versus perceived savings
on Time-of-Use rate acceptance (co-authored with Nicole D. Sintov) ................................................. 57
4.1 Abstract ................................................................................................................................. 57
4.2 Introduction .......................................................................................................................... 57
4.2.1 Perceptions affecting TOU acceptance ......................................................................... 58
4.3 Methods ................................................................................................................................ 58
4.3.1 Procedures .................................................................................................................... 58
4.3.2 Measures ....................................................................................................................... 59
4.3.3 Participants ................................................................................................................... 63
4.3.4 Analyses ........................................................................................................................ 65
4.4 Findings ................................................................................................................................. 66
4.4.1 Correlations of perceived savings with actual changes in usage and bills .................... 67
4.4.2 Modeling Acceptance of TOU ....................................................................................... 68
4.4.3 Mediation Analysis ........................................................................................................ 71
4.5 Discussion .............................................................................................................................. 73
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
5

5 References .................................................................................................................................... 75
6 Chapter 1 Article Appendices........................................................................................................ 90
6.1 Chapter 1, Appendix 1: Fixed Effects Analysis of PV adoption rates .................................... 90
7 Chapter 1, Supplementary Appendix: Synthetic Control Analysis of PV adoption rates .............. 91
7.1 Method ................................................................................................................................. 91
7.2 Results ................................................................................................................................... 91
8 Chapter 2 Article Appendices........................................................................................................ 94
8.1 Chapter 2, Appendix 1: Descriptive statistics for EV adoption survey sample ..................... 94
8.2 Chapter 2, Appendix 2: Scree plot for symbolic attributes of EVs ........................................ 94
8.3 Chapter 2, Appendix 3: Factor loadings of symbolic attributes for EVs ............................... 96
9 Chapter 3 Article Appendices........................................................................................................ 97
9.1 Chapter 3, Appendix 1: Models of TOU acceptance with unstandardized coefficients ....... 97
9.2 Chapter 3, Appendix 2: Mediation analyses with Rate Satisfaction as the Dependent
Variable ............................................................................................................................................. 98
10 Chapter 3, Supplementary Appendix: Additional predictors of TOU acceptance included in
initial analyses, to be formed into a second journal publication .......................................................... 99
10.1 Initial Abstract ....................................................................................................................... 99
10.2 Initial Literature Review Framing for Perceived vs. Actual Savings ...................................... 99
10.2.1 Usage patterns .............................................................................................................. 99
10.2.2 Perceived control ........................................................................................................ 100
10.3 Additional Segments from Initial Literature Review ........................................................... 103
10.3.1 Income ........................................................................................................................ 103
10.3.2 Enabling technology .................................................................................................... 103
10.3.3 Environmental concern ............................................................................................... 104
10.4 Methodology ....................................................................................................................... 104
10.4.1 Procedures .................................................................................................................. 104
10.4.2 Actual usage and costs ................................................................................................ 104
10.4.3 Perceptions ................................................................................................................. 104
10.4.4 Demographics ............................................................................................................. 105
10.4.5 Physical constraints ..................................................................................................... 105
10.4.6 Initial analyses ............................................................................................................. 106
10.4.7 Initial results ................................................................................................................ 107
10.4.8 Initial Discussion .......................................................................................................... 111
10.4.9 Initial Conclusion ......................................................................................................... 112
Dissertation by Lee V. White
6

Abstract
Households account for over a third of CO 2-e emissions in the US, and in aggregate can
contribute significantly to efforts to mitigate climate change by reducing the carbon footprint
associated with their daily activities. However, we do not yet fully understand what drives
households to adopt new and emerging technologies to reduce emissions, nor what
encourages households to adopt new behaviors that can reduce emissions. Decisions at the
household level can be affected by policies at the national, state, and local level offering
market-based incentives or streamlining regulatory environments, and can simultaneously be
affected by the personal values, beliefs, and self-identities of household members. This
dissertation examines household adoption of solar photovoltaic panels, electric vehicles, and
new electricity use behaviors, broadly framing analyses within the attitude-behavior-context
theory to consider both personal level drivers and contextual policy and societal drivers of
adoption.
I conclude that household decisions to take part in climate change mitigation efforts are
powerfully driven by perceptions of individual benefits including financial savings and
ability to showcase care for the environment. Local governments may be able to remove
contextual barriers to solar adoption, but this could not be confirmed nor disproved in the
current analysis due to limitations on available city-level data. Adoption of electric vehicles
and new electricity use behaviors was confirmed to be powerfully driven by perceptions of
symbolism and savings, respectively. Overall, findings emphasize the need to include
consideration of personal factors when developing policy settings, rather than relying on
purely financial instruments. Findings also emphasize the need to include mechanisms to
strengthen household understanding of energy usage and costs, to facilitate stronger
economic evaluations by households of their own energy use decisions.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
7

1 Introduction
Anthropogenic climate change is a defining environmental issue of the 21
st
century, and the
energy supply sector is the largest contributor to greenhouse gas (GHG) emissions (Brucker
et al., 2014). Climate change cannot be addressed without fundamental changes in the
electricity sector, including a shift to low-GHG generation technologies (Brucker et al.,
2014). An electricity system reliant on 80% renewable generation capacity is theoretically
feasible in the US, but this change would require a more flexible electricity system (NREL,
2012); that is, the way that energy is used in society will need to change to incorporate
measures such as smart grids and demand-side response (IEA, 2011).
Emissions of carbon dioxide and other gases with direct or indirect radiative forcing
potential, known collectively as GHGs, have caused significant observable warming in the
global climate system since the 1950s (IPCC, 2013). Continued climate change is expected to
significantly disrupt societies around the world in addition to causing irreversible ecosystem
harm including species loss. Impacts on human society include rising sea levels impacting
coastal settlements by damaging infrastructure and housing, less predictable rainfall
decreasing food crop outputs, more extreme weather events increasing the incidence of
severe natural disasters such hurricanes, and an overall warming trend increasing the
geographical spread of insect borne diseases (IPCC, 2014).
Changes aimed at reducing GHG emissions associated with energy use span all levels of
government and all sectors of society. Policies can be implemented at national, regional, or
local levels of government, can target changes using market-based incentives or regulations,
and can target industrial, utility, agricultural, and residential sectors. Actions to reduce
emissions can range from replacing utility scale coal plants with fields of solar panels to
encouraging installation of distributed solar on individual residences, and from structural
changes such as cap and trade policies to individual decisions to buy locally made products
with lower embodied transport energy. These changes at all levels of society will be
necessary for an eventual shift to a system that does not require oil or coal for fuel. The faster
these changes can be made, the more likely it is that emissions mitigation efforts will be
successful in keeping warming below the 2
o
C threshold, beyond which climate feedback
effects will accelerate climactic changes in potentially dangerous ways (IPCC, 2013).
1.1 What this dissertation considers
In this dissertation, I focus on examining energy technology choices in the household sector.
Households are where many energy use decisions come together, and in the US the
residential sector accounts for 27% of CO 2-e emissions while passenger cars account for 43%
of transport CO2-e emissions on top of that (US Environmental Protection Agency, 2015a).
Dietz et al. (2009) have estimated that national implementation of measures to support
household behavior change and use of already existing energy-efficiency technologies could
eliminate 7.4% of overall US carbon emissions.
Thus, it is critical to understand how household decisions to adopt new energy technologies
arise. Household decisions are expected to be influenced by market-based and regulatory
policies introduced at the national, regional, and city levels, and by personal factors including
identity, perceptions, and altruism of household members.
Market-based policies at all levels of government assume that household decisions will be
economically rational; for example, if the cost of a residential solar system is reduced (or the
Dissertation by Lee V. White
8

remuneration for electricity sold from it is increased) so that a positive return on investment
can be attained, then it is expected that the rate of household installations of solar would rise.
This was the logic behind the rates set for Germany’s feed-in tariffs, which were set based on
the cost to generate electricity from residential solar panels (Mendonça, Jacobs, & Sovacool,
2010). Financial incentives are also common throughout the US at both the city and state
level (Li & Yi, 2014; REN21, 2014).
Policies can also aim to reduce barriers or increase convenience associated with adopting
emissions-reducing technology, such as policies to support adoption of electric vehicles by
allowing their drivers to use high-occupancy vehicle lanes – a policy which has been
associated with higher adoption of these vehicles (Sheldon & DeShazo, 2017). Examinations
of household adoption of green technologies have long considered that removing barriers and
increasing convenience through regulatory steps or specific policy settings are extremely
important steps to take alongside providing incentives (Paul C. Stern, 1999).
A large body of literature additionally discusses the impact that personal-level attitudinal
factors can have on decisions to adopt green technology. Specifically, altruism, perceptions,
and self-identify can be powerful drivers of pro-environmental behavior (Steg, Bolderdijk,
Keizer, & Perlaviciute, 2014; E. Van der Werff, Steg, & Keizer, 2014; Ellen Van der Werff
& Steg, 2015), above and beyond financial motivations. These findings have drawn on long-
established psychological frameworks including the Value-Belief-Norms theory and the
Norm Activation Model (Schwartz, 1977; P. C. Stern, 2000).
My dissertation focuses on policies that either are implemented at the city level or that
feasibly could be implemented at the city level, as this level is a much more agile policy
arena compared with the state or national level and is able to implement policy changes much
more quickly. I focus on evaluating household decisions to adopt emissions-reducing energy
technologies and related behaviors, considering impacts of the policy and attitudinal factors
outlined above. Specifically, the first chapter considers both market incentives and attempts
to streamline regulatory environments, the second chapter considers personal evaluations of
both costs and ability of emissions-reducing technologies to positively reflect self-identity
alongside convenience measures, and the third chapter considers both actual financial returns
and perceived financial returns of adopting a pro-environmental behavior with additional
measures evaluating perceived convenience of the behavior.
1.2 Why I focus on the household and local levels
Since time is an important factor in mitigation efforts, strategies for change at multiple levels
must be pursued simultaneously. There have been many efforts over past decades to craft an
internationally unified agreement to address climate change, but the agreements tentatively
reached have been too limited to substantially reduce GHG emissions (Rosen, 2015). One
major issue with reaching agreements at an international level is achieving consensus
between countries on which measure of equity should be used to allocate emission budgets
between countries (Rose, Stevens, Edmonds, & Wise, 1998). Many countries and regions
have proceeded to implement policies to reduce greenhouse gas emissions even in absence of
an international commitment (REN21, 2014). In the United States (US), many of these
policies have targeted utilities to increase their shares of renewable generation, although other
countries such as Germany have placed more of an emphasis on household generation
contributions (Hoppmann, Huenteler, & Girod, 2014).
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
9

Extensive research exists examining policy efforts to mitigate GHG emissions at the national
and state levels (Hoppmann et al., 2014; Menz & Vachon, 2006; REN21, 2014; Wiser,
Namovicz, Gielecki, & Smith, 2007; Yin & Powers, 2010), but sparsity of data has slowed
examination of the effectiveness of policies at the municipal or regional level (Feiock,
Krause, Hawkins, & Curley, 2013; Millard-Ball, 2012). There is also extensive literature
examining what drives pro-environmental behavior at the household level (Abrahamse, Steg,
Vlek, & Rothengatter, 2005; Kastner & Stern, 2015; Steg et al., 2014; Ellen Van der Werff,
Steg, & Keizer, 2013b), but as new technologies emerge and new policies are trialed it is
imperative to build on this behavioral literature. Moreover, stronger bridges are needed
between policy and behavioral literatures pertaining to GHG mitigation. In particular, there
has been a call to consider people as actors from a multitude of perspectives including as
energy consumers, as recipients of policy incentives and targeting, and as potential producers
of electricity at the micro-scale (Paul C. Stern, 2014).
This dissertation contributes to forming such a bridge, and focuses on the nexus of household
energy decisions and local policies to support GHG emission reductions. In this dissertation I
provide insights into how people within household units make decisions regarding energy
use, with a consideration of which policy settings may facilitate greater emissions reduction
in future. I consistently use empirical methods to evaluate the strongest predictors of
environmentally beneficial decisions. The household and local level are often neglected when
considering emissions reductions, despite households accounting for well over a third of
electricity consumption (EIA, 2016). My dissertation addresses this gap from planning,
policy, and behavioral perspectives, providing insights critical to understanding how to
transition to a low-carbon society in the near-future.
1.3 Specific household activities selected
Emissions associated with household energy use can be reduced when people install
distributed renewable generation technologies on their homes, when people switch their mode
of transportation to no longer rely on fossil fuels, and when people change their behavior to
reduce the energy they need to accomplish daily tasks or to use it at times better aligned to
renewable electricity generation peaks. This dissertation examines three of these activities
that can be performed by people within household units; (1) installation of solar photovoltaic
(PV) panels, (2) adoption of electric vehicles (EVs), and (3) change of electricity use habits
prompted by time-of-use (TOU) rates. The three activities selected for examination were
chosen due to both their current promotion through policy efforts and their potential to
change energy systems synergistically.
Residential PV, in addition to having the ability to displace household electricity use from the
grid and thereby reduce the emissions footprint of a household, has the potential to supply the
grid at times when air conditioning use and other mid-day activities demand large amounts of
electricity. Germany and California, among other jurisdictions, have aggressively promoted
residential PV through feed-in tariff and net metering policies respectively (CEC & CPUC,
2016; Hoppmann et al., 2014). California provides an excellent testbed to examine issues
beyond system cost affecting the spread of residential PV, which may provide insights to the
spread of other distributed generation and emerging energy technologies in future. In the first
chapter I examine whether local permitting processes may pose a barrier to residential PV
installations, laying groundwork for future examinations of local policy efforts to promote
renewable and distributed technologies.
Dissertation by Lee V. White
10

EVs have the potential to displace demand for gasoline-fueled cars. If adopted en masse, EVs
could drastically reduce transport emissions. As such they are being actively promoted within
California and many other jurisdictions, both for GHG reduction and air quality goals (IEA,
2012). California Governor Jerry Brown recently formalized a commitment to put 5 million
electric vehicles on the state’s roads by 2030 (Dillon, 2018). However, even as EV
technology begins to mature there still remain questions about what drives people to adopt an
EV rather than a gasoline vehicle. Further understanding these drivers not only provides
insights into EV adoption, but may increase knowledge of what motivates people to buy
green durable goods in general. EVs also have the potential to be utilized as demand-side
response tools, since they can be set to charge only when there is ample supply in the grid
that may otherwise require ramping down of generation supply despite available wind or
solar resources.
Household willingness to change electricity use habits facilitates demand-side response,
which is critical for increasing the flexibility of electric grids. This flexibility will be
necessary to smooth grid functioning as adoption rates grow for technologies such as
residential PV and EVs (IEA, 2011). Electricity output from variable sources such as solar
and wind cannot be timed on the supply-side, so to best utilize generation from these sources
more flexibility is needed on the demand side. This can come in the form of storage,
including both centralized storage and distributed storage such as the batteries of EVs, or in
the form of curtailment and load-shifting to better align timing of demand to timing of
supply.
My third and final dissertation chapter considers predictors of household acceptance of time-
of-use (TOU) rates, a financial demand-side response instrument that charges higher rates to
use electricity in times when supply is predicted to be low, or to require steep ramping.
California in particular is currently exploring TOU as a response to steep ramping needed in
the evening hours as daytime solar generation ceases with sundown (CPUC, 2018). This
current state-driven experimentation with TOU provides a rich dataset to understand drivers
of household TOU acceptance and in turn to gain insights into what may affect household
acceptance of a wider range of demand-side response measures.
1.4 Theoretical contributions and chapter outline
In this dissertation I contribute to theory examining what motivates households to reduce
their GHG emissions, with particular focus on what local governments can do to remove
contextual barriers and on which personal factors affect household attitudes towards adopting
measures to reduce emissions. I build on the fundamental concept discussed by Stern (1999)
that pro-environmental behaviors such as green technology adoption are influenced by both
personal and contextual factors. In particular, I frame my approach within the attitude-
behavior-context theory, which advances that the decision to undertake a pro-environmental
behavior will be influenced by both personal attitudes and by contextual factors. This
provides a framework within which I contribute to bridging literature examining local policy
implementation and its contextual effects with literature examining pro-environmental
behavioral and attitudinal drivers. The emphasis on both personal and contextual factors is
paralleled by Geels’ discussion of socio-technical transitions, where he proposes explicitly
incorporating the user side into analyses of whether newly emerging and niche technologies
will gain prevalence and drive shifts to new systems (Geels, 2004).
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
11

A consideration of contextual factors provides a broad frame within which to consider the
prevalence of barriers to adopting green technologies – in particular, households wishing to
adopt PV face barriers in terms of costs and in terms of regulations imposed by city
governments regarding building modifications (Burkhardt, Wiser, Darghouth, Dong, &
Huneycutt, 2015). This provides the framing for the first chapter, which considers whether
removal of permitting barriers increases rates of PV installation. The second chapter
(focusing on EV adoption) considers both context in the form of costs and charging
limitations, and personal attitudes in the form of perceived symbolism of EVs. This allows
comparison of whether the pro-environmental behavior of EV adoption in the Los Angeles
context is currently driven more by contextual or attitudinal factors. Finally, the third chapter
focused on TOU adoption compares attitudinal factors in the form of perceptions of savings
to a personal-level context in the form of observable spending on electricity to determine
which is a more powerful driver of adoption intentions.
1.4.1 Chapter 1: Drivers of Residential Solar Installations: Do Local Permitting Processes
Have an Effect?
I examine the impacts of local regulations on city-level observed installation rates of
residentially-sized solar PV. This chapter focuses on expected effects of removing contextual
barriers, following implications in the attitude-behavior-context theory that households will
not adopt PV if they face high barriers even if they have positive attitudes towards PV (Paul
C. Stern, 1999). As national and state level leadership take inconsistent action to introduce
policies that mitigate climate change, local policies have been emphasized as an area with
great potential for emissions reduction (Bulkeley, 2010; Fiorino, 2014). PV provides an
excellent case to study the effects of policies aimed specifically at removing barriers, as
several recent studies have highlighted that local government permitting policies may have
strong effects on the price and ease of installing PV (Burkhardt et al., 2015; Dong & Wiser,
2013; Friedman et al., 2013; Seel, Barbose, & Wiser, 2014).
I hypothesize that local government policies to streamline permitting will increase the rate of
PV installations by reducing cost and time needed to install these systems. In Chapter 1, I
examine the effects of permitting policies using both fixed effects analysis (section 2:
Chapter 1) and synthetic control methods (section 7: Chapter 1 Supplementary Appendix).
Both fixed effects and synthetic control analyses are unable to reject the null hypothesis that
streamlining permitting has no effect on PV installation rates, though these analyses do rule
out large impacts of streamlined permitting on installation rates. This neither confirms nor
disproves the theoretical expectation that removing a barrier to PV permitting will increase
installation rates, as the visibility of permitting barriers to households was expected but not
confirmed. Chapter 1 highlights the need for more frequent and consistent data collection on
policies at the local level that would facilitate identification of smaller effects if they exist,
and indicates that drivers of PV adoption at the city level may not correspond with drivers of
PV adoption at zip code and individual levels identified in previous work such as income and
availability of PACE financing (Drury et al., 2012; Kirkpatrick & Bennear, 2014; Kwan,
2012).
Dissertation by Lee V. White
12

1.4.2 Chapter 2: You are what you drive: Environmentalist and social innovator symbolism
drives electric vehicle adoption intentions (publication co-authored with Nicole D.
Sintov)
In Chapter 2, I use survey response data to determine how financial factors, perceived
convenience, and self-identity impact people’s intentions to adopt an EV. EVs have the
potential to dramatically reduce emissions, but even in environmentally-friendly California
adoption rates are still low – only 0.7% of new vehicles sold in California in 2014 were fully
electric (EIA, 2014). Since EVs are still a relatively new technology, there have been limited
studies conducted to determine what motivates households to purchase EVs rather than other
types of vehicles. I draw on existing theoretical frameworks that posit the importance of
symbolic attributes alongside instrumental attributes in predicting intent to adopt sustainable
innovations such as EVs (E. H. Noppers, Keizer, Bolderdijk, & Steg, 2014).
This chapter examines whether EV uptake is driven primarily by removal of contextual
barriers such as high purchase cost or primarily by symbolic attributes affecting attitudes to
the vehicles, and also considers policy implications. I advance the literature by refining the
concept of symbolic attributes to include specific reflections of self-identity, and find that
perceiving EVs to reflect environmentalist self-identity is the strongest and most consistent
predictor of EV adoption intentions. Further, individuals who are concerned about climate
change perceive stronger reflections of environmentalist and innovator identities in EVs.
This chapter emphasizes that policies designed to increase EV adoption should not only focus
on addressing contextual barriers such as cost and charging convenience, but should also
target attitudinal factors and in particular symbolic perceptions by including policy measures
such as campaigns supporting people to view themselves as environmentalists. Encouraging
EV adoption by tapping into self-identity constructs that can drive adoption separate from
cost concerns may increase in importance as EVs become a more mature technology less
feasible to support through financial incentives. This importance of self-identity aligns with
prior work stressing the centrality of altruism and norms in adoption of pro-environmental
behaviors, and in particular the finding that people are more likely to take action to address
an environmental problem if they feel they can make a meaningful and positive impact (Ellen
Van der Werff & Steg, 2015).
Chapter 2 is published in Transportation Research Part A, full citation information: White, L.
V., Sintov, N. D., 2017. You are what you drive: Environmentalist and social innovator
symbolism drives electric vehicle adoption intentions. Transportation Research Part A: Policy
and Practice 99, 94-113.)

1.4.3 Chapter 3: Controlling household electricity loads: The effect of actual versus
perceived savings on Time-of-Use rate acceptance (co-authored with Nicole D.
Sintov)
Finally, in Chapter 3 I examine factors that affect people’s acceptance of TOU rates, which
can facilitate integration of high shares of variable renewable generation capacity into the
grid. I draw together prior literatures that find (1) TOU adoption is financially motivated
(Mostafa Baladi, Herriges, & Sweeney, 1998; Nicolson, Huebner, & Shipworth, 2017; K.
Train & Mehrez, 1994), and (2) people are typically poor at estimating the energy use (Attari,
DeKay, Davidson, & Bruine de Bruin, 2010; Brounen, Kok, & Quigley, 2013), hence energy
costs, of individual appliances. Demand-side response measures such as TOU are seen as
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
13

critical to integrating high shares of variable renewable generation into the grid (IEA, 2011;
Procter, 2013). However, less than 20% of households typically choose to opt-in when TOU
pricing is offered (Todd, Cappers, & Goldman, 2013).
Here I examine whether perceived or actual costs of switching to TOU represent the greatest
barrier to stronger acceptance of TOU and by extension other demand-side response
measures. My finding that perceived savings is the most powerful predictor of TOU
acceptance raises concerns that residents may enroll in TOU on the basis of perceived savings
without actually seeing any savings. This suggests that policies designed to support
household DSR should include support for utilities to provide participating households with
mechanisms for accessible feedback, such as comparing their bills and usage timing to
previous years, to reduce the likelihood of people falsely perceiving savings. If people falsely
perceive savings, they will not be able to make the best possible decisions regarding which
electricity rate they should use to allow ideal allocation of funds between electricity and other
budget items. Further, people with poor understanding of energy-using activities will not
respond to the TOU price incentive in an economically efficient fashion..
1.5 Conclusion: policy lessons learned
Overall, the three chapters emphasize the need to consider non-monetary drivers of
household pro-environmental behavior when developing policy settings, and further
emphasize the need for greater tracking of changes in policy settings over time at the local
level. The second and third chapters reveal the importance of household perceptions, both of
symbolism and of savings, in driving household willingness to adopt new technologies and
new ways of interacting with electricity systems. Factors motivating households to reduce
GHG emissions are confirmed to be highly complex, and to encompass behavioral factors
that are at present not always well considered in policy design compared to more typically
evaluated financial factors. At the individual level, perceptions and symbolism dominate
energy decisions, emphasizing that attitudes toward GHG mitigation technologies are an
important driver of adoption that must be accounted for during policy development. The
findings of this dissertation further emphasize the need to bridge policy and behavioral
literatures on technology adoption.
Based on the findings across the three chapters, in future policy-makers may be able to
facilitate greater emissions reduction by developing policies using information measures to
tap into environmentalist self-identities, using educational strategies to strengthen the link
between perceived and actual energy use, and including settings that facilitate collection of
consistent and comparable data on policy changes across cities. The first two recommended
policy measures, drawn from the EV and TOU chapters, follow prior findings that carefully-
framed information campaigns can increase energy conservation (Asensio & Delmas, 2016)
and that messaging to activate social norms is an effective way to promote a wide variety of
energy-saving behaviors (Ellen Van der Werff & Steg, 2015). The recommendation that
policy changes be better tracked at city level, drawn from the PV chapter, builds on existing
literature that has faced challenges identifying outcomes of local policies due to lack of data
(Feiock et al., 2013; Millard-Ball, 2012). If local policies and their outcomes are not tracked
in a way allowing comparison, analyses cannot feasibly identify which policies have the most
powerful effects on desired outcomes such as climate change mitigation contributions.
Dissertation by Lee V. White
14

The first chapter’s analyses were unable to disprove the null hypothesis, and this may well
have been due to the inability of these analyses to include household level personal indicators
such as whether neighbors had solar PV and whether decisions such as planning for
retirement influenced adoption decisions. Though local government policies streamlining
permitting remove a barrier to PV installation, this barrier may not be visible or widely
known by households prior to beginning installation processes. It may be that local
governments could achieve greater changes in PV installation rates by addressing
informational barriers, or by pursing policies such as whole-city zero-emission strategies
intended to create positive community attitudes towards PV installation. These strategies are
not explored in the present dissertation, but are implied by the attitude-behavior-context
framework as next steps for analysis given the lack of significant findings regarding
contextual barrier removal in the form of streamlining permitting. Recent meta-analysis
examining energy-relevant investment decisions at the household level also emphasizes the
importance of beliefs about consequences for and beyond the household, and the importance
of receiving energy consulting, for encouraging adoption of GHG mitigating technologies
(Kastner & Stern, 2015).
The second chapter reveals that people are motivated to reduce their GHG emissions not only
by financial gains, but also by perceived ability to make a visible statement of being
environmentalists. This suggests that EV adoption in California has progressed to the stage
where attitudes as a driver overpower financial and logistical barriers. These findings
regarding attitude do not suggest that financial incentives should be curtailed, but rather
indicate that for visible durable goods such as EVs the image projected by the purchase is
also an important factor weighed when making buying decisions. This aligns with recent
findings that the more positive people expect to feel following pro-environmental action, the
stronger their intention is to act pro-environmentally (Taufik, Bolderdijk, & Steg, 2016).
Chapter 2 thus also emphasizes existing calls to pursue policy settings that include campaigns
to increase positive attitudes towards pro-environmental actions and technologies (Taufik et
al., 2016).
The third chapter does not consider GHG emission-reducing technology directly, but does
reveal that household-level motivation to change interaction with electricity systems is
primarily financially motivated. However, since there is only a weak link between actual and
perceived savings, despite the financial motivation of household units these findings
emphasize the need include information measures within policy settings to increase
household knowledge of ways to save both kWh and $ when performing electricity-using
activities. The results of this analysis highlight a barrier that may not have been previously
perceived, namely, households are still very poorly positioned to understand which
electricity-using activities incur high costs. Encouraging household level behavior change to
reduce or shift timing of electricity use fundamentally cannot reach full effectiveness and
efficiency while people still lack strong understanding of electricity use, which is invisible,
abstract, and only recently commonly tracked at a more precise level than monthly use.
Policies aimed at increasing flexibility on the demand-side need to consider including
mechanisms to address this information gap, to allow the residential sector to effectively
participate.
A transition to a low-carbon society will require a shift to common use of energy
technologies with low or zero GHG emissions. Literature considering the wider fabric of
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
15

socio-technical shifts historically emphasizes that we need to consider the actual distribution
and use of the technology by individuals (Geels, 2004). This parallels calls within the
behavioral literature to consider both personal and contextual factors that drive decisions to
adopt new technologies (Paul C. Stern, 1999). Combined, the three chapters of this
dissertation provide insights into what is driving adoption of PV and EV technologies that
have only recently matured, and emphasize that policies to encourage future uptake of these
will need to consider factors beyond the economic. In particular, transition to a low-carbon
society can be accelerated by targeting policies to increase energy literacy and tap into
environmentalist self-identities, following prior literature examining drivers of pro-
environmental behavior (Schley & DeKay, 2015; Ellen Van der Werff & Steg, 2015).
Overall, factors motivating households to reduce GHG emissions are complex, and there is a
need for policies promoting adoption of pro-environmental technologies to encompass
behavioral factors alongside economic ones.

Dissertation by Lee V. White
16

2 Chapter 1: Drivers of Residential Solar Installations: Do Local
Permitting Processes Have an Effect?

By Lee V. White
Commonly used abbreviations: CEC: California Energy Commission; CPUC: California
Public Utilities Commission; ICLEI: “Local Governments for Sustainability” network; IREC:
Interstate Renewable Energy Council; IOU: Investor Owned Utility; POU: Publicly Owned
Utility

2.1 Abstract
Local governments are relatively agile policy makers and are expected to be able to make
significant contributions to climate change mitigation through local legislation. In addition to
policies to reduce energy use (e.g., promoting walkable development), local governments can
introduce policies that make it easier for people within households to generate emission-free
electricity from solar photovoltaic (PV) panels. Streamlining permitting processes for solar
PV can significantly reduce overall costs of installing these systems and potentially reduce
time and cost burdens. This study uses fixed effects modelling to examine impacts of local
policies introduced to promote solar PV by cities in California. To fill a gap in longitudinal
data on implementation dates of local policies to support PV, I utilize a combination of
surveys, partial databases, and publicly available city-level information to build a complete
picture of changes in relevant policies from 2005 to 2013. Modelling results are unable to
reject the null hypothesis that the implementation of streamlined permitting has no effect on
residential PV installation rates. This indicates that further exploration is needed to determine
what policies, if any, cities can implement at the local level to significantly increase
residential solar installations, and highlights the need for consistent tracking of city policy
changes.
2.2 Introduction
Urban areas account for up to 76% of CO 2 emissions globally (Seto et al., 2014). This,
combined with the comparatively fast pace at which city governments are able to enact
policies to improve sustainability (Feiock et al., 2013), positions cities as a potentially critical
arena for implementing measures to reduce greenhouse gas (GHG) emissions. City
governments can take steps to reduce emissions from transportation, to reduce waste, and to
improve efficiency of household operations, and may also be able to promote installation of
renewable generation technologies such as solar photovoltaic (PV) panels that can displace
emissions associated with electricity use (Seto et al., 2014). Encouraging distributed
renewable generation, particularly solar photovoltaic (PV) panels, is one of the ways that
cities can reduce greenhouse gas emissions associated with their jurisdiction. Determining
whether city government policies have a measurable impact on residential solar PV
installations could reveal whether this is an area where city governments should direct greater
attention in future.
Over the past several decades, city governments have become increasingly involved in efforts
to reduce GHG emissions and mitigate climate change, and energy use and energy efficiency
are often key areas targeted by city-level sustainability policies (Bulkeley, 2010). Cities have
continued to gain prominence as important actors to mitigate climate change since
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
17

international and national-level agreements and policies have often stalled during
negotiations (Betsill & Rabe, 2009; Bulkeley & Betsill, 2013). However, measuring the
impacts of these city-level efforts has been challenging due to lack of consistent data, and
quantitative studies have not been able to confirm that city actions such as implementing
Climate Action Plans are causal drivers of emissions reduction (Millard-Ball, 2012).
The primary mechanism by which local governments are expected to be able to increase PV
installations is by removing relevant barriers, whether these are financial, regulatory, or time
barriers. If households face barriers to performing pro-environmental behavior such as PV
installation, for example time or cost barriers, then this will constrain their ability to shift to
these pro-environmental behaviors (Paul C. Stern, 1999). Permitting application procedures,
review times, and fees have been identified as significant barriers to installing residential
solar (Pitt, 2008), and several recent papers indicate that local governments could reduce
residential solar PV costs by streamlining permitting processes and may also be able to
reduce the time and inconvenience of installing a PV system (Burkhardt et al., 2015; Dong &
Wiser, 2013; Seel et al., 2014).
This paper examines whether cities which have lessened barriers posed by permitting
procedures observe higher rates of PV installation following removal of these barriers while
accounting for differences inherent to each city and between utilities. Previous studies
examining residential PV permitting have examined data from only one or two years of city
permitting practices. The current paper contributes to this literature by using fixed effects
analysis to examine, for a sample of California cities through the years 2005 to 2013, the
impacts on existing trends in PV installation when cities increase the availability of online
information for permitting and achieve average turnaround times for residential solar permit
issuance of three days or less. No data set previously existed recording changes in cities’
solar permitting practices over several years, and the current paper further contributes to the
literature by developing this data set for cities in California.
2.3 Predictors of residential PV adoption
Several recent papers have examined individual-level and zip-code level predictors of
residential PV adoption (Davidson, Drury, Lopez, Elmore, & Margolis, 2014; Kwan, 2012;
Robinson & Rai, 2015). However, none of these residential PV studies have included
predictors for the effect of variation in permitting practices between cities on PV adoption
rates. One prior study examining streamlined permitting impacts in the context of comparing
city vs. state policies found no impact of permitting policies (Li & Yi, 2014), but did not take
into consideration factors such as how long streamlined permitting policies had been in place,
and did not restrict the sample to residentially sized PV systems. Further, the cross-sectional
research design used is inherently unable to take into account whether implementing a
streamlined permitting policy increased the rate of solar PV installations relative to existing
trends in each city.
PV adoption at the individual level is predicted by financial aspects (Robinson & Rai, 2015).
Zip-code level analysis to predict PV adoption in California found that education, race, and
home value were good predictors as expected, and that number of rooms in the home, heating
source, and house age were additionally key predictors (Davidson et al., 2014). The cost of
electricity and the presence of financial incentives have also been identified as important
predictors of zip-code level PV installation rates (Kwan, 2012). Higher educational
Dissertation by Lee V. White
18

attainment is associated with more PV installations, as is lower housing density and higher
income (Kwan, 2012). Older residents (age group 45-54 vs. 25-34 and 35-44) are associated
with greater PV installation (Kwan, 2012), and additionally many PV owners report that they
started considering PV when contemplating retirement investments (Rai, Reeves, &
Margolis, 2016). Zip codes with median incomes of $45,000-$100,000 have been found to be
more strongly associated with PV installation than those with median incomes of $25,000-
$45,000 or >$100,000 (Kwan, 2012). Availability of Property-Assessed Clean Energy
(PACE) financing has also been found to increase installation rates of residential solar PV
(Kirkpatrick & Bennear, 2014).
Permitting can add unnecessary complexity for both households and installers, and soft costs
for PV systems including permitting and paperwork can make up 30-40% of the total cost for
a PV system (Treadwell, Gishri, Wheeland, Anders, & Kaatz, 2012). Variations in permitting
procedures between jurisdictions with the best and worst permitting practices are associated
with average installation cost differences of $0.18/W installed for residential PV systems
across the US, and streamlining permitting processes could reduce PV costs in California
specifically by $0.27-$0.77/W (Burkhardt et al., 2015; Dong & Wiser, 2013). The time that it
takes to issue permits can vary from a few hours to over a month, and longer plan check
times can introduce both additional costs and additional frustrations for bringing PV systems
online (Dong & Wiser, 2013; IREC, 2013). The ease of accessing information about solar
permitting procedures also varies between cities, with some making information readily
available online whereas others give little or no indication of what is required to receive a
permit for residential solar (IREC, 2013). High fees, long permitting times, and lack of
widely accessible permitting information may all contribute small but potentially significant
barriers to the already expensive and lengthy process of installing residential solar PV (IREC,
2013).
2.4 Method
The current research focuses specifically on solar installations in cities within the state of
California, a state that exhibits strong leadership in promoting renewable generation. Multi-
year panel data on city policy status was not readily available, and was compiled from
multiple sources as described in section 2.4.2; the scarcity of city-level data is an ongoing
problem with analysis of city sustainability policies (Feiock et al., 2013; Millard-Ball, 2012).
The initial set of cities for which PV installation rates and policy data were sought contained
114 localities with populations over 65,000 for years 2005-2013, but only 98 of these had PV
installation records for all years. Since ACS 1 year data was used for consistency, only cities
with populations consistently over 65,000 had demographic data available for all years. Of
these 98, only 57 had definitive accessible permitting records, and only 23 had definitive
information regarding permitting fees specifically for residential solar PV for all years of
interest. Thus, fully balanced panel data including streamlined permitting indicators were
available for 57 cities, and permitting fees and PACE indicators were examined using a fully-
balanced sub-sample of 23 cities. Table 2.1 provides descriptive statistics for the main sample
of 57 cities.
The following section describes data collection. Data describing variation in city permitting
practices over time is sparse, particularly for years before 2011, so all indicators used for
streamlined permitting and permitting fees were collected individually for each city where
available, after seeding with cross-sectional data available for select years. This is further
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
19

described in sections 2.4.3 and 2.4.4. To determine whether cities with complete data
available regarding permitting speed and process had different underlying PV installation
rates compared to cities that had no such data available, a fixed effects regression was run
estimating the impact of a missing permit data dummy and year dummies on installed PV
capacity. Fixed effects analysis is discussed in detail in Section 2.4.1. No significant impact
was found for missing vs. present permitting data on installation rates.
Table 2.1: Descriptive statistics for main sample (57 cities)
Variables Units Mean SD Min Max
Watts of PV installed
annually

Watts/capita (i.e., W/capita or
kW/1000 people)
4.20 7.04 0.017 64.39
Number of PV systems
installed annually

PV systems/1000 people 0.90 1.41 0.01 11.72
Streamlined permitting 1 if permitting was streamlined,
0 otherwise
0.11 0.31 0 1
Permitting fees <$400
1
1 if permitting fee under $400,
0 otherwise
0.28 0.48 0 1
PACE
1
1 if PACE program in place,
0 otherwise
0.24 0.42 0 1
Electricity price
2
$/kWh 0.15 0.02 0.08 0.18
Financial incentives
2
$/kW installed 1.84 1.06 0.00 4.50
Median age
2
Years 34.10 3.40 25.70 43.00
Share of registered
democrats
Percent 43.21 8.12 28.00 58.00
Share of households
with income > $100,000
Percent 26.76 10.19 5.30 54.10
Share of households
with income
$100,000 to $50,000
Percent 31.27 4.08 21.50 45.30
Share with bachelor’s
degree or higher
Percent 30.47 13.90 7.30 69.9
1
Values for sub-sample (26 cities),
2
prior to non-linear transformations

2.4.1 Empirical approach
Cities that chose to streamline their residential solar permitting procedures may be inherently
different from those that did not, and these differences may drive higher solar installation
rates even prior to streamlining permitting. This poses a problem of selection bias, which can
be partially addressed by using fixed effects analysis to control for time- and city-invariant
differences (see Equation 2.1). Fixed effects modelling assumes that there are no time- and
city-specific unobserved factors correlated with the treatment, i.e., no “contemporaneous
shocks”. This is a more plausible assumption to meet than the assumption necessary for
cross-sectional analysis that there is no unobserved heterogeneity between cities. The model
is represented by the Equation 2.1, which corresponds to the predictors outlined in Figure 2.1:
Equation 2.1:

=

+

+

+

+

+

Where PVit represents the number of systems installed in a given year; αi represents a city-
specific intercept; γ t represents year fixed effects; P it represents permitting practices of
different cities; Cit represents other city-level differences expected to impact PV
development, including income and median age; and U it represents utility characteristics
Dissertation by Lee V. White
20

including average electricity price in each year and level of financial rebate for solar offered
each year. The coefficient β gives the average effect of permitting practices. Errors for the
fixed effects regression were clustered at the city level, to avoid overstating of effects due to
serial correlation in the time-series data.
Figure 2.1: Factors expected to drive city-level differences in residential PV installation
rates. Signs indicate expected direction of effect on PV installation rates

2.4.2 Streamlined Permitting
There is no existing database that tracks city-level changes in solar PV permitting practices
over many years, for cities in California or any other state, so an indicator was constructed
from existing cross-sectional data sets and internet archives
1
. IREC outlines several areas that
are considered “best practices” for permitting and as such are expected to predict PV
adoption. These are: (1) post requirements online, (2) implement an expedited permit process,
(3) enable online permit processing, (4) ensure a fast turnaround time, (5) collect reasonable
permitting fees, (6) do not require community-specific licenses, (7) offer a narrow inspection
appointment window, (8) eliminate excessive inspections, and (9) train permitting staff in
solar (IREC, 2013).
Of these, (1) and (4) are used to construct the streamlined permitting indicator, and (5) is
considered separately. Inspection information records for cities are even scarcer than
permitting practice information, so no measures could be constructed to examine the impacts
of (7) or (8), and training and community-specific licenses are too variable to construct
indicators for. (3) was excluded since in the earlier years of the study, online permit
processing would have been extremely rare for any type of submission. This yields a
streamlined permitting indicator that requires cities to meet two criteria that are consistent

1
https://archive.org/ was used to find archived web pages for each city’s building department. Any html
webpages are time-stamped with the date when the snapshot was taken, and so could be used to determine
whether the city had a link available for a residential PV checklist in any given year. Archived PDFs needed to
be viewed more cautiously, as they did not contain timestamps – to some extent, this precluded determining
whether cities met more strenuous criteria (such as only requiring approval from one building department)
even if it could be confirmed that they had a PV checklist online in a given year.
Cit
Income (+)
Political party (county level) (+)
Median age (+)
Education (+)
PACE (+)
PVit
Number of PV systems
installed/year
and
kW of PV installed/year
P
it

Streamlined permitting (+)
Low permitting fees (+)
U it (utility characteristics)

Electricity prices (+)
Utility financial incentives (+)
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
21

across the guidelines from IREC and California’s recent AB 2188
2
, (1) that cities should
make checklists for photovoltaic permitting requirements available online and (2) that cities
should aim for turnaround times of 3 days or less.
Turnaround time and posting of requirements online were taken as the two criteria to
construct a streamlined permitting indicator for several reasons. Firstly, the “Requirements
Posted Online” criterion provided a largely consistent way to determine the earliest year
during which city permitting qualified as “streamlined”, as without this criterion there would
be no guarantee that any written record of city permitting practices ever existed. Since web
archives are not necessarily comprehensive, a city was only considered to have “failed” the
“Requirements posted online” criterion if an archived page of the city’s building department
was found in any given year that included no links, and made no reference, to any form of
solar PV checklist or application procedure. If no such page could be found, and if no reports
existed to clarify when PV permitting had first been streamlined, then the city was excluded
from the sample.
Secondly, including turnaround time provided an indication of whether cities were generally
minimizing delays and associated costs for solar installers and thus homeowners. Permits are
also often needed before other stages of the solar installation process can be completed, such
as rebate applications (Treadwell et al., 2012), so reducing delays in this area is likely to have
associated impacts on the overall installation timeframe. “Rapid Turnaround Time” was taken
as 3 days or less, reflecting the threshold stated both in AB 2188 and IREC recommendations
(IREC, 2013; J. Kaatz & Anders, 2015). A more lenient specification whereby “fast
turnaround” was 10 days or less was also tested in fixed effects models, and no differences
were found regarding which predictors were significant.
To create a data set recording whether or not cities met the criteria for streamlined permitting
during any given year between 2005 and 2013, VoteSolar’s Project Permit was used as a base
to determine whether a city met the criteria for streamlined permitting in 2013. Project
Permit, a crowed-sourced map of city permitting practices across the US which is
continuously updated, has been used in previous research examining local permitting
practices (Burkhardt et al., 2015). Cities that passed both “Requirements Posted Online” and
“Rapid Turnaround Time” requirements for Project Permit then formed a shortlist, and every
city on that shortlist was examined individually through web archives and relevant reports.
This newly constructed indicator covers more years than any previous measure for
differences in city permitting practices. Table 2.2 shows the dates for cities included as
treated units; control units were those cities confirmed to have not met both permitting
criteria prior to 2013.
2.4.3 Permitting Fees
An initial base of permitting fee data was built from two existing datasets, then supplemented
with information sourced from web archives and news reports. The first of these two sets was
a substantial database of California city solar permitting fees that exists thanks a project led
by Kurt Newick in which the Sierra Club gathered permitting fee information for many
California cities between 2005 and 2012 (Kurt Newick pers. comm, 2016; Mills & Newick,

2
AB 2188 requires all cities in California to adopt a checklist of PV system requirements for expedited review,
allow electric submission of permitting documents, and use only a single inspection on installed systems
before allowing grid connection to proceed (CSE, 2016)
Dissertation by Lee V. White
22

2008; Mills, Newick, Stewart, & Compeán, 2009). The second data set used was Project
Permit, which provided a snapshot of whether cities “passed” or “failed” the test for
reasonable permitting fees in 2013. Project Permit used a cut-off of $400 to either pass or fail
cities on the criterion of “fair permitting costs”. If a city charged fees under $400 when
surveyed by the Sierra Club, then still charged fees under $400 when surveyed by Project
Permit, it was assumed that the city also charged reasonable permitting fees during the
intervening time unless there was evidence that the city had instituted a potentially temporary
fee waiver which may have expired in the interim (i.e., any case where city fees were
observed to be $0 during some years).
Coding for permitting fees was restricted to whether cities had fees under or over $400 in any
given year, with cities charging fees under $400 being coded as “1” and cities with fees over
$400 being coded as “0”. The cut-off of $400 is based on recommendations by IREC (IREC,
2013). It was assumed that if a city had reasonable permitting fees (under $400) when
Newick first led a survey of them in 2005 and 2006, then they would have also had
reasonable permitting fees prior to this first survey; this is considered a reasonable
assumption because there was little publicity of permitting fees prior to Newick’s work (Mills
& Newick, 2008).
Newick notes that cities typically either charge solar fees based on the time it takes staff to
process permits (leading to reasonable permitting fees), or based on a percentage of the value
of the installed solar system (leading to much higher permitting fees). It was thus assumed
that, if a city was observed to drop its fees by several hundred dollars to fall below the $400
threshold, the city was highly likely to have switched from a “valuation” method to a
“processing time” method of calculating permitting fees. Thus, cities with a substantial
observed drop in fees were assumed to have charged higher fees consistently before this drop
and were coded accordingly.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
23

Table 2.2: Permitting timing, availability, and fee information for the treated cities included
in the final sample. Cities with incomplete fee information were excluded from the sub-
sample analysis, as were cities that waived permitting fees for some years then reintroduced
fees.
City Permitting
three days
or less
Permitting
information
online
Met both
permitting
criteria
Major permitting fee changes where
known
Antioch 2013 2013 2013 From $745 in 2007 to $243 in 2008
Berkeley 2011 2009 2011 Waived fees 2008 only
Fremont 2009 2009 2009 From $850 in 2006 to $237 in 2007
Lancaster 2010 2010 2010 $61 flat fee introduced 2010
Long Beach 2009 2009 2009 From $599 in 2009 to $233 in 2010
Sacramento 2009 2009 2009 Waived fees 2007-2009 only
San Diego 2012 2006 2012 Unknown
San Francisco 2007 2008 2008 $85-$170 all years
Santa Barbara 2006
1
2007 2007 No changes known ($430)
Santa Clara 2009 2009 2009 Unknown
Vista 2009
1
2011 2011 Unknown
1
or earlier
2.4.4 Potential model issues
2.4.4.1 Omitted variable bias
Use of fixed effects controls for time-invariant unobservables, but there may exist time-
varying unobservables that are correlated with both PV installation rates and the decision to
adopt streamlined permitting. This would create an omitted variable violation, causing
estimates to be biased. The main concern in this model is unobserved variables which may
impact both local government decisions to streamline permitting and simultaneously affect
resident decisions to install solar PV. One such unobservable may be solar company
marketing activities, which could vary across cities and time such that cities with unobserved
higher solar marketing may have correspondingly higher PV installation rates. This variation
could not be included in the model, despite previous findings that marketing can act as a
“spark event” for encouraging solar installation (Rai et al., 2016).
The link between solar marketing activities and city decisions to streamline permitting is less
established, but streamlined permitting procedures are expected to reduce costs for solar
installers (Miller, 2014). Case studies of Sacramento, San Francisco, and Lancaster PV
permitting reform all discuss influences of the solar industry (ICMA, 2015; NREL, 2011; SF
Environment, 2012). In all three cases, discussions or roundtables with local stakeholders
influenced plans to streamline solar permitting procedures to reduce time and complexity for
the permit process. Thus, a decision by a solar installer to increase activity in any given city
may be correlated with both PV installations and with city government interest in
streamlining permitting. Both of these correlations are expected to be in the positive
direction, which would inflate coefficient estimates, and positive results would thus need to
be viewed cautiously.
2.4.4.2 Reverse Causation
There is an additional risk of reverse causation, particularly given that case studies of San
Diego and Santa Clara report streamlining permitting procedures after facing difficulties
processing increasing numbers of PV permit applications under their prior systems (City of
Dissertation by Lee V. White
24

San Diego and the California Center for Sustainable Energy, 2009; SunShot and North
Carolina Solar Center, 2013). Solar permitting in San Diego used to be over the counter, but
when the city began receiving 40 permit requests each month it no longer had sufficient staff
to process them over the counter and permit times slowed, which likely contributed to fully
streamlined permitting being rolled out in 2012 (City of San Diego and the California Center
for Sustainable Energy, 2009). It is conceivable that other cities may also have chosen to
implement streamlined permitting if there was, or was expected to be, a large increase in PV
installations, to prevent their planning staff from being overwhelmed; thus, it may be that
increasing PV installations prompt streamlined permitting rather than vice versa.
If permitting is streamlined in expectation of higher PV installation rates in future, a positive
coefficient would still be expected for the streamlined permitting variable – that is, it would
still be expected that the rate of PV installation would increase following implementation of
streamlined permitting. This requires caution interpreting positive result coefficients.
Likewise, if an increase in PV installations prompts implementation of streamlined
permitting, then the years following the streamlining of permitting would be expected to have
an even higher rate of PV installations if the rate were in turn influenced by permitting
processes. However, if streamlining permitting does not have an effect on PV installation
rates, then the rate would instead be expected to continue at the high level which prompted
permitting to be initially streamlined. In the latter case, the null hypothesis that streamlining
permitting has no effect on PV installation rates cannot be ruled out.
2.4.5 Dependent variable: Rate of PV system installation
Two dependent variables were examined: (1) watts (W) per capita of PV installed each year,
and (2) the number of PV systems per 1,000 people installed each year. Streamlined
permitting is expected to encourage a greater number of PV systems to be installed, but is not
necessarily expected to cause each individual system to be larger, hence both the number of
PV systems and total installed capacity were examined to ensure that the effects of the policy
on household preferences for installation were captured. These were both considered on a per
capita basis to account for differences in the size of the cities being examined, with
population data for cities in California taken from the US census bureau.
PV systems between 1-10 kW were included in sums of city PV capacity. Though net
metering incentives are extended to systems up to 1 MW in size, guidelines for expedited
solar permitting are generally created only for smaller systems under 10 or 15 kW (CPUC,
2016; J. Kaatz & Anders, 2015). Previous studies examining residential solar have also
limited analysis to systems 10 kW and under, in absence of records indicating whether a
given installation is residential or commercial (Kwan, 2012). As residential systems are an
average size of around 5 kW, a cutoff of 10 kW is considered sufficiently representative.
Berkeley Lab’s “Tracking the Sun” dataset was used to source zip code level information on
installation rates of solar PV systems (Lawrence Berkeley National Lab, 2016). Data for
California in 2014 and 2015 is not provided by this data set, due to concerns about accuracy
as record keeping transitioned between California agencies with the expiry of SB 1. A
supplementary analysis was additionally performed using quarterly California Solar Initiative
(CSI) data between 2007-2013. Although CSI data is included in the Tracking the Sun data
set, running a supplementary analysis limiting data to that gathered by one program limits the
potential for apparent results to be due to discrepancies in data collection between different
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
25

agencies or different incentive programs. No differences were found regarding which
variables had statistical significance between using CSI and Tracking the Sun data sets, nor
for quarterly vs. annual analysis of Tracking the Sun data for 2007-2013.
2.4.6 City characteristics
2.4.6.1 PACE
PACE financing was coded 1 for each year or quarter once a city had passed a resolution to
adopt its first PACE program. Specific programs included in searches included the HERO
program, CaliforniaFIRST, AllianceNRG, and FigTree, and PACE programs such as
BerkeleyFIRST that are city or region specific were also included. PACE measurement is
complicated by the fact that in July 2010, the FHFA issued supervisory guidance that PACE
could introduce soundness concerns for mortgages, and this had the effect of stalling or
temporarily suspending some PACE programs. However, many jurisdictions that had
implemented PACE in early years restarted their PACE programs shortly after this
disruption, and many other cities continued to join PACE programs in later years despite the
lack of resolution of the FHFA guidelines (Joe Kaatz & Anders, 2014). Thus, the most
consistent treatment of PACE was determined to be assuming that programs continued
uninterrupted from their start month. A sub-sample was run excluding all cities that began
their PACE programs before 2011 (reducing the sample from 23 to 17 cities), but the results
of this analysis yielded coefficient sizes and significances comparable to the main analysis.
2.4.6.2 Income
Greater income is expected to be associated with higher installation rates of residential PV.
Estimates of the share of households with income over $100,000 and between $50,000-
$99,000 were taken from the US Census American Community Survey (ACS), using ACS 1-
year estimates both for consistency throughout all time periods and to make use of most
current data in each year.
2.4.6.3 Education
Cities with more educated populations are expected to have higher installation rates of
residential PV. Estimates of the share of the population in each city over 25 with a Bachelor’s
degree or higher were taken from the ACS 1-year estimates.
2.4.6.4 Age
Cities with residents of a greater median age are expected to have higher rates of residential
solar installation. ACS 1-year estimates were used to determine the median age in each city
for each year. A square transformation was applied to this variable as increasingly higher
median age was expected to have an increasing positive impact on increases in PV
installation rates.
2.4.6.5 Voting preferences
Cities with a higher share of democrats are expected to have higher installation rates of
residential solar. County-level registered voter preferences were assigned to each city,
recording the percent registered as democrats as recorded in the Secretary of State voter
registration statistics (California Secretary of State, 2015).
2.4.7 Utility Characteristics
ArcGIS was used to match the 2015 California Energy Commission (CEC) maps of
California utility service areas with the 2015 boundaries of cities from the U.S Census
Bureau (CEC, 2015). Utilities were spatially joined wherever they intersected with cities, to
Dissertation by Lee V. White
26

identify which utilities served each city and also if multiple utilities served some cities. Cases
where overlap had occurred were then examined individually, and if over 90% of the city’s
area overlapped with a single utility then that utility was assigned to the entire city. Two
cities did not meet this criteria (Modesto and Mission Viejo, where the majority utility
covered less than 80% of city’s area), and were dropped from the sample.
2.4.7.1 Financial incentives
The California Solar Initiative (CSI) was established in 2006 by SB 1, and all cities within
Investor Owned Utility (IOU) territories have been eligible for rebates through the CSI since
2007, and prior to that were eligible for rebates through the Emerging Renewables program.
Cities served by Publicly Owned Utilities (POUs) were required by SB1 to have access to
rebates from January 2008 onwards, and POUs serving cities in the sample had all
implemented rebate programs prior to 2008 except for Alameda Power & Telecom. Due to
California’s stringent support of residential solar PV, nearly all cities had access to rebates
through state-mandated programs for the entirety of the time of analysis.
Although nearly all utilities offered financial incentives for residential solar PV over the
years studied, the California legislation both allowed for variations between utilities and
required that the financial incentives offered decline over time. Following the introduction of
SB1, all utilities used incentive schedules that declined through set tiers as thresholds for
participation were met. That is, the rebate rate would decline after a given quantity of kW had
been installed in the utility’s territory.
These rate schedules, and the rate for each utility in any given year, were gathered from
archived utility and DSIRE webpages. With the exception of LADWP, all utilities recorded
rates in terms of $/kW installed solar; thus the rates in $/kW were used for modelling input
(no LADWP cities were included in the final sample). A logarithmic transformation was
applied to this variable as increasingly higher incentives are expected to have a decreasing
impact on increases in PV installation rates.
2.4.7.2 Electricity prices
EIA form 861 was used to determine average bundled residential prices for each utility. Each
city was assigned the average price in that year for the utility it is served by. A logarithmic
transformation was applied to this variable, as increasingly higher prices are expected to have
a decreasing impact on increases in PV installation rates.
2.4.7.3 Third Party Ownership
Analysis controlled for availability of TPO by excluding utilities from analysis if they had
ever prohibited leasing. IOUs allowed connections by TPO systems from 2008 onwards, but
some POUs did not allow TPO for several more years. Access to TPO was coded 0 for all
cities prior to the year 2008, and 1 for all cities for the year 2008 and onwards. This was then
refined by stated POU policies on utility websites, as found from web archives and press
releases announcing changes to leasing availability. When coding was completed, it was
notable that only one municipal utility (Roseville) disallowed TPO for multiple years after it
was widely available to IOU customers, and Roseville was removed from the sample.
2.5 Analysis
Following construction of the dataset identifying which cities had implemented streamlined
permitting, it was possible to visually compare trends of cities that did vs. did not implement
streamlined permitting. This simple visual comparison is presented in Figure 2.2. There is
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
27

clearly a large amount of variation that cannot be attributed simply to cities implementing or
not implementing streamlined permitting procedures, but this variation appears to be much
greater after 2009. Many cities implemented streamlined permitting near the year 2009 (see
Table 2.2), so relatively similar trends revealed by visual inspection of Figure 2.2 prior to this
year indicate that fixed effects analysis may offer further insight as to the impact of
streamlined permitting on PV installation rates. It is notable also that Figure 2.2 shows the
cities which implemented streamlined permitting during the study period actually had lower
per-capita installation rates by 2013 than some of the cities that didn’t. Four cities (including
Berkeley) were excluded from the fixed effects analysis (and from Figure 2.2) as visual
inspection indicated that they had unusual trends not in line with those of other cities pre-
2009.
Figure 2.2: PV installation trends in cities that did vs. did not implement streamlined
permitting

Fixed effects analysis is used to further investigate the impact of streamlining permitting,
while controlling for differences between cities. The results of this analysis cannot rule out
the null hypothesis that streamlining permitting has no effect on solar installation rates.
Figures 2.3 and 2.4 plot the coefficients from this fixed effects analysis with error bars at the
95% confidence level, and it is notable that in both cases the estimates are not precise enough
to determine whether the effect is positive or negative. Of the 53 cities retained for analysis,
only 11 introduced streamlined permitting during the period of the study; this limits the
explanatory power of the model. The coefficient for streamlined permitting remains
statistically insignificant for both dependent variables (W/capita installed annually and
number of PV systems/capita installed annually, for both main sample and the sub-sample
(including PACE and permitting fees). Additional tests were run to examine whether the
model was sensitive to specifying the dependent variable using number of owner-occupied
houses or number of single-unit owner occupied houses in each city rather than city
population, and two additional models were run examining quarterly (rather than annual)
installations from 2007-2013 using both CSI and Tracking the Sun datasets. The model was
also examined excluding POUs from the analysis, since there is a possibility that data
Dissertation by Lee V. White
28

collection methods for these utilities differed from those of IOUs, which are expected to be
more consistent because IOUs are all mandated to collect and send data to the CPUC
following set guidelines. These model variants all failed to reject the null hypothesis that
streamlining permitting has no impact on residential solar installation rates.
Of the control variables, only two were statistically significant. Median age had an impact at
or close to zero on both W/capita and systems/capita installed annually for both samples,
while having a higher share of households with income between $100,000 - $50,000
predicted lower rates of PV installations for only the sub-sample analysis including
permitting fee and PACE impacts. Neighborhoods with higher shares of incomes above
$100,000 or higher shares of incomes $50,000-$99,000 were not associated with higher rates
of PV installation, despite expectations that this would be an important predictor based on
findings at the individual level that financial factors are an important predictor of PV
installations.
Figure 2.3: Main sample (N=53 cities): Coefficients from fixed effects model of factor
impacts on PV installation rates, using cluster robust standard errors clustered at city level.
All models included fixed effects dummies for year and city, with 95% confidence intervals.
Table provided in Section 6.1: Chapter 1, Appendix 1.

Streamlined permitting
Electricity price (log)
Financial incentives (log)
Share democratic
Age (sq)
Share with income > $100,000
Share with income $100,000 to $50,000
Education
-3 -2 -1 0 1
systems/1000 capita
-15 -10 -5 0 5
W/capita
kW systems
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
29

Figure 2.4: Sub-sample (N=23 cities): Coefficients from fixed effects model of factor impacts
on PV installation rates, using cluster robust standard errors clustered at city level. All
models included fixed effects dummies for year and city, with 95% confidence intervals.
Table provided in Section 6.1: Chapter 1, Appendix 1.

2.6 Conclusions and Policy Implications
As national and state level leadership take inconsistent action to introduce policies that
mitigate climate change, local policies have been emphasized as an area with great potential
for emissions reduction (Bulkeley, 2010; Fiorino, 2014). It was hypothesized that local
government actions to remove barriers to residential solar PV by streamlining permitting
procedures would lead to increased PV installation rates, by reducing cost and time needed to
install these systems. However, analyses in the current paper were unable to reject the null
hypothesis that streamlining permitting has no effect on PV installation rates. Modeling did
rule out the possibility of very large effects of city level policies on PV installation rates; if
city streamlining of permitting policies was a powerful driver of residential PV installation
rates, then an effect was expected to be seen despite the small sample size gathered, and
potential modeling issues identified were expected to inflate any observed effect rather than
mask it.
This paper both highlights the need for more frequent and more consistent data collection on
policies at the local level that would facilitate identification of smaller effects if they exist,
and indicates that drivers of PV adoption at the city level may not correspond with drivers of
PV adoption at zip code and individual levels identified in previous work such as income and
availability of PACE financing (Drury et al., 2012; Kirkpatrick & Bennear, 2014; Kwan,
2012). The lack of centralized databases of city policies and climate change metrics has made
it challenging to conduct quantitative assessment of the impact of city-level strategies to
mitigate climate change (Bulkeley, 2010; Feiock et al., 2013), and to the author’s knowledge
this study is the first to examine whether changes in city policies have quantifiable effects on
Streamlined permitting
Permitting fees <$400
PACE
Electricity price (log)
Financial incentives (log)
Share democratic
Age (sq)
Share with income > $100,000
Share with income $100,000 to $50,000
Education
-4 -2 0 2 4
systems/1000 capita
-20 -10 0 10 20
W/capita
kW systems
Dissertation by Lee V. White
30

residential PV installation rates using several years of policy data. While analysis was unable
to reject the null hypothesis that cities streamlining permitting has no impact on residential
PV installation rates, this work nonetheless advances the literature by providing the first
attempt to address this question using quantitative time-series data.
Given the lack of support for rejecting the null hypothesis, it should also be considered that
streamlining permitting may in fact not have any impact on residential PV rates. For example,
it may be that consistency between jurisdictions is more important than any single
jurisdiction streamlining its permitting process. Future work may be able to examine the
impacts of consistency of permitting as opposed to streamlining of permitting, particularly
given the growing prevalence of standards such as Solar ABCs’ Standardized Permit Process
Reports and legislation such as California’s AB 2188
3
. PV installation decisions may also be
influenced by neighbor effects and the impacts of marketing activities by solar PV installers,
which were not possible to include in the current city-level analysis but likely do have some
impact on PV installation decisions (Rai et al., 2016).
2.7 Limitations and future research
The study was limited by a relatively small sample of only 53 cities over 9 years, only 11 of
which implemented streamlined permitting during the course of the study. This may have
limited the ability of fixed effects analysis to identify effects. Further, it is possible that the
model contained both omitted variable bias and potential reverse causation, though both of
these modelling issues were expected to inflate coefficients rather than suppress them
(discussed in section 2.4.4). Greater access to data allowing measurement of the variation
between solar company marketing effects in different cities may result in a model with
greater overall explanatory power for predicting PV installation rates.
Several permitting-related factors could not be included in the present analysis. In particular,
no measures were included regarding the length of time and cost of inspections of PV
systems following initial issuance of the permit, though these inspections are also expected to
present a barrier to PV installation (IREC, 2013). Further, the permitting was not assessed
based on the number of departments involved in review of the systems, which may also
introduce additional barriers. Additionally, lack of detailed time-series data precluded TPO
from inclusion in any analyses, despite the expectation that availability of different ownership
options may be an important predictor of residential PV uptake. PV systems typically have a
high upfront cost, and coupled with the long payback times for these systems it is likely that
upfront costs have placed limits on installations (Corfee et al., 2014). Future work may be
able to incorporate the availability of different ownership options to a greater degree.

3
AB 2188 requires all cities in California to adopt a checklist of PV system requirements for expedited review,
allow electric submission of permitting documents, and use only a single inspection on installed systems
before allowing grid connection to proceed (CSE, 2016)
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
31

3 Chapter 2: You are what you drive: Environmentalist and social
innovator symbolism drives electric vehicle adoption intentions
(publication co-authored with Nicole D. Sintov)

By Lee V. White and Nicole D. Sintov
3.1 Abstract
Electric vehicles (EVs) have the potential to dramatically reduce vehicle emissions
contributing to climate change without significantly reducing convenience or mobility.
Despite their potential, EV market share remains low, necessitating research to identify
factors that could encourage more widespread adoption. For instance, concern about climate
change is associated with intent to adopt an EV, but little is known about mechanisms
through which this concern may translate into action. This study builds on previous work
investigating the roles of symbolic and instrumental attributes in low-emission vehicle
adoption, focusing exclusively on EVs to better understand perceptions associated with their
unique technical capabilities. Prior work has examined symbolism rather generally (e.g., in
terms of status). We examine specific aspects of self-identity that EVs may reflect,
representing the extent to which consumers perceive EVs as symbols that they are
environmentalists and/or social innovators. In addition, extending prior work, we quantify the
relative influence of these separate aspects of symbolism on EV adoption intentions alongside
instrumental, psychological, and demographic factors. We find differing impacts of these two
symbols on EV adoption intentions. Environmentalist symbolism is consistently the strongest
predictor of adoption, across three dependent variables. Innovator symbolism predicts
willingness to lease/buy an EV, trailing only environmentalist symbolism in effect size, and
outperforming instrumental attributes as well as psychological and demographic factors.
Additionally, we examine a potential mechanism through which concern about climate
change may translate into EV adoption intentions: we find that seeing EVs as
environmentalist and social innovator symbols partially mediates the relationship between
concern about climate change and EV adoption intentions. These results have implications
for EV marketing and policy, and suggest that emphasizing the potential for EVs to reinforce
specific self-identities may be a more promising strategy to increase adoption rates than
emphasizing instrumental attributes such as fuel efficiency.
3.2 Introduction
Significant reductions in anthropogenic emissions of greenhouse gases, including carbon
dioxide, are necessary to stem climate change and its associated consequences (Allen et al.,
2009; Meinshausen et al., 2009). In the United States (US), the transportation sector accounts
for roughly 28% of all greenhouse gas emissions, with light duty vehicles making up 62% of
transportation sector emissions (US Environmental Protection Agency, 2015b). Gasoline
vehicles emit myriad additional pollutants with deleterious health effects, such as nitrogen
oxide and fine particulate matter (Brugge, Durant, & Rioux, 2007), positioning the
transportation sector in an important role for public health (World Health Organization, 2005;
Zhang & Batterman, 2013).
Alternative fuel vehicles offer one promising solution to these issues. In particular, fully
electric vehicles (henceforth referred to as “EVs”) receive 100% of their energy from the
electrical grid, allowing for very low emissions if they are charged on low carbon-intensity
Dissertation by Lee V. White
32

electric grids. Notwithstanding the US’s currently coal-heavy electricity portfolio, the
growing share of renewable energy sources and improving efficiency of power plants allow
EVs to reduce total average emissions to nearly half those of a gasoline-fueled vehicle (US
Department of Energy, 2015a).
Additionally, efforts are being made to integrate EVs into the power grid using “Vehicle to
Grid” technology, whereby EV batteries are used as distributed storage. In light of this
emerging technology, a large EV fleet could offer additional benefits to the power sector
including power grid regulation, spinning reserve, peak load shaving, and load leveling (Tan,
Ramachandaramurthy, & Yong, 2016). In some situations, it may even be possible for EVs to
obviate the need for additional electricity generation, for instance, by discharging unused
battery energy back onto the grid during peak demand periods (Jochem, Babrowski, &
Fichtner, 2015).
Despite these advantages and various government subsidies, EVs accounted for only 0.7% of
US market share in 2015 (IEA, 2016). Although the availability of financial incentives is
positively correlated with EV adoption rates, price signals represent only one predictor of EV
adoption; these decisions cannot be understood without considering additional factors,
including symbolic attributes, that consumers perceive as important in these purchases
(Heffner, Kurani, & Turrentine, 2007; Kurani, Turrentine, & Heffner, 2006; Steg, 2005).
Developing a more comprehensive understanding of the motivators and barriers to consumer
adoption of EVs is imperative to improving their market penetration.
3.2.1 Contributions
Previous work examining symbolic attributes related to EV adoption intentions has
operationalized symbolic attributes somewhat generically in terms of whether EVs convey a
positive message about their drivers (E. Noppers, Keizer, Bockarjova, & Steg, 2015;
Schuitema, Anable, Skippon, & Kinnear, 2013). This leaves an open question of what
particular positive messages drivers may be seeking. However, little research has investigated
specific aspects of symbolism and how these are perceived to reflect on one’s self-identity;
the few studies that have done so have either studied hybrid electric vehicles (hybrids; see
last paragraph in this section for why this may not accurately represent EVs), or have used
analytic approaches that fell short of examining multiple components of symbolism while
accounting for other important predictors of adoption simultaneously (Axsen & Kurani, 2013;
Krupa et al., 2014). We advance the literature on symbolism by operationalizing EV
symbolic attributes as the extent to which people perceive EVs to reinforce specific aspects of
self-identity. This study examines the extent to which EVs may symbolize two specific
aspects of self-identity: environmentalist and social innovator. Further, we evaluate the
separate influences of these symbolic attributes in multiple regression models alongside
instrumental, demographic, and psychological predictors of adoption.
The climate change mitigation potential of EVs is a commonly cited driver of consumer
interest in these vehicles. Previous studies have established the influence of concern about
climate change on EV adoption intentions (Carley, Krause, Lane, & Graham, 2013; Egbue &
Long, 2012; Skippon & Garwood, 2011). We extend this by examining whether symbolic
attributes may be a mechanism through which concern about climate change translates into
willingness to adopt EVs. This area of investigation has implications for efforts to promote
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
33

EV uptake, as appeals to reflect self-identity may be more effective in promoting EV
adoption than elevating levels of concern over climate change (Steg et al., 2014).
Additionally, this study focuses exclusively on battery-operated EVs, excluding hybrids.
Unlike EVs, hybrids still utilize the familiar technology of combustion engines, and can be
refueled at regular gas stations. Technical attributes of EVs and hybrids differ considerably,
particularly regarding concerns about vehicle range (Rezvani, Jansson, & Bodin, 2015).
Hybrids may be owned and operated with very little change of habit, and as such may
represent an easier-to-adopt pro-environmental behavior (PEB) compared to EVs. Hybrids
also still produce emissions directly, so may reflect environmentalist self-identity to a lesser
extent compared to EVs. For these reasons, results from studies focused on hybrid adoption
may not fully transfer to EVs.
3.2.2 Predictors of EV adoption intentions
Predictors of EV adoption can be divided across multiple dimensions. Following Steg’s
(2005) categorization of instrumental (associated with practical concerns such as cost) and
symbolic (associated with imagery and identity aspects of EVs) attributes, below we review
the evidence for these two types of predictors. We additionally examine evidence for
demographic and psychological predictors. Groups of predictors examined are summarized in
Figure 3.1.
Figure 3.1: Expected predictors of EV adoption intentions, and their hypothesized directions

Dissertation by Lee V. White
34

3.2.2.1 Symbolic attributes
3.2.2.1.1 Symbolic attributes and inconsistencies in the literature
Steg et al. (2005) introduce the term symbolic attributes to refer to the group of factors
affecting car choices due to emotions and symbolism associated with cars. Symbolic
attributes associated with EVs have been linked to the concept of identity, such that the
symbolism associated with EVs can “construct and express identity” (Rezvani et al., 2015)
and “define and express self-identity and social status” (Burgess, King, Harris, & Lewis,
2013), with symbolic attributes being “related to a sense of self or social identity that is
reflected by, or built from the possession of new technologies” (Schuitema et al., 2013).
Previous studies have established the importance of symbolic attributes for predicting EV
adoption (E. H. Noppers et al., 2014; E. Noppers et al., 2015), and have found that symbolic
attributes can mediate the effect of perceived instrumental attributes on EV purchase
(Schuitema et al., 2013).
However, the term “symbolic attributes” has been used somewhat inconsistently across
studies to describe groups of questions which seem to tap slightly different psychological
constructs. Furthermore, most previous assessments of symbolic attributes have fallen short
of assessing a range of unique symbols relevant to self-identity. See Table 3.1 for a summary
of questions from prior studies which have purported to measure symbolic attributes or
imagery associated with EVs. Most studies applying the term “symbolic attributes” have
operationalized symbolism quite broadly, with questions such as “An electric car gives me
status“ (E. H. Noppers et al., 2014; E. Noppers et al., 2015; Schuitema et al., 2013). In some
cases, symbolism has been operationalized to reflect (environmental) benefits of cars without
specific reference to status, image, or self-identity (Nayum & Klöckner, 2014). What might
the broad category of “status” mean to different individuals? The questions used in previous
work generally do not tap into specific types of status, nor aspects of self-identity, that people
may be trying to reflect (E. H. Noppers et al., 2014; E. Noppers et al., 2015; Schuitema et al.,
2013). It is important to define these specific aspects, as individuals may value some aspects
of self-identity more highly than others when considering adopting EVs. Therefore, for the
purposes of this paper, we define symbolic attributes as attributes which reflect specific
aspects of self-identity.
A few studies have assessed specific aspects of self-identity reflected by EVs, but have not
accounted for the influences of these variables alongside those of instrumental EV attributes
in models of adoption intentions (Table 3.1). For instance, Krupa and colleagues (2014)
assessed environmental and technological imagery, but only included environmental imagery
in a multivariable model (which was not significantly associated with adoption intentions),
and examined intent to adopt compact hybrids rather than EVs. Axsen and Kurani (2013)
assessed environmental, intelligent, and responsible imagery associated with EVs, but these
questions asked about others’ perceptions of the vehicle itself, not individuals’ perception of
how the vehicle reflects on them. As well, they primarily relied on chi-squared analyses.
These approaches limit conclusions about the extent to which specific aspects of symbolism
predict EV adoption (and adoption intentions), as they have not fully adjusted for the impacts
of other established predictors of adoption (for review, see Rezvani et al., 2015).

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
35

Table 3.1: Operationalization of low-emission vehicle symbolism in prior quantitative work,
compared to present study operationalization
Authors (year)
location, sample

Vehicle
type

Questions Term
used by
authors to
describe
construct
Analyses
applied to
symbolism
questions
Noppers et al. (2014)
Netherlands,
community sample

EVs “The electric car shows who I am”
“The electric car enhances my
social status”
(a total of eight items that load
onto the same factor, not all
detailed in paper)

Symbolic
Attributes
Bivariate
correlations;
Multiple
regression
Noppers et al. (2015)
Netherlands,
commercial sample
EVs To what extent do you think that
the following characteristics are
advantages of an electric vehicle?
‘An electric car gives me status’
‘An electric car enables me to
distinguish myself from others’
‘I can show who I am with an
electric car’
‘An electric car fits me’
‘An electric car says something
about me’
Symbolic
Attributes
Multiple
regression
Schuitema et al.
(2013)
United Kingdom,
recent car buyers

Hybrids
and EVs
‘Compared to a normal car, plug-
in hybrid electric cars/plug-in
fully electric cars not suitable for
my lifestyle’
‘I would feel proud of having a
plug-in hybrid electric car/plug-in
fully electric car outside my
house’
‘I would feel embarrassed to drive
a plug-in hybrid electric car/plug-
in fully electric car’
Symbolic
Attributes
OLS
regression
testing
mediation
effects
Krupa et al. (2014)
US, online sample

Hybrids ‘Owning a [hybrid] would make a
statement regarding my strong
environmental values’
‘Owning a [hybrid] would make it
clear to others that I am on the
forefront of new technology’

Imagery Spearman
correlations;
Multiple
logistic
regression
including
one
symbolism
item
Nayum & Klockner
(2014)
Norway, community
sample

Fuel-
efficient
cars
How important did you find
following aspects for you when
you made purchasing decision of
your new car?
Environmentally friendly
materials
Fuel economy
CO 2 reducing tires
The energy label of the car
Symbolic
Attributes
Structural
Equation
Modelling
Dissertation by Lee V. White
36

Greenhouse gas emissions
Emission of polluting chemicals
Axsen & Kurani
(2013)
US, new vehicles
drivers

All
passenger
vehicles
including
EVs
Other people will think this
vehicle looks…
…intelligent
…responsible
…supportive of the environment
…supportive of the US
…powerful
Imagery Chi-square
analysis
Present study (2017),
US, power utility
customers
EVs ‘Owning an EV demonstrates to
others that I care about the
environment’
‘Changing from a gasoline-
powered vehicle to an EV will
lessen my impact on the
environment’
‘Driving an EV means that I am
doing the right thing’
‘Driving an EV means that I am a
trendsetter for environmentally
friendly technologies’
‘Driving an EV means that I am
socially responsible’
Symbolic
Attributes
reflecting
aspects of
self-
identity
Multiple
OLS
regression,
mediation
analyses

We identified three aspects of symbolism suggested by prior literature. In particular, previous
work suggests that EVs have the potential to communicate a variety of specific meanings on
behalf of their owners, including care for the environment, support for new technology, and
concern for general social welfare (Heffner et al., 2007). We consider these as three symbolic
reflections of self-identity: environmentalist, innovator, and socially responsible citizen, and
examine them in the following sections 3.2.2.1.2 to 3.2.2.1.4.
3.2.2.1.2 EVs as symbols for environmentalists
Self-identity can be defined as “a set of meanings attached to roles individuals occupy in the
social structure, and unique ways in which they see themselves in these roles" (Barbarossa,
Beckmann, De Pelsmacker, Moons, & Gwozdz, 2015). Previous research has established that
pro-environmental self-identity, or the extent to which one sees oneself as an
environmentalist, is a predictor of PEB in general (Ellen Van der Werff et al., 2013b;
Whitmarsh & O’Neill, 2010). EVs can serve as highly visible symbols for consumers wishing
to advertise environmentalist self-identities to others, and may also reinforce existing
environmentalist self-identities by acting as a symbol to oneself (E. H. Noppers et al., 2014;
Sexton & Sexton, 2014).
Supporting this idea, studies have found associations between environmental self-identity and
favorability towards EVs (Axsen, TyreeHageman, & Lentz, 2012; Graham-Rowe et al.,
2012), higher ratings of EV instrumental and symbolic attributes (Schuitema et al., 2013),
and perceptions that EVs are symbols of environmental concern (Axsen & Kurani, 2013).
Environmentalist self-identity has also been found to influence consumer attitudes towards
EVs, and in turn has been found to enhance intentions to adopt EVs (Barbarossa et al., 2015).
Hence, the extent to which respondents perceive EVs as symbols to reflect environmentalist
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
37

self-identity should enhance adoption likelihood. Although previous studies have shed light
on related questions, this particular topic remains under-explored.
3.2.2.1.3 EVs as symbols for innovators
As noted by Heffner et al. (2007) with reference to hybrid vehicles, emerging vehicle markets
have the potential to embody new combinations of meanings. In addition to acting as a
symbol for environmentalists, EVs may also act as symbols for consumers who wish to show
an affinity for technological innovation as part of their self-identities.
Consumers often adopt technologically innovative systems due to enjoyment of innovations
and technical aspects, even in absence of economic benefits or environmental concerns
(Schelly, 2014). Cars are publicly consumed products, and social identity concerns have been
found to drive consumer adoption of such products, being both highly visible and projecting
an image of technological innovativeness (Grewal, Mehta, & Kardes, 2000). For instance,
some consumers are willing to pay more for EVs if they perceive them as superior to existing
technology (Hahnel, Ortmann, Korcaj, & Spada, 2014), and symbolic attributes more
strongly predict interest in EVs by potential early adopters if instrumental attributes are
evaluated poorly (E. Noppers et al., 2015). People have also expressed that they see symbolic
meaning in hybrids for marking their self-identity as part of a “technological vanguard”
(Heffner et al., 2007). Noppers et al. (2014) concluded that EVs signal a status motivation
separate from environmental concern, and although unmeasured, that EVs may also signal
innovativeness, suggesting that future research explore this topic.
If consumers enjoy interacting with new technology and view EVs as a way to set new
trends, this may increase their enjoyment of driving an EV, which in turn could make EV
adoption more likely (Steg et al., 2014). In fact, some consumers who perceive that EVs are
not easy to use have stronger adoption intentions than those who think they are easy to use,
which may indicate that they see status in adopting a new and somewhat challenging
technology (Peters & Dütschke, 2014).
3.2.2.1.4 EVs as a symbol of social responsibility
EVs can embody meanings which are only secondarily related to the environment. Notably,
EVs can embody support for the nation (Axsen & Kurani, 2013), or opposition to war and
imported oil (Heffner et al., 2007). Consumers who purchase EVs with these types of
meaning in mind may feel that EVs serve as a way to show social responsibility, for instance
by supporting issues that are not directly associated with “green” imagery. In fact, consumers
without previous histories of environmental purchasing have responded positively to EVs as a
way to construct new identities encompassing moral concern and care for others (Heffner et
al., 2007).
Based on prior literature, we advance the following hypotheses:
EVs will be perceived as reflecting multiple unique aspects of self-identity (H1a), in
particular, environmentalist, innovator, and socially responsible self-identities (H1b)
Perceptions that EVs symbolize unique aspects of self-identity will each positively
predict intent to adopt an EV (H2a), with environmentalist symbolism acting as the
strongest symbolic predictor of EV adoption intentions (H2b)
Dissertation by Lee V. White
38

3.2.2.2 Concern about climate change and the role of symbolic attributes
In EV research, concern about climate change is often assessed by asking whether individuals
view climate change as a problem (Carley et al., 2013; Krupa et al., 2014) and/or by asking
whether the emissions reduction potential of EVs is seen as a benefit (Egbue & Long, 2012;
Krupa et al., 2014; Skippon & Garwood, 2011). Both measures are associated with stronger
EV adoption intentions (Carley et al., 2013; Egbue & Long, 2012; Krupa et al., 2014;
Skippon & Garwood, 2011). However, previous literature has fallen short of determining
whether or how concern about climate change may act to influence adoption through linkages
with sense of self. For instance, environmentalist consumers may be especially motivated to
adopt EVs as an emissions reduction tool (Steg et al., 2014) to respond to their concerns over
climate change.
Self-identity has been found to mediate the relationship between values and behavior,
shedding some light on psychological mechanisms through which environmental concern
may translate into PEB (E. Van der Werff et al., 2014; Ellen Van der Werff et al., 2013b;
Whitmarsh & O’Neill, 2010). It follows that viewing EVs to symbolize aspects of self-
identity may similarly mediate the association between concern about climate change and EV
adoption, though this has not previously been tested. The saliency of EVs as a symbol for
environmentalist self-identity is expected to be stronger amongst individuals who are
concerned about climate change, and in turn, stronger perceptions of environmentalist
symbolism are expected to increase adoption intentions. A similar link is expected for those
viewing EVs as a symbol for innovator self-identity, who may look to EVs as a new
technology that can also help address societal issues.
H3a: Concern about climate change will positively predict intent to adopt an EV, and
this relationship will be mediated by the extent to which EVs are perceived to symbolize
environmentalist self-identity
H3b: Viewing EVs as a symbol for innovator self-identity will also mediate the
relationship between concern about climate change and EV adoption, although to a
lesser extent than viewing EVs as a symbol for environmentalists
3.2.2.3 Instrumental attributes
Instrumental attributes describe practical issues such as economic and convenience concerns
associated with cars. EV adoption is influenced by these instrumental attributes, including
purchase costs, range, and charging locations (Axsen & Kurani, 2013; Caperello & Kurani,
2012; Carley et al., 2013; Graham-Rowe et al., 2012; Anders Fjendbo Jensen, Cherchi, &
Mabit, 2013; Krupa et al., 2014; Wang, Hawkins, Lebredo, & Berman, 2012). Following the
definitions used in Noppers et al. (2014), we consider economic and convenience measures as
instrumental attributes, and review relevant prior work below.
3.2.2.3.1 Economic concerns
Many consumers perceive the purchase cost of an EV as a barrier to adoption. For instance,
in one US-based survey, over half of respondents listed the purchase price as a major
disadvantage (Carley et al., 2013). Moving beyond perceptions to behavior, actual EV
purchase data from several countries offering support policies and infrastructure for EVs
revealed that government rebate programs encouraged EV adoption (Sierzchula, Bakker,
Maat, & Van Wee, 2014). However, a recent US study found that purchase incentives
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
39

predicted greater adoption likelihood for plug-in hybrids but not for battery EVs (Vergis &
Chen, 2015), suggesting that context may play a role in the influence of purchase price.
Related to fuel costs, on average across the US it costs about half as much per gallon-
equivalent to fuel a car using electricity rather than gasoline (US Department of Energy,
2015b). Although consumers often rate fuel savings as a major benefit of EVs (Carley et al.,
2013; Egbue & Long, 2012), this is not always reflected in intent to purchase EVs (Carley et
al., 2013). This may be related to the fact that the savings achievable by switching from
gasoline to electricity as a fuel are not immediately apparent, and consumers are sometimes
concerned about or fail to fully appreciate fuel savings repaying the higher purchase cost over
the vehicle’s lifetime (Graham-Rowe et al., 2012). Consumers have also been found to have
little idea how much they spend on gasoline in any given month (Turrentine & Kurani, 2007),
and may be unable to determine actual fuel cost savings associated with an EV. Studies have
found varying degrees of success in increasing adoption intentions by providing consumers
with five-year fuel savings or total cost of ownership data (Dumortier et al., 2015; Nixon &
Saphores, 2011), highlighting that price is not the only determinant of EV purchase, and that
people are not making purely rational decisions in purchasing EVs.
3.2.2.3.2 Convenience: Charging and Range
Perceptions of the convenience of EVs also impact adoption intentions, with reduced range
and long charging times being two major factors that reduce likelihood of adoption. Limited
EV range has had a major negative impact on adoption intentions in some studies (Anders F.
Jensen, Cherchi, & de Dios Ortúzar, 2014; Anders Fjendbo Jensen et al., 2013) and range
limitations have been linked to consumer frustration in others (Graham-Rowe et al., 2012).
However, some studies have found no relation between actual driving patterns and intent to
purchase an EV (Carley et al., 2013; Kurani, Turrentine, & Sperling, 1996), and current EV
drivers are generally satisfied with available range (Franke & Krems, 2013; Pearre, Kempton,
Guensler, & Elango, 2011). In sum, the findings related to range are somewhat inconclusive.
It is possible that different individual or household level characteristics affect the impact of
range on adoption. For instance, individuals living in households with access to multiple
vehicles did not find EV range limitations to be a barrier (Kurani et al., 1996), as they could
use a gasoline vehicle for longer trips. Additionally, some consumers may view adaptation to
instrumental limitations such as shorter range as a symbolic benefit of EVs (Axsen & Kurani,
2013). Further research is needed to understand for whom EV limited range is a benefit vs. a
barrier.
The ability to charge an EV with minimal disruption to typical routines has emerged as
another important convenience factor associated with higher likelihood of adoption.
Consumers who needed to alter their daily routines to allow EVs time to charge at
workplaces reported frustration at the limitations imposed on their movement (Graham-Rowe
et al., 2012), and those who trialed EVs reported higher willingness to pay if they had the
option to charge at work or access to public chargers (Anders Fjendbo Jensen et al., 2013).
Charging at home has been found to be most preferred, with workplace charging rated second
(Skippon & Garwood, 2011). Further, the ability to charge at home, thereby avoiding
refueling at gas stations, may be considered a positive attribute of EVs relative to gasoline
vehicles (Kurani et al., 1996). Overall, findings support convenient charging as a factor that
encourages EV adoption.
Dissertation by Lee V. White
40

3.2.2.4 Demographic characteristics
Previous empirical studies examining the relationship between demographic characteristics
and EV adoption have yielded inconclusive or conflicting results. With respect to gender for
instance, men have generally been found to express more interest in EVs than women (Egbue
& Long, 2012; Peters & Dütschke, 2014; Plötz, Schneider, Globisch, & Dütschke, 2014).
However, interest does not always translate into intent to purchase (Egbue & Long, 2012).
Furthermore, in some cases, gender has not been found to affect interest in EVs at all (Anders
F. Jensen et al., 2014), and another study controlling for psychological variables yielded
slightly higher adoption intentions among women (Nayum & Klöckner, 2014). Thus, the
literature does not provide a conclusive prediction of the impact of gender on EV adoption
intentions.
Demographic findings related to age, income and education are similarly mixed. Regarding
age, whereas older individuals have generally been found to express greater interest in EVs
(Barth, Jugert, & Fritsche, 2016; Egbue & Long, 2012), adoption intentions are generally
higher among middle-aged individuals (Peters & Dütschke, 2014; Plötz et al., 2014).
Additional studies have found that younger individuals express stronger intentions to
purchase an EV (Carley et al., 2013; Hidrue, Parsons, Kempton, & Gardner, 2011), and
higher willingness to pay (Achtnicht, 2012). However, other work controlling for
psychological factors found no significant impact of age (Nayum & Klöckner, 2014). With
respect to income, some studies have found higher earnings to have no impact (Carley et al.,
2013; Egbue & Long, 2012), or only marginal impacts (Barth et al., 2016) on adoption
intentions, yet another found a small positive effect even when controlling for psychological
factors (Nayum & Klöckner, 2014). Finally, higher education levels have been found in
various studies to predict interest in EVs but not necessarily purchase intent (Egbue & Long,
2012), to have a weak positive effect on adoption (Nayum & Klöckner, 2014), to have a
negative effect on adoption intent (Carley et al., 2013), and to have no effect on adoption
(Moons & De Pelsmacker, 2012; Peters & Dütschke, 2014). Concluding, the findings on
associations between EV adoption intentions and gender, age, income, and education are
mixed, necessitating further work to clarify our understanding of these relationships.
3.2.2.5 Past pro-environmental behavior
Broad-ranging research highlights the importance of accounting for past behavior in
predicting future behavior (Ouellette & Wood, 1998). We consider that past PEB may impact
EV adoption intentions. Supporting this notion, Whitmarsh and O’Neill (2010) note that past
PEB significantly influences intentions to perform PEBs in the future. Past PEB has been
linked to stronger intentions to use EVs (Moons & De Pelsmacker, 2012), but has not
previously been considered as a predictor of EV adoption intentions alongside EV symbolic
attributes (Axsen & Kurani, 2013; Krupa et al., 2014).
3.2.2.6 Social norms
Social norms (broadly, the social pressure to behave “appropriately”), can be encouraged by
observing others meeting norms for a given situation (Keizer et al., 2013). Thus, if consumers
perceive that many people around them are taking positive action for the environment and
society by adopting EVs, they may perceive it as the “appropriate” thing to do, and develop
stronger intentions to adopt EVs themselves. Supporting this idea, social norms related to
EVs have been found to predict EV adoption intent (Peters & Dütschke, 2014). However,
previous studies have not attempted to directly influence social norms as they pertain to EVs,
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
41

for instance, through messaging that aims to give consumers the impression that their peers
are engaging in PEB through EV purchases. Our study incorporated a messaging experiment
in which respondents were randomly assigned to receive a message focused on either
financial benefits of EVs or social norms surrounding EVs, the latter of which was intended
to examine this potential.
H4: Receipt of social norms messaging will predict stronger intent to adopt an EV than
receipt of financial messaging
3.3 Methods
The following section describes how we used survey data from a sample of Los Angeles
(L.A.) County residents to test our hypotheses.
3.3.1 Procedures
Surveys were mailed to a random sample of residential customers of a large power utility in
southern California in 2014. Sequentially, surveys first assessed dependent variables
assessing dimensions of EV adoption intentions, followed by instrumental attributes,
symbolism, and psychological characteristics, with demographic variables assessed last.
Additionally, households were randomly assigned to receive one of two persuasive cover
letters introducing the study and inviting participation: half the sample received a social
norms-focused persuasive message, emphasizing the increasing popularity, preponderance,
and public acceptance of EVs in L.A. The other half received a financially-focused
persuasive message, emphasizing the financial benefits of EVs including lower fuel and
maintenance costs.
3.3.2 Participants
A total of 481 surveys were returned, including 218 from respondents who had received
financial messaging and 263 from respondents who had received social norms messaging. Of
these, 437 respondents answered all three dependent variables. 124 individuals with missing
data on predictor variables were then dropped. An additional two respondents who provided
outlier responses of over 10,000 miles to questions asking for perceived EV range and daily
driving distances were coded as having missing responses to these questions, raising the
missing count to 126. This yielded a final sample of 355 observations, which includes 155
and 200 respondents who received financial and social norms persuasive messages,
respectively.
The sample had a median age of 53.0 years and was roughly half female. Nearly three-
quarters had bachelor’s degrees, and approximately two-thirds identified as Caucasian, with
one-third identifying as minority or multi-racial. Compared to L.A. county, our sample was
older, had higher income and education levels, and contained a lower proportion of
individuals identifying as a minority ethnicity or multi-racial. See Table 3.2 for a summary of
sample characteristics.

Dissertation by Lee V. White
42

Table 3.2: Participant demographic characteristics relative to Los Angeles County.
Characteristic Sample (%)
a
Los Angeles County
b
(%)
Ethnicity
African American
Asian / Asian American
Caucasian
Latino
Native American / Pacific Islander
Multiracial
Other

2.3
11.9
67.6
10.7
0.6
4.9
2.0

8.1
13.7
27.5
47.9
0.2
2.1
0.2
Educational Attainment (>= 25 yrs)
Less than High School
High school diploma
Some college / Associate’s degree
4-year college degree
Graduate / professional degree
High School diploma or higher
Bachelor’s degree or higher

1.1
4.3
22.1
37.7
34.8
98.9
72.5

23.4
20.5
26.5
19.4
10.2
76.6
29.7
Annual household income
<$25,000
$25,001-$50,000
$50,001-$75,000
$75,001-$100,000
>$100,000

9.9
13.6
20.4
13.6
42.5

Median household income:
$55,909
Home ownership rate 60.8 46.9
Home type
Single Family Home
Apartment/Condo
Duplex, Triplex
Townhouse
Mobile Home
Other

52.8
35.9
5.9
4.5
0.6
0.3

49.7
34.3
8.0
6.5
1.5
Gender (% male) 51.8 49.3
Age (median) 53.0 35.1
a
Not all respondents answered all demographic questions. Available data for respondents
who answered questions is as follows: N ethnicity = 346, N education=353, N income=294, N home
ownership=355, Nhome type=354, Ngender=284, Nage=246.
b
(US Census Bureau, 2013).
3.3.2.1 Drop-out analyses
Wilcoxon rank-sum tests were used to compare the median distribution of each dependent
variable (see section 3.3.3.1 below for variable details) among the final main dataset (355
observations) to those excluded due to missing data (126 observations in total, though only
103-106 observations could be included in each test as many incomplete responses did not
include the necessary dependent variable). The Wilcoxon rank-sum test is the nonparametric
alternative to an independent samples t-test, and was used given the non-normal distributions
of the dependent variables. No differences between completers and non-completers were
observed for willingness to lease/buy. However, the distribution of completers was
significantly different from that of non-completers for willingness to pay and impressions,
such that completers scored significantly higher (for willingness to pay M pay complete = 5.46,
Mpay dropped = 4.76, Mdnpay complete = 5, Mdnpay dropped = 4, U = 15934.5, p < 0.03; for
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
43

impressions M impressions complete = 3.99 M impressions dropped = 3.64, Mdn impressions complete = 4,
Mdn impressions dropped = 4, U = 15205.5, p < 0.01). This suggests that the respondents included in
the present study were more favorable towards EVs, indicating that the results may not
generalize to the general population. Results should be interpreted with caution in light of
this.
3.3.3 Measures
The survey assessed three dependent variables and a number of independent variables,
grouped into the following categories: instrumental attributes, symbolic attributes,
demographic factors, and psychological factors. Table 3.3 provides summary statistics for
these variables.
3.3.3.1 Dependent variables
The survey assessed three dependent variables selected to reflect increasing levels of
commitment to EV adoption
4
. First, respondents provided their general impression of EVs on
a 5-point scale ranging from 1 = Strongly Unfavorable to 5 = Strongly Favorable. Second,
they rated their willingness to lease or buy an EV as their next personal vehicle on a 10-point
scale ranging from 1 = Extremely Unlikely to 10 = Extremely Likely. Finally, participants
were asked to imagine they were interested in a car model that was also available in an all-
electric vehicle. They were asked how much more they would be willing to pay for the EV
compared to the conventional gas vehicle, with all other features of the vehicle being
identical. Responses were provided on a 13-point scale ranging from 1 = $0 to 13 = $6000
5
.
In all cases, a higher number corresponds to stronger EV adoption intentions.
3.3.3.2 Symbolic attributes
The survey included five questions focused on specific aspects of self-identity that EVs have
been found to reflect (see Table 3.4 for question text). Responses were given on a 7-point
Likert scale ranging from 1 = Strongly Disagree to 7 = Strongly Agree, with higher numbers
indicating stronger endorsement of self-identity symbols in EVs.
3.3.3.3 Instrumental attributes
3.3.3.3.1 Cost
A total of three questions asked respondents to rate perceived EV costs compared to gasoline
vehicle costs, using a 5 point Likert response scale where 1 = Much less and 5 = Much more.
Questions separately assessed perceptions of (1) purchase costs, (2) maintenance costs, and
(3) fuel costs.
3.3.3.3.2 Charging Convenience
Charging convenience was assessed with a binary variable. Respondents were asked “Do you
have access to an electrical outlet where your car is parked at your primary residence?” with
responses of no coded as “1” and yes as “2”.
3.3.3.3.3 Range and driving distance
Respondents were asked two open ended questions regarding driving distances. One question
assessed perceived EV range, (“Approximately how many miles do you believe a fully-

4
Following Schuitema, we use the common assumption that Likert scales can be treated as interval-level data
and analyzed using parametric methods (Schuitema et al., 2012; Kline, 2000; Nunally, 1978).
5
This type of willingness to pay scale has been considered to lead to more valid values than using open-ended
prompts, with respondents providing fewer zero values and with predictors explaining a greater amount of
variance in willingness to pay (Donaldson, Thomas, & Torgerson, 1997).
Dissertation by Lee V. White
44

charged EV can drive before the battery is drained?”). A second question assessed estimated
daily travel mileage (“On an average day, approximately how many miles do you drive in a
car that you lease or own? Include trips made for work, school, shopping, errands,
entertainment, etc.”). Responses to both questions were open-ended and requested in miles,
and were coded as numeric integers. Additionally, we examined the difference between
perceived EV range and daily driving distance (calculated as perceived EV range – estimated
daily travel mileage, coded as a numeric integer); this was intended to measure perceptions
that an EV could fulfill average daily driving needs on a single charge. Our main analyses
(Tables 3.5 and 3.6) include only the perceived EV range variable. However, we also ran the
models using the difference between perceived EV range and estimated daily travel mileage,
and found similar patterns of results.
3.3.3.4 Psychological characteristics
3.3.3.4.1 Concern about climate change
We adapted a scale from Bostrom et al. (2012) to examine concern about climate change. Our
scale comprised five questions (“To what extent are you concerned about air pollution in your
city? “, “How serious a threat is climate change to humankind?”, “How well is climate
change understood by science?”, “How much does the idea of climate change fill you with
dread?”, and “To what extent do you have moral concern about climate change?”). Each was
rated on a 7-point Likert scale where a higher number corresponded to a stronger concern
about climate change. Factor analysis showed that the variables loaded onto one factor. To
create a climate change concern scale, we took the mean of responses to available items
6
.
3.3.3.4.2 Past pro-environmental behavior
Past PEB was assessed with three items. Respondents rated, on a scale of 1 (Strongly
Disagree) to 7 (Strongly Agree), whether they made an effort to recycle, whether they had
purchased lightbulbs that were more expensive but saved energy, and whether they had
encouraged family or friends not to buy environmentally harmful products. A scale was
created by taking the mean of available responses (at least two of the questions had to be
answered to receive a scale score).
3.3.3.4.3 Relevance of cars in general to self-identity
Respondents rated, on a Likert scale of 1 to 7 with 1 = Strongly Disagree and 7 = Strongly
Agree, the statement “I think the kind of car a person drives says something about the
person”. This “car identity” variable separates out variance associated with the belief that cars
impact identity in general (i.e., generic symbolic attributes), from variance associated with
the ability of EVs to reinforce aspects of self-identity corresponding to one or more symbolic
reflections of self-identity.
3.3.3.5 Demographic characteristics
Income. Income was provided on a categorical scale (see Table 3.2). For analyses, we created
a binary variable, split roughly at the median income for L.A. county of $55,909 (US Census
Bureau, 2013). Those with incomes of $50,000 or greater received 1, and others received 0.

6
Only respondents who answered more than half of items (at least 3) were retained in the final sample and
received a scale score.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
45

Education. Education was assessed using a categorical scale (see Table 3.2 for response
options). For analyses, we created a binary variable, coded 1 if respondents had a four-year
college degree or higher, and 0 otherwise.
Age. The survey assessed birth year. Age (in years) was calculated accordingly and coded as
a continuous variable.
Gender. Male gender was coded as 1, and female was coded as 0.
Table 3.3: Descriptive statistics for key variables (main sample, N=355)

Response
range

Mean Std.
Dev
Dependent Variables
(N = 355)
Impressions 1-5 3.99 1.13
Willingness to lease/buy 1-10 5.48 3.07
Willingness to pay 1-13 5.46 3.36
Symbolism of unique
aspects of self-identity
(N = 355)
Environmentalist
a
1-7 5.22 1.39
Social Innovator
b
1-7 5.07 1.08
Instrumental Attributes
(N = 355)
Purchase cost 1-5 3.85 0.87
Maintenance cost 1-5 2.93 1.17
Fuel cost 1-5 1.77 1.04
Charging convenience 1 / 2 1.47 0.50
Estimated EV range 15-600 126.10 89.71
Estimated daily driving mileage 0-250 24.95 24.00
Psychological
characteristics
(N = 355)
Past PEB
c
1.67-7 5.70 1.00
Concern about climate change
d
1.4-7 5.39 1.18
Car identity 1-7 5.08 1.42
Demographic
characteristics
(N = 224)
e

Income 0 / 1 76
Education 0 / 1 77
Gender (% male) 0 / 1 50
Age 22-93 48.25 15.79
a
Cronbach’s alpha = 0.87,
b
Spearman’s rho = 0.35,
c
Cronbach’s alpha = 0.56,
d
Cronbach’s alpha =
0.87,
e
Note that the demographic sub-sample of 224 respondents is presented here, for consistency
with the results presented in Table 3.6 which were generated based on this sub-sample.

3.3.4 Data preparation
To test H1a and H1b we conducted iterative principal factor analysis using our full sample of
355 respondents. To test H2a and H2b, we conducted OLS multiple regression analyses. To
test H3a and H3b, we conducted mediation analyses by (1) using causal chain path diagrams
(after Baron and Kenny, 1986) and (2) calculating indirect effect sizes and using the
distribution of the product of coefficients method to test their significance (Tofighi &
MacKinnon, 2011). To examine H4, we conducted an additional univariate OLS analysis.
In constructing OLS models, we visually inspected normal quantile plots of residuals against
the inverse normal for each dependent variable to identify skewness or kurtosis. Excessive
skewness was found for impressions of EVs, so the variable was square-transformed to adjust
for this. The Huber-White test was used to assess heteroscedasticity in our models, which was
found for willingness to pay and willingness to lease/buy. Hence, Huber-White standard
errors were used for all three models, to ensure unbiased estimation of variance even in
presence of heteroscedasticity (White, 1980). Huber-White standard errors were additionally
Dissertation by Lee V. White
46

applied to mediation analysis testing following tests that indicated heteroscedasticity in main
models and these are reflected in both the path diagrams and the indirect effect significance
tests.
3.4 Results
3.4.1 Factor analysis of symbolic attributes
To examine H1a, we conducted iterative principal factor analysis constrained to three factors.
Promax rotation was then applied. Results of this analysis are presented in Table 3.4. Factors
were extracted based on Cattell’s Scree test, which, being less impacted by the number of
variables (Zwick & Velicer, 1986), can be more accurate than traditional K1 methods
(Fabrigar & Wegener, 2011; Osborne & Costello, 2009) although both methods in this case
would yield two factors. The Scree plot indicates a two-factor solution (see Section 8.2:
Chapter 2, Appendix 2) as inter-factor differences in eigenvalues fall off markedly from the
third factor on. This two-factor solution is not in support of H1b, which predicted three
specific aspects of self-identity would be symbolized by EVs. Constraining the solution to
two factors returned similar results, provided in Section 8.3: Chapter 2, Appendix 3.
Following the recommendations of Osborne and Costello (2009), we used a minimum item
loading cut-off of 0.40 to indicate that an item belongs to a factor. The first three questions
loaded onto Factor 1, “environmentalist” symbolism, representing the perception that EVs
reflect environmentalist self-identity. We therefore define the environmentalist symbol as the
average score of the three underlying items. Cronbach’s alpha for these three items was 0.87.
The remaining two questions, which were expected to load separately as innovator and social
responsibility symbolism, both loaded onto Factor 2. The constituent items both reference the
broader social fabric, and furthermore suggest an element of social innovation, indicated by
technology “trendsetting” and taking responsible action for the good of society. Thus, Factor
2 is considered to represent “social innovator” symbolism. A spearman correlation examining
the correlation between the two constituent items found a statistically significant relationship
between the two variables (ρ = 0.35, p = 0.000).

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
47

Table 3.4: Promax rotated factor loadings and uniqueness of symbolic attributes items based
on iterative principal factor analysis constrained to three factors
Question Factor 1:
Environmentalist
Factor 2:
Social Innovator
Factor 3:
Unsupported
Uniqueness
(1) Owning an EV
demonstrates to others
that I care about the
environment
0.81 -0.06 0.13 0.34
(2) Changing from a
gasoline-powered
vehicle to an EV will
lessen my impact on
the environment
0.81 0.04 -0.07 0.32
(3) Driving an EV
means that I am doing
the right thing
0.71 0.26 0.04 0.20
(4) Driving an EV
means that I am a
trendsetter for
environmentally
friendly technologies
0.26 0.59 0.01 0.40
(5) Driving an EV
means that I am
socially responsible
-0.02 0.41 0.21 0.72
Eigenvalues: 2.48 1.70 0.40

3.4.2 Factors impacting EV adoption intentions
Given the findings in section 3.4.1, to address H2a and H2b, we investigated the unique
impacts of “environmentalist” and “social innovator” symbolism on EV adoption intentions,
rather than “environmentalist, “innovator” and “social responsibility” symbolism separately
as initially proposed. To test H2a and H2b, we built three OLS multiple regression models,
each of which examined one of our dependent variables: impressions of EVs, willingness to
lease/buy an EV, and willingness to pay more for an EV than an equivalent conventional
vehicle. Predictors included symbolic attributes (environmentalist and social innovator
symbolism), instrumental attributes (purchase cost, maintenance cost, fuel costs, access to
charging outlet, and perceived EV range for a single charge), psychological characteristics
(concern about climate change, past PEB, and perceived relevance of cars to identity), and
demographic characteristics (income, education, age, and gender). Results are shown in Table
3.5.
Across all three dependent variables, the extent to which EVs were seen as symbolizing
environmentalist self-identity had greater predictive power than any other variable in any of
the models, supporting H2b (see Table 3.5). In addition, the extent to which respondents
perceived EVs to symbolize social innovator self-identity had a unique effect on impressions
and intent to lease/buy an EV (and for both DVs was the second strongest predictor after
environmentalist symbolism), supporting H2a.

Dissertation by Lee V. White
48

Table 3.5: Results of multiple regression analysis predicting EV adoption intention (N =355)
Impressions
(sq)
Willingness to
lease/buy
Willingness to
pay
Symbolic
attributes
Environmentalist 0.48
***
0.28
***
0.27
***

(0.32) (0.12) (0.17)

Social Innovator 0.13
*
0.27
***
0.04
(0.39) (0.16) (0.18)

Instrumental
attributes
Purchase cost -0.01 -0.09
*
0.03
(0.40) (0.14) (0.20)

Maintenance cost -0.07 -0.21
***
-0.17
**

(0.31) (0.13) (0.15)

Fuel cost -0.12
*
-0.01 -0.08
+

(0.34) (0.14) (0.16)

Charging convenience -0.03 0.06 0.20
***

(0.64) (0.25) (0.31)

Estimated EV range 0.01 -0.05 0.02
(0.00) (0.00) (0.00)

Psychological
characteristics
Concern about climate
change
-0.02
(0.36)
0.04
(0.14)
0.08
(0.19)

Past PEB scale 0.07 0.09
+
0.02
(0.41) (0.16) (0.17)

Car identity 0.04 -0.00 0.01
(0.23) (0.10) (0.12)
R
2
0.43 0.44 0.27
N 355 355 355
Standardized beta coefficients; Standard errors in parentheses
+
p < 0.10,
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

Instrumental attributes were also significant predictors of EV adoption intentions, but in
general they had weaker predictive strength than the symbolic attributes. Neither purchase
cost nor fuel costs were large predictors, though higher perceived purchase costs predicted
lower willingness to lease/buy, and lower perceived fuel costs predicted more positive
impressions. Perceiving lower maintenance costs had larger impacts that were significant for
both willingness to lease/buy and willingness to pay. Additionally, charging convenience
predicted willingness to pay with a decent effect size, and had marginal significance as a
predictor of willingness to lease/buy. Surprisingly, perceived EV range was not significantly
associated with any of the dependent variables; nor was daily driving distance or the
difference between estimated range and daily driving distance (tested, but not reported here).
We did not find large impacts of psychological characteristics; concern about climate change
had no effect on EV adoption intent, and past PEB was only a marginal predictor of greater
willingness to lease/buy.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
49

Because there were a large number of respondents missing demographic data, we evaluated
the impact of demographic variables on EV adoption intentions as a sub-analysis (Table 3.6;
see Section 8.1: Chapter 2, Appendix 1 for a summary of descriptive statistics on sub-
sample). Specifically, we limited this sub-analysis to a sub-sample of the 224 respondents
who provided complete data for all dependent variables, independent variables, and
demographics. We added income, education, gender, and age to the OLS model described
above.
In general, the same pattern of results emerged with respect to symbolic and instrumental
attributes. Symbolic attributes, in particular the environmentalist symbol, remained strong
predictors for all three dependent variables, indicating that H2a and H2b are robust to
inclusion of demographic variables. Table 3.6 additionally revealed that higher incomes were
associated with higher willingness to pay. Being female or older was associated with lower
willingness to pay. Demographic variables did not have an impact on the other dependent
variables.

Table 3.6: Results of multiple regression analysis predicting EV adoption intentions
including demographic characteristics (N = 224)
Impressions
(sq)
Willingness to
lease/buy
Willingness to
pay
Symbolic attributes
Environmentalist 0.52
***
0.27
***
0.29
***

(0.46) (0.18) (0.23)

Social Innovator 0.05 0.21
**
0.04
(0.53) (0.22) (0.22)

Instrumental
attributes
Purchase cost 0.04 -0.10
+
0.01
(0.57) (0.20) (0.28)

Maintenance cost -0.12
*
-0.23
***
-0.18
**

(0.37) (0.18) (0.20)

Fuel cost -0.10
+
0.05 -0.05
(0.39) (0.16) (0.18)

Charging convenience -0.02 0.09 0.15
*

(0.85) (0.38) (0.43)

Estimated EV range -0.03 -0.08 -0.01
(0.00) (0.00) (0.00)

Psychological
characteristics
Concern about climate
change
-0.05
(0.50)
0.01
(0.19)
0.01
(0.24)

Past PEB scale 0.08 0.15
*
0.02
(0.56) (0.24) (0.25)

Car identity 0.07 0.07 0.03
(0.28) (0.12) (0.14)

Demographic
characteristics
Income 0.00 -0.03 0.13
*

(0.99) (0.44) (0.46)
Dissertation by Lee V. White
50

Education 0.06 0.00 0.05
(0.92) (0.40) (0.48)

Gender 0.06 0.05 0.19
**

(0.84) (0.36) (0.43)

Age -0.08 -0.04 -0.16
*

(0.03) (0.01) (0.01)
R
2
0.43 0.39 0.28
N 224 224 224
Standardized beta coefficients; Standard errors in parentheses
+
p < 0.10,
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

3.4.3 Concern about climate change: mediation
To test H3a and H3b, we conducted a single-level mediation analysis. Concern about climate
change was the independent variable; the three dependent variables were impressions,
willingness to lease/buy, and willingness to pay; and two mediators were environmentalist
and social innovator symbolism. We began by estimating the three basic mediation equations
for each dependent variable-mediator pair (Baron & Kenny, 1986):
Equation 3.1:

=

+

+

Equation 3.2:

=

+

+

Equation 3.3:

=

+

+ ′

+

where y i represents EV adoption intentions and τ represents the direct effect of the
independent variable (concern about climate change) on adoption intentions; mi represents the
mediator (environmentalist or social innovator symbolism) and α represents the direct effect
of the independent variable on the mediator; and β represents the effect size of the mediator
holding the effect of the dependent variable constant, τ’ represents the effect size of the
independent variable on EV adoption intentions holding the effect of the mediator (either
environmentalist or social innovator symbolism) constant, and x i represents the independent
variable.
Based on the α and β estimates produced by equations 3.2 and 3.3 for each dependent
variable-mediator pair, we calculated the coefficient of the indirect effect (αβ) for each
mediator using the product of coefficients method outlined in MacKinnon et al. (2002). Next,
we tested the significance of each indirect effect size using the distribution of the product of
coefficients method (Tofighi & MacKinnon, 2011). Mediation was considered to occur if the
indirect effect (αβ) was significant, and if additionally the mediation path through both α and
β was significant (see Figure 3.2). Table 3.7 shows the effect sizes, standard errors, and
significance of the indirect effects.
All six dependent variable-mediator pairs showed significant indirect effects (Table 3.7), and
in addition significant paths existed along α and β in path diagrams (Figure 3.2), indicating
that mediation occurred for all six pairs. Large indirect effects found for environmentalist and
social innovator symbolism support H3a and H3b. In addition, indirect effect sizes for
environmentalist symbolism were consistently larger than for social innovator symbolism,
further supporting H3a.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
51

Table 3.7: Indirect effect sizes with standard errors in parentheses. Significance tested using
the distribution of the product of coefficients method.
Impressions Willingness to
lease/buy
Willingness to pay
Environmentalist 0.35 (0.04)*** 0.77 (0.10)*** 0.65 (0.13)***
Social Innovator 0.15 (0.03)*** 0.57 (0.09)*** 0.27 (0.09)**
+
p < 0.10,
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

Figure 3.2 (a, b, and c) presents models of mediation paths. Each path τ describes the direct
effect of concern about climate change on each dependent variable, as described in equation
3.1, while the path τ’ describes the effect of concern about climate change on each dependent
variable when the impact of either environmentalist or innovator symbolism is controlled for.
Baron and Kenny (1986) specify the use of unstandardized coefficients for this comparison.
In Figure 3.2, paths α and β are always positive and statistically significant, and τ’ is always
smaller than τ such that the model passes the difference in coefficients test (Baron & Kenny,
1986; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). Thus, in conjunction with
the findings on the significance of indirect effect sizes presented in Table 3.7, Figure 3.2
supports H3a and H3b that the relationship between concern about climate change and EV
adoption intentions is mediated by environmentalist and social innovator symbolism.
Figure 3.2a-c: Causal chain path diagrams depicting the mediating roles of environmentalist
and innovator symbolism in the relationship between concern about climate change and EV
adoption intentions.
7

2a) Dependent variable: EV impressions

7
Mediation of the relationship between concern about climate change and EV adoption
intention are shown for three dependent variables (impressions, willingness to lease/buy,
willingness to pay). Labels (α, β, τ, τ’) correspond to the paths described in equations 4.1,
4.2, and 4.3. The direct effect is shown by τ, and the indirect path passes through the
mediator via α and β. τ’ represents the size of the former direct path (τ) when the indirect
path through α and β is simultaneously taken into account.
Dissertation by Lee V. White
52

2b) Dependent variable: willingness to lease/buy EV

2c) Dependent variable: willingness to pay

3.4.4 Normative messaging intervention
3.4.4.1 Validity check
A validity check was included whereby responses to two questions were expected to differ as
a function of the persuasive message received. These include the fuel and maintenance cost
questions (see Section 3.3.3.3.1), which we expected those who received the financial
message to rate lower (indicating lower costs associated with EVs vs. conventional vehicles).
Using Wilcoxon rank-sum tests (the non-parametric alternative to the independent samples t-
test, given the non-normal distributions of the relevant variables), we did not find statistically
significant differences on responses to the fuel cost question. However, those who received
financial messaging reported marginally lower maintenance costs than those who received
social messaging (U = 13909.5, p < 0.09, Mfinancial = 2.83, Mdnfinancial = 3, Msocial = 3.02,
Mdn social = 3), which was in the predicted direction. This provides modest evidence that
respondents read messages.
3.4.4.2 Influence of messaging on adoption intentions
A set of regression models was conducted to test H4. Impressions, willingness to lease/buy,
and willingness to pay were the dependent variables, and type of messaging received was the
sole independent variable. Contrary to expectations, type of messaging had no impact on
adoption likelihood (β = 0.05, sd = 0.13 p < 0.38 for impressions, β = -0.01, sd = 0.36, p <
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
53

0.83 for willingness to lease/buy; and β = 0.07, sd = 0.39, p < 0.21 for willingness to pay).
We additionally tested whether messaging type acted as a moderator of the relationships
between each of the dependent variables and income, symbolic attributes, and instrumental
attributes, and found no evidence for moderation in any of these cases.
3.5 Discussion
This study builds on previous findings that symbolic attributes are important predictors of EV
adoption intentions (E. H. Noppers et al., 2014; Schuitema et al., 2013). We refined the
concept of symbolic attributes, defining this construct as “attributes which reflect specific
aspects of self-identity”. We advance prior work which has fallen short of examining
symbolism associated with specific aspects of self-identity, the impacts of such symbolism on
adoption intentions, or symbolism alongside other important predictors of EV adoption
(Axsen & Kurani, 2013; Krupa et al., 2014). Specifically, we define and evaluate two specific
aspects of self-identity that EVs may be perceived to reflect: environmentalist and social
innovator.
3.5.1 Specification of symbolic attributes reflecting aspects of self-identity
Environmentalist, innovator, and social responsibility symbolism of self-identity were
expected to load onto separate factors. Instead, we found two factors: environmentalist and
social innovator symbolism, supporting H1a, but not H1b.
Innovator and social responsibility may load together due to the fact that both questions
tapped a social element, i.e., by asking about being a “trendsetter”, and by asking about
taking responsible action for societal good, both of which indicate setting a positive example
for others in a social group. This aligns with prior suggestions that EVs and hybrids have the
potential to embody new combinations of meanings, and in particular it appears they may
bridge areas of innovation and social responsibility (Heffner et al., 2007). Hybrids are
sometimes seen by their owners as “…symbols of advocating to vehicle manufacturers. By
purchasing [a hybrid], households see themselves as providing support to automakers that
have developed hybrid technology, and punishing those who have not” (Heffner et al., 2007).
This supports the idea that EVs may embody self-identities focused on both innovation as
well as feelings of socially responsibility, i.e., that EVs act as symbols for those who want to
set socially responsible trends.
Also notable is that seeing driving an EV as “doing the right thing” loaded with
environmentalist symbolism in the factor analysis. This highlights that respondents saw
performing PEBs as doing the right thing, which may also indicate past exposure to pro-
environmental social norms. California has been a leader of environmental policy since the
1960s (Sperling & Eggert, 2014), which has involved conveying various pro-environmental
messaging to residents over the years.
3.5.2 The predictive strength of symbolic attributes reflecting self-identity
Both environmentalist and social innovator symbolism are strong and independent predictors
of EV adoption intentions, supporting H2a. This finding expands the literature, which has
identified the importance of similar constructs separately but not examined them together
(Axsen & Kurani, 2013; Krupa et al., 2014; E. H. Noppers et al., 2014). We confirm that
environmentalist symbolism in particular is a powerful predictor for EV adoption intentions,
and although previous research examining impacts of environmental imagery on hybrids used
different analysis methods (Krupa et al., 2014), the greater significance of environmentalist
Dissertation by Lee V. White
54

symbolism as predictor in our model may indicate that this symbolic attribute has a greater
impact on adoption intentions for EVs compared to hybrids; this possibility should be
considered in future studies of low-emissions vehicles.
Additionally, our finding in support of H2b that environmentalist symbolism is the strongest
predictor of EV adoption intentions across three dependent variables, over and above EV
instrumental attributes as well as demographic and psychological characteristics, emphasizes
that this particular type of symbolism should be included in future models. This construct is
distinct from the measures of concern about climate change that have often previously been
examined (Carley et al., 2013; Egbue & Long, 2012; Krupa et al., 2014; Skippon &
Garwood, 2011), and also distinct from constructions of environmental attributes examined
previously
8
(E. H. Noppers et al., 2014; Schuitema et al., 2013) due to its measurement of EV
reflections of environmentalist self-identity. Environmental self-identity has been linked to
feeling a moral obligation to act pro-environmentally, which in turn affects PEBs such as EV
adoption (Ellen Van der Werff, Steg, & Keizer, 2013a). The predictive strength (for EV
adoption intentions) of seeing EVs to symbolize environmentalist self-identity extends these
findings, and may in part be due to desires to show that one has acted on this type of moral
obligation.
Providing context for our findings on social innovator symbolism, altruistic signaling can
serve as a way to gain social status, particularly if such a signal is costly (costly signaling;
Griskevicius et al., 2010). This may explain the significance of innovator symbolism for
predicting willingness to lease/buy an EV. Paying additional money to set an environmentally
friendly trend may be a new form of altruistic signaling offered by EVs, as suggested by the
observation in Heffner et al. (2007) that EV purchases may be made with the intent to reward
automakers for providing environmentally-friendly options. This is supported by previous
observations that EV symbolism was rated higher by potential early adopters if they
perceived instrumental attributes to be lower (Noppers et al., 2015). Additionally, consumers
who perceived that EVs were not easy to use had stronger adoption intentions (Peters &
Dütschke, 2014), which may indicate another form of “costly” signaling in terms of
embracing a technological learning curve alongside additional monetary costs for trendsetting
technologies.
There is some indication that decisions to purchase an EV may rest on a temporally-sensitive
series of symbolic meanings. Specifically, we find that perceiving EVs as social innovator
symbols is a strong predictor of willingness to lease/buy, but a relatively weak or
insignificant predictor for impressions and willingness to pay. This suggests a shift in
particular aspects of self-identity that are most valuable at different stages of adoption
decision-making. Further research is needed to understand the process of EV adoption
decision making over time, as well as the role that experience with EVs plays at these various
stages.
3.5.3 Concern about climate change: mediation
Previous literature has found that concern about climate change directly predicts EV adoption
intentions (Carley et al., 2013; Egbue & Long, 2012; Krupa et al., 2014; Skippon &
Garwood, 2011), but this has not previously been examined alongside seeing EVs as social

8
Noppers et al. (2014) examined environmental attributes in terms of “low CO2 emissions” and
“environmentally friendly”.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
55

innovator or environmentalist symbols. Additionally, self-identity has been found to fully
mediate the relationship between valuing the environment and actions to alleviate
environmental harm (Ellen Van der Werff et al., 2013b), so we expected that social innovator
and environmental symbolism associated with EVs would mediate the relationship between
concern about climate change and EV adoption intentions (H3a and H3b). Our findings
confirmed this hypothesis. This extends the literature by uncovering a mediating effect of
self-identity symbolism associated with EVs, i.e., that seeing EVs to reflect environmentalist
and social innovator self-identities mediates the relationship between concern about climate
change and intent to purchase an EV.
On a practical note, our findings indicate that those concerned about climate change value the
potential of EVs to reaffirm their environmentalist and social innovator identities. This has
implications for EV marketing and uptake strategies. Educating people about the realities of
climate change, and explaining why it needs to be mitigated, may not be sufficient to
encourage people to take mitigating action if they do not feel that protecting the environment,
or being a social innovator, are aspects of their identity. A different approach, perhaps with
greater appeal to the ability of EVs to reflect environmentalist or social innovator identities,
may be more effective in encouraging adoption.
3.5.4 Effects of messaging
To our surprise, we did not find any effects of persuasive messaging on adoption intentions,
thus H4 was not supported. This may be because social norms of EV ownership were already
apparent among our sample of L.A. residents at the time of data collection, even if they did
not receive the social norms messaging in our study. California is at the forefront of EV
adoption across the US with EVs constituting 1.5% of new vehicle sales in 2014 when data
were collected (California New Car Dealers Association, 2014), vs. 0.7% across the US (IEA,
2016). In addition, L.A. is a recognized leader of progressive municipal energy policy
(Pincetl, Graham, Murphy, & Sivaraman, 2016). For EV policy in particular, in 2012 L.A. set
a target to have 80,000 EVs by 2015, and promoted electric buses and trucks alongside
initiatives to increase availability of charging stations, while additionally working with a
regional collaborative to promote EVs (IEA, 2012). Thus, L.A. residents are likely to have
seen EVs around town, which could plausibly convey positive social norms about them. This
could have muted the impact of the social norms messaging, either because the message did
not convey new norms, or because the financial group also subscribed to such norms.
Alternatively, one brief message may not have been sufficient to influence social norms. It is
also possible that the financial messaging had a positive impact on EV adoption intentions on
par with effects of social norms messaging, hence failing to produce a distinguishable
difference between the two messaging types. Further research should test marketing
approaches to encourage EV uptake to discern what strategies work best, and for which
consumers.
3.6 Conclusion
Previous studies have operationalized symbolic attributes broadly in terms of EVs as status
symbols, but have not examined the predictive power of EVs reflecting specific aspects of
self-identity. We advance this literature by refining the concept of symbolic attributes to
include specific reflections of self-identity, examining whether EV adoption intentions are
impacted by seeing EVs as symbols for environmentalist and social innovator self-identities.
We evaluate the relative importance of these symbolic attributes reflecting specific aspects of
Dissertation by Lee V. White
56

self-identity alongside a set of additional variables that prior studies have linked to EV
adoption, and find that perceiving EVs to reflect environmentalist self-identity was the
strongest and most consistent predictor of EV adoption intentions.
Concern about climate change has previously been found to influence EV adoption
intentions, but mechanisms for this concern translating into action have not previously been
examined. We find that seeing EVs as environmentalist and social innovator symbols
mediates the relationship between concern about climate change and EV adoption intentions.
That is, individuals who are concerned about climate change perceive stronger reflections of
environmentalist and innovator identities in EVs, and this in turn leads to stronger intent to
adopt EVs. This indicates that efforts intended to increase EV adoption should couch EVs as
reflections of these aspects of self-identity, not so much as tools to address climate change
per se.
3.7 Limitations and Future Directions
Our findings should be viewed in light of several limitations. First, our study relied on self-
report, which is not as strong an outcome as observed behavior. Future work using actual EV
purchase data is needed to address these limitations. Second, relative to L.A. County, our
sample was older, and included higher proportions of people with higher educational
attainment, incomes, and those who identified as Caucasians. Third, a large number of initial
respondents were dropped due to missing data. Drop-out analyses indicated that those
retained in the final sample were more favorable towards EVs than those who were dropped,
so our results may not capture the full range of variation in EV adoption likelihood. Hence,
results may not generalize to all populations or settings. As consumers who are highly
educated have been noted to be interested in EVs as early adopters (Carley et al., 2013),
results may overestimate adoption willingness of those in later adopter groups (according to
diffusion of technology). Additional work in non-early-adopter markets is needed to assess
the generalizability of our results.
Fourth, we used a limited number of questions to assess symbolism. A larger number of
questions designed to measure a greater spread of distinct symbolic reflections on self-
identity could reveal additional clusters of meaning that influence adoption, such as
technological or innovator symbolism separate from social and responsibility aspects, or use
of EVs to signal anti-war ideologies (Heffner et al., 2007; E. H. Noppers et al., 2014). In
addition, future work should also assess self-identity to provide further insights into how
concern about climate change, self-identity, and symbolism work together to impact EV
adoption intentions. Next, our assessment of instrumental attributes was limited in that it only
considered estimated range, costs, and charging convenience, whereas previous studies of
EVs have assessed additional instrumental attributes such as vehicle performance, comfort,
and driving experience (E. Noppers et al., 2015; Schuitema et al., 2013); future studies would
be well advised to consider all of these instrumental factors. Finally, we assessed estimated
EV range, rather than asking respondents to report subjective concerns regarding range.
Although we found that estimates of EV range, estimates of typical daily driving needs, and
the difference between these two variables did not impact EV adoption intentions, it may still
be the case that concerns regarding EV range could impact adoption intentions. Future work
should consider perceptions of EV range limitations as well as associated feelings of anxiety
or concern.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
57

4 Chapter 3: Controlling household electricity loads: The effect of actual
versus perceived savings on Time-of-Use rate acceptance (co-authored
with Nicole D. Sintov)

By Lee White and Nicole Sintov
Note: Chapter 3 initially included more predictors, but following feedback received at
conference the chapter was split in half. The initial analyses are outlined in Section 10:
Chapter 3 Supplementary Appendix.
4.1 Abstract
Demand-side response (DSR) measures are critical to integrating large shares of variable
renewable generation into electric grids. Time-of-use rates (TOU) are a relatively common
type of DSR that can help us understand what predicts household acceptance of new
electricity rate paradigms. We investigate drivers of TOU acceptance among 8945
participants who opted-in to a utility pilot program, and find perceived savings are the most
powerful driver of acceptance of TOU. Further, we find only weak relationships between
perceived savings and actual changes in bills and usage. Households may thus be at risk of
joining TOU programs based on perceived savings without achieving actual savings,
negatively impacting household returns. We find that perceived savings mediates the impacts
of rate understanding and ease of shifting use on TOU acceptance; policies to support
measures such as in-home displays alongside DSR programs could increase DSR acceptance
through mediation effects, and these measures additionally have the potential to better align
actual and perceived savings.
4.2 Introduction
Replacing fossil fuels with renewable generation capacity in the electricity sector is an
essential step to mitigate emissions contributing to climate change. Large shares of renewable
generation will require an increase in electric grid flexibility, due to the variable nature of
solar and wind (IEA, 2011). Demand-side response (DSR) measures aim to engage electricity
consumers to shift or reduce demand, to better match available supply of electricity from
variable sources. The residential sector accounts for 30% of electricity consumption across
OECD countries and 30-40% in the US (EIA, 2016; IEA, 2017), yet there is limited empirical
evidence on what motivates households to participate in DSR. Extensive studies of household
conservation behaviors provide insight into how households think about energy use (Attari et
al., 2010; Delmas, Fischlein, & Asensio, 2013), but do not help us understand what might
prompt households to engage in utility-sponsored DSR programs that explicitly link the
timing of household electricity use to utility generation profiles.
This paper examines household acceptance of static time-of-use (TOU) rates, a DSR measure
that aims to shift electricity use away from high-demand (i.e., “on-peak”) times by charging
higher rates for on-peak usage. We define TOU acceptance to encompass both satisfaction
with rate and intent to remain on TOU following a trial period. Static TOU rates use fixed
rate schedules that charge different prices for electricity use by season, day of week, and hour
of day to discourage on-peak use. Though static TOU rates have been in use for decades and
are well established to shift the time of household electricity demand in response to price
signals (Faruqui & George, 2005; Faruqui & Malko, 1983; Faruqui & Sergici, 2010), there is
limited understanding of what motivates households to participate in these programs. Thus,
Dissertation by Lee V. White
58

further work examining TOU is needed to produce insights into how future DSR measures
should be designed to maximize household acceptance. Such work is critical for designing
DSR policies and programs that households will adopt and continue with, which is in turn
critical to integrating large shares of variable renewable generation capacity into electricity
grids.
4.2.1 Perceptions affecting TOU acceptance
Household perceptions of energy costs are expected to be a powerful predictor of acceptance
of price-based DSR instruments such as TOU (Mostafa Baladi et al., 1998; Nicolson et al.,
2017; K. Train & Mehrez, 1994). Expectations of monetary saving predict enrollment in
TOU pilots (Mostafa Baladi et al., 1998; K. Train & Mehrez, 1994), and fear of increased
bills can greatly dampen willingness to switch to TOU (Nicolson et al., 2017). However,
households often have limited understanding of their energy use (Attari et al., 2010; Brounen
et al., 2013), and individuals in TOU pilots have perceived their baseline usage patterns to
allow savings even when they do not (Mostafa Baladi et al., 1998). This raises the key
question of the extent to which perceptions of savings predict TOU acceptance compared to
actual, measurable changes in bills and usage. Household perceptions of electricity and water
use have repeatedly been found to be inaccurate (Attari, 2014; Attari et al., 2010; Iwata,
Katayama, & Arimura, 2015). Thus, there is a need to understand whether household
acceptance of TOU (and other DSR measures) is driven more strongly by perceived savings
or by actual decreases in electricity bills and usage.
When evaluating electricity costs, individuals may be more likely to recall perceived savings
rather than actual savings; perceived savings are the attempted recall of actual savings, and in
the case of energy use and costs, these perceptions are affected by cognitive processes,
including cognitive accessibility of electricity usage (Schley & DeKay, 2015). Hence,
perceived savings may have a greater effect on TOU acceptance than actual bill decreases. In
addition, greater ability to adjust actual usage away from TOU on-peak times, i.e., more
flexible demand, has been found to predict TOU acceptance (Caves, Herriges, & Kuester,
1989; Ericson, 2011; K. Train & Mehrez, 1994). However, not all households are willing or
able to change their electricity use habits, even if they perceive that they can (Hargreaves,
Nye, & Burgess, 2013). Thus, perceived ease of shifting electricity use may also predict TOU
acceptance. Further, households on TOU are better positioned to reduce bills if they
understand when on-peak rates occur. Perceived and actual understanding of TOU rates are
thus expected to predict TOU acceptance.
4.3 Methods
We use data from households that participated in a static TOU pilot program administered by
a large power utility in the southwestern US.
4.3.1 Procedures
4.3.1.1 TOU Pilot
Households were shifted to TOU rates in June or July of 2016, and remained on these rates
until June or July 2017. TOU rate 1 and 2 on-peak times covered different hours depending
on weekend vs. weekday (see Table 4.1). On-peak hours were in the evening, as the pilot
region has a high penetration of solar generation leading to increased need for ramping
generation in the evening as the sun sets. During on-peak hours, a higher price was charged
per kWh compared to off-peak or super off-peak hours. The exact cost per kWh varied
depending on rate and season, but costs were the same for weekdays/weekends for each peak-
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
59

bracket and season. Rate 1 had six evening hours on-peak for summer and winter weekdays,
and no on-peak times on weekends. Rate 2 had three evening hours on-peak for summer and
winter weekdays, and no on-peak times on weekends.
4.3.1.2 Survey
A survey gauging customer experiences was administered between October and December
2016. Customers thus had 3-6 months’ experience with TOU rates before the survey was
administered, which occurred solely or primarily during summer months. Survey data were
collected using a mixed-method approach including email, mail, and phone to increase
response rates. Survey responses rates were 82% overall.
4.3.2 Measures
The following sub-sections detail the measures constructed for use in analysis. Raw data was
gathered from household smart meter records, household bills, and responses to the survey.
4.3.2.1 Electricity usage data
The utility provided hourly electricity usage data for each household for the months July-
December during the baseline years 2014 and 2015, and for the pilot year 2016. Only the
summer months (July, August, and September) were included in analysis, because 1) all
households had experience with TOU during summer months and 2) summer bills tend to be
higher in this region where there is heavy air-conditioner use and heating often utilizes gas
rather than electricity (EIA, 2017a, 2017b). Summer hourly consumption data are used to
form two variables: 1) on-peak use at baseline (kWh, daily average); and 2) decrease in on-
peak use (kWh, daily average).
On-peak use at baseline (Usage control variable): Mean daily on-peak usage is calculated
separately for each year of data during summer (July, August, and September). The mean
daily on-peak usage variable takes into account the different lengths of on-peak time, i.e.,
Rate 1 on-peak use is coded with 6 hours on-peak and Rate 2 on-peak use is coded with 3
hours on-peak. For those in the control group, “on-peak” times were assigned artificially.
Control group households were randomly assigned to either Rate 1 or Rate 2 rate structure,
and on-peak usage was defined as usage happening during the hours designated as on-peak
by that rate.
Decrease in on-peak use: Difference in on-peak use between baseline and TOU pilot averages
is taken as baseline daily on-peak average use minus TOU pilot daily on-peak average use.
The following equation describes the decrease in on-peak usage calculation: ΔU = ((U 2014 +
U2015)/2 - U2016), where U represents mean daily on-peak usage in kWh in each year as
denoted by subscript, and Δ represents ‘change in’.
4.3.2.2 Electricity bills
Spending on electricity was measured using actual billing data for each household, for years
2014-2016 as provided by the utility. As with usage data, only the summer months were
included in analysis. All households included in the analysis had billing data for at least two
of the three months in summer, as households with further missing billing data also had
missing usage data.
Decrease in bills: The bill amounts used for analysis were the actual bills that customers
received. That is, they reflect the true amount that customers paid each month. Customers
participating in the TOU pilot received financial incentives to participate in the form of bill
Dissertation by Lee V. White
60

credits, which are reflected in bill totals and could not be parsed out. However, this measure
is suitable for comparing household perceptions of costs to actual bills, given that bill
amounts used in analysis are those actually paid by households. A mean summer bill variable
was constructed by taking the mean of summer bills for households that had bills for at least
two of the three months of summer (July, August, and September). The decrease in bills is
described by the following equation: ΔB = ( (B2014 + B2015)/2 - B2016), where B represents
mean bill amounts in $ each year as denoted by subscript, and Δ represents ‘change in’.

4.3.2.3 Survey measures
Dependent variables: Two dependent variables are used to measure TOU acceptance: (1) rate
satisfaction and (2) intent to remain on TOU. Each variable was assessed using a single item
with responses provided on 11-point Likert scales as detailed in Table 4.1.
Independent variables: Table 4.1 further describes the questions used to measure the primary
independent variables from the survey: perceived savings, perceived ability to reduce evening
use, perceived understanding of rate, and actual understanding of rate.
• Perceived savings was assessed with a scale formed by taking the mean of three
questions (α=0.91) as detailed in Table 4.1.
• Perceived ability to reduce evening use was measured with a single question,
described in Table 4.1.
• Perceived understanding of rate was assessed with a scale formed by taking the mean
of four questions (α=0.90) as detailed in Table 4.1. Questions regarding understanding
of rate were only asked of participants on TOU rates, so no responses from the control
group are available for these.
• Actual understanding of rate was measured using a survey question that asked
respondents to mark all the hours during which their electricity was most expensive
on summer weekdays, i.e., which hours they recalled being on-peak. A new variable
was created to represent percentage of hours correctly recalled, such that if a
household had on-peak hours 2-8pm and only marked on-peak hours as 5-8pm they
received a score of 50%.
Demographic control variables: A set of three demographic variables expected to influence
TOU acceptance was also included in the models. Homeownership was included as those
who own homes rather than rent have more control over the building envelope and
appliances, and thus greater control over their energy use (Buchanan, Banks, Preston, &
Russo, 2016). Homeownership was coded as a dichotomous variable with 1 indicating
homeownership and 0 indicating otherwise. Income was included as those with higher
incomes may be less sensitive to energy prices (Brounen et al., 2013; P. Reiss & White,
2003). Enrollment in electric utility financial aid programs (that give households discounts on
electricity bills if income falls below certain limits based on the number of household
members) serves as an indicator of low income, with those enrolled in a financial aid program
coded 1 and others coded 0. Finally, education was included to control for potential impacts
associated with greater energy literacy or rate understanding. Education was coded as
dichotomous variable split at the sample median, with 1 indicating that respondents had a
two-year college degree or more education (four-year college, graduate school, or
professional degrees) and 0 indicating that respondents had less formal education.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
61

Table 4.1: Detailed description of survey measures and variables
Measure Survey questions Response
Dependent Variables
Intent to remain on TOU I want to stay on my new rate plan after this
study ends
11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
Satisfaction with rate plan How satisfied are you overall with your
current electricity rate plan?
11-point Likert scale,
0-10 with 0 indicating
“not satisfied at all”
and 10 indicating
“completely
satisfied”
Independent Variables
Perceived ease of reducing
evening use
Since the beginning of this summer, how
easy was it for you or members of your
household to take action to reduce your
electricity usage in the afternoon and
evenings?
11-point Likert scale,
0-10 with 0 indicating
“not at all easy” and
10 indicating
“extremely easy”.
Perceived savings

Note: Only calculated for
households with 2 of the 3
questions answered,
otherwise households were
dropped

The rate provides me with opportunities to
save money
11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
The rate works with my household’s
schedule
11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
The rate is affordable 11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
Perceived understanding of
rate

Note: Only calculated for
households with 3 of the 4
questions answered,
otherwise households were
dropped

My electricity bill helps me understand the
time of day I’m spending the most money on
electricity
11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
The peak and off-peak time periods are easy
to remember
11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
The rate is easy to understand 11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
and 10 indicating
“completely agree”
My electricity bill is easy to understand 11-point Likert scale,
0-10 with 0 indicating
“do not agree at all”
Dissertation by Lee V. White
62

and 10 indicating
“completely agree”
Actual understanding of
rate
To the best of your knowledge, during what
times on weekdays in the summer months is
electricity the most expensive to use based
on your current study rate plan? Mark all the
hours when electricity is most expensive to
use.
Free response
checkboxes for all
hours in a 24-hour
period

A factor analysis was run to determine the item structure for the three and four items
belonging to the perceived savings and perceived understanding scales, respectively. An
iterated principal factor analysis was run including all seven items. Cattell’s Scree test
(Figure 4.1) was used to determine the number of factors to retain, as this can be a more
accurate method of determining number of factors present than relying on Eigenvalues
(Fabrigar & Wegener, 2011; Osborne & Costello, 2009). A total of two factors were retained;
promax rotation was then applied, and results are presented in Table 4.2. Findings support
two factors onto which the items load as expected from a theoretical standpoint.
Figure 4.1: Scree plot of eigenvalues after iterated principal factor analysis

Table 4.2: Factor loadings for perceived savings and perceived understanding following
promax rotation
Question Factor 1: Perceived
understanding
Factor 2:
Perceived savings
Uniqueness
The rate provides me with opportunities to
save money
0.23 0.67 0.29
The rate works with my household’s
schedule
0.02 0.86 0.24
The rate is affordable 0.05 0.90 0.13
My electricity bill helps me understand the
time of day I’m spending the most money
on electricity
0.85 0.01 0.27
The peak and off-peak time periods are
easy to remember
0.70 0.11 0.39
The rate is easy to understand 0.74 0.10 0.34
My electricity bill is easy to understand 0.80 0.10 0.23
Eigenvalues 4.53 0.56
0 1 2 3 4 5
Eigenvalues
0 2 4 6 8
Number
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
63

4.3.3 Participants
Data are from households that participated in a pilot program administered by a large power
utility in the southwestern US. The utility sent invitations by direct mail and email soliciting
opt-in to the 2016 TOU pilot to roughly 197,000 households, and 14% of these accepted.
Some households that accepted the offer were not enrolled because they were ineligible for
various reasons (e.g., already participating in a special rate program, or having solar
photovoltaic panels installed at their home). Households were offered a financial incentive
(credits to their electricity bill) for participation. Low-income households and those with
elderly members were deliberately oversampled. The utility assigned 21534 households to
either TOU rate 1 (N=4709), rate 2 (N=6365), rate 3 (N=3746), or the control group that
opted-in to TOU but was not placed on a TOU rate (N=6714). Rate 2 was assigned more
participants due to additional utility analyses being conducted on data collected from this
group. Due to roll-out issues and additional complexities specific to rate 3, only rates 1 and 2
and the control group are examined here. Rate 3 was considered more complex because it
contained five TOU rate brackets (super off-peak, off-peak, mid-peak, on-peak, and super on-
peak) compared to the three rate brackets in Rates 1 and 2 (super off-peak, off-peak, and on-
peak). Additionally, roll-out of Rate 3 was delayed such that not all households had begun
receiving billing by July 2016, whereas all Rate 1 and 2 households were fully switched to
TOU rates and billing by July.
4.3.3.1 Dropped participants
By December 2016, 2787 households had dropped out of the pilot due to relocating,
ineligibility, or choosing to drop out, leaving 18747 households enrolled. Of these, a total of
16181 households responded to the survey. Before receipt by the author, the utility removed
respondents who answered 5.4% or less of the survey items. Respondents were also removed
if they provided the same rating (e.g., 10) for all items across the three multi-item measures
(over 4 questions) in the survey. Additionally, respondents were removed if they selected all
items in a ‘select-all-that-apply’ question in which some categories were mutually exclusive,
e.g., if when asked “what kept you from shifting usage in the evening” respondents selected
both “nothing keeps me from shifting my usage” and “my schedule doesn’t allow me to
reduce my usage”.
This yielded a sample of N=16073 households, with 5198 in the control group, 3522 on Rate
1, and 4593 on Rate 2; the 2760 households on Rate 3 were not used, leaving an initial
sample of N=13313 households. Additional households were then dropped due to missing
billing data, usage data, or survey responses. Specifically, households were dropped for the
following reasons:
Usage data: First, households that were missing electricity usage data on any days in July,
August, and September in 2014, 2015, or 2016 were dropped from analysis. Across all years,
a total of N=1339 households were dropped from the dataset due to incomplete usage data;
the missing values were predominantly clustered across several days or weeks at the
beginning of the recorded period, indicating that either no account was established for that
address (i.e., residents moved in during study period) or the house did not have a smart meter
at the beginning of the time period.
Dissertation by Lee V. White
64

Billing data: Second, customers with a summer bill decrease or increase of over $500
between baseline and pilot year were also dropped, as $500 represents an unusually large
monthly bill change for a standard household (N=7 households).
Survey data: Finally, households were dropped if they did not provide an answer to either
dependent variable, answered fewer than 2 of the 3 perceived savings questions, fewer than 3
of the 4 perceived understanding questions, or if they were missing data on any of the key
questions used to measure perceived ease of reducing evening use, actual rate understanding,
education, income, or home ownership. This resulted in dropping 3080 households
(excluding the questions related to understanding which were only asked of TOU rate
participants, not controls). The remaining sample was N=9220 households.
Customer engagement with pilot: The survey included a question asking participants if they
recalled participating in the rate study. Of the N=9220 households with complete data, N=275
households enrolled in TOU rates indicated that they did not recall participating in the study
or did not answer the question regarding memory of participation. These households were
dropped from further analysis, bringing the main sample size for all groups (control and TOU
rate households) to N=8945.
TOU participant sub-sample: Among the main sample of N=8945, certain survey questions
were only asked of households assigned to TOU rates (N=5358). Specifically, the dependent
variable of intent to remain on TOU and the independent variables relating to understanding
of TOU rate were not applicable to the control group. Thus, a sub-sample was formed for
TOU households. This TOU sub-sample contains 5249 households, as 109 households had
incomplete data for survey questions regarding perceived or actual understanding of rates.
Samples used in analyses: The final analyses thus utilize two sample groups: a main sample
used for Model 1 that includes the control group, and a sub-sample used for Models 2 and 3
that only includes those who were assigned to TOU rates (summary statistics in Table 4.3).
The main sample (N=8945) contains 3587 respondents in the control group, 2325 assigned to
TOU rate 1, and 3033 assigned to TOU rate 2. The sub-sample of only TOU households
(N=5249) contains 2,279 on TOU rate 1 and 2,970 on TOU rate 2. Mediation analyses used
only the sub-sample (N=5249), due to the focus of these analyses on rate understanding
questions that were only asked of TOU participants.
4.3.3.2 Dropout analyses
Wilcoxon rank-sum tests were used as a non-parametric alternative to t-tests to compare
those retained in the final sample to those dropped due to missing data across two measures:
(1) the dependent variable of rate satisfaction (main sample, N=8945), (2) the dependent
variable of intent to remain on TOU (sub-sample, N=5249). Point-biserial effect size
(denoted by r) was generated and is reported alongside Wilcoxon test statistics. For the main
sample (N=8945), households with complete data across the survey and usage/billing data
had significantly lower rate satisfaction than those with missing data (N=8945, M=6.73,
Mdn=7, SD=2.55 for complete and N=3913, M=6.89, Mdn=7, SD=2.66 for incomplete) (z=-
4.29, p=0.00, r=-0.03). Only respondents who were assigned to a TOU rate responded to the
intent to remain on TOU rate question. Among this sub-sample, those with complete usage,
survey, and billing data (N=5249, M=6.83, Mdn=8, SD=3.00) had slightly lower scores than
non-completers (N=2629, M=6.98, Mdn=8, SD=3.10) on intent to remain on TOU, a
statistically significant difference (z=-3.12, p=0.00, r=-0.02). Caution is therefore advised in
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
65

interpreting results, as those who remain in the samples used for analyses are less favorable to
TOU than those dropped.
Table 4.3: Summary statistics
Sub-sample
(N=5249)

Mean (Std.
Dev)
Main sample
(N=8945)

Mean (Std. Dev.)

Response Range
1

I'm satisfied with my current rate plan 6.60 (2.58) 6.73 (2.55) 0 to 10
I'd like to stay on my new rate 6.83 (3.01) 6.96 (2.91) 0 to 10
Perceived savings 6.59 (2.63) 6.66 (2.62) 0 to 10
Perceived ease of reducing evening use 6.38 (2.52) 6.35 (2.55) 0 to 10
Perceived understanding of rate 7.23 (2.28) NA 0 to 10
Decrease in bills ($) -22.69 (52.62) -15.14 (51.49) -463.97 to 414.55
Decrease in on-peak use (kWh, daily
av.)
0.22 (1.80) 0.13 (1.90) -18.84 to 15.55
Actual understanding of peak times 0.55 (0.40) NA 0 to 1
On-peak use at baseline (kWh, daily
av.)
6.32 (5.13) 6.62 (5.51) 0.00 to 59.75
Education (% high education) 48 48 0 or 1
Home owned (% own home) 74 74 0 or 1
Low-income customer (% low-income) 40 41 0 or 1
1
Minimum and Maximum values are given for the main sample (N=8945), except for values relating to
understanding which were only measured in the subsample and hence are provided for the subsample (N=5249)
4.3.4 Analyses
While billing and usage data followed normal distribution patterns, data gathered through the
survey did not. Thus, when selecting methods for factor analysis, correlation analysis, and
dropout analysis, non-parametric methods (i.e., those that do not make assumptions about the
probability distributions of the data) were used.
4.3.4.1 Regression
Ordinary Least Squares (OLS) multiple linear regression was run using the statistical analysis
software STATA to test the core hypothesis that perceived savings was the most powerful
predictor of TOU acceptance, above and beyond decreases in bills and on-peak usage (see
Figure 4.2 for additional predictors and expected directions). Multiple linear regression
allowed for simultaneous examination of comparative effect sizes of additional predictors
expected to impact TOU acceptance, while also controlling for the effect of these additional
predictors relative to perceived savings. Post-estimation tests using both visual inspection of
residual versus fitted values and Cameron and Trivedi’s omnibus test indicated
heteroscedasticity of error terms, so Huber-White standard errors were applied to correct for
this. Non-normality was also revealed by post-estimation tests; while this is not expected to
affect the accuracy of co-efficient estimates, it indicates that tests of statistical significance
should be interpreted with caution.
4.3.4.2 Mediation
To test whether the policy-targetable predictors of perceived understanding, actual
understanding, and perceived ease of reducing evening use were having an indirect impact on
intent to remain on TOU, mediation analysis was performed with perceived savings as the
mediator. Mediation analysis was performed using causal chain path diagrams (Baron &
Kenny, 1986; Tofighi & MacKinnon, 2011). Given the non-normality and heteroskedasticity
Dissertation by Lee V. White
66

present in the data, a robust median estimation method was used rather than OLS (Yuan &
MacKinnon, 2014). Given data non-normality and heteroskedasticity, bootstrap methods such
as those used by Preacher and Hayes were considered to not be appropriate (Yuan &
MacKinnon, 2014). A single-mediator model was used, defined by three equations,
Equation 4.1:
Equation 4.2:
Equation 4.3:
where y i represents the dependent variable, x i represents the independent variable, m i
represents the mediator, ß01- ß03 represent intercept terms, and e1i – e3i represent error terms.
In Equation 4.1, which estimates the impact of the independent variable on the dependent
variable, τ represents the direct effect of the independent variable in absence of the mediator.
In Equation 4.2, α represents the impact of the independent variable on the mediator.
Equation 4.3 estimates the simultaneous effects of the independent variable and mediator on
the dependent variable (Baron & Kenny, 1986) where τ’ represents the effect of the
independent variable on the dependent variable holding the effect of the mediator constant,
and β represents the effect of the mediator. The coefficient of the indirect effect of the
mediator (αβ) was calculated from α and β in Equations 4.2 and 4.3 above using the product
of coefficients method (MacKinnon et al., 2002). Additionally, the significance of the indirect
effect was tested using the distribution of the product of coefficients method (Tofighi &
MacKinnon, 2011). Confidence intervals for indirect effects were computed based on α and β
and their standard errors (Tofighi & MacKinnon, 2011).
4.4 Findings
Our core hypothesis is that household perceptions of savings will be the most powerful driver
of TOU acceptance (H1a), over and above actual changes in bills (H1b) and usage (H1c).
Figure 4.2 details these and additional hypotheses regarding TOU acceptance, specifically
that perceived ease of reducing evening use is expected to have a positive impact on TOU
acceptance (H2a), as is perceived understanding of rate (H2b) and actual understanding of
rate (H2c). We further hypothesize that perceived savings will mediate the relationships
between each of (1) perceived ease of shifting on-peak use (H3a), (2) perceived
understanding of rate (H3b), and (3) actual understanding of rate (H3c) and TOU acceptance.
Figure 4.2: Hypothesized predictors of TOU acceptance
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
67

*Data available for the subset of respondents on TOU rates (N=5249), as detailed in the findings and methods section
4.4.1 Correlations of perceived savings with actual changes in usage and bills
To confirm that discrepancies in regression coefficients of perceived savings vs. actual bill
decreases are not attributable to multicollinearity, we first examine correlations between our
core variables for H1a-H1c. Based on prior research finding large discrepancies between
perceived and actual energy use, specifically in terms of usage by individual appliances
(Attari et al., 2010), we expect only a weak correlation between overall perceived savings and
both changes in bills and changes in on-peak usage. We use Spearman’s rank correlation
(N=8945, including households assigned to the control, TOU rate 1, and TOU rate 2) to test
the correlations of perceived savings with both decrease in bills and decrease in on-peak use.
On-peak use, as opposed to overall use, was examined due to its relevance to TOU rates
specifically; it would be possible to achieve lower bills by shifting use to off-peak times even
if overall use did not decrease. Perceived savings is correlated positively with decreases in
bills at the 5% level (ρ=.17 p=.000), and positively with decreases in on-peak usage (ρ=.05
p=.000). Though statistically significant, these are small correlations. Figure 4.3 shows these
relationships visually.
Figure 4.3: Visual representation of correlations between perceived savings and (a) decrease
in bills ($); (b) decrease in on-peak use (kWh, daily average)
Dissertation by Lee V. White
68

a b
4.4.2 Modeling Acceptance of TOU
We test our core hypothesis, that perceptions of savings will be the most powerful driver of
TOU acceptance (H1a) – even compared to actual decreases in bills (H1b) and usage (H1c)—
using three ordinary least squares (OLS) multiple regression models with Huber-White
standard errors. We additionally test hypotheses H2a, H2b, and H2c as presented in Figure
4.3. Table 4.4 reports results of these analyses with standardized coefficients to enable
comparison of coefficients within models; unstandardized coefficients are presented in
Section 9.1: Chapter 3, Appendix 1 for comparability between models. In Model 1, the
dependent variable is satisfaction with current rate, and the sample used for analysis (main
sample N=8945) includes the control group, i.e., those who were not assigned to any TOU
rate but remained on their pre-existing rate. Models 2 and 3 examine the dependent variable
of intent to remain on TOU following the pilot, a question asked only of pilot participants,
and hence only include households assigned to a TOU rate (sub-sample N=5249).
Independent variables included in all models are: perceived savings, perceived ease of
reducing evening use, actual decrease in bills, actual decrease in on-peak use, and a set of
control variables (i.e., on peak use at baseline, homeownership, education level, low income
status). Model 1 also includes a rate type variable indicating whether respondents were
control, TOU rate 1, or TOU rate 2 participants, and Models 2 and 3 include a rate type
variable indicating whether participants were assigned to TOU rate 1 or rate 2. Model 3 adds
variables for perceived and actual understanding – questions that were only asked of TOU
pilot participants.
We test variance inflation factors (VIF) to check all models for possible multicollinearity, and
find that VIF values remain well under 10 for all variables, indicating no multicollinearity
concerns (Neter, Wasserman, & Kutner, 1989). All three models have high explanatory
power for their respective dependent variables (R
2
=0.57-0.58, see Table 4.4). Effect sizes
(partial η
2
, denoted by

) to evaluate the proportion of model variability explained by each
predictor are calculated using standard variance estimators for standard errors, and are
calculated after fitting linear models. A small effect size is between 0.01 and 0.03, a medium
between 0.04 and 0.26, and a large >0.26 (Miles & Shevlin, 2001).
The following sub-sections discuss TOU acceptance predictor results pertaining to: (1)
testing H1a- H1c, perceived savings vs. actual decreases in bills and on-peak use; (2) testing
H2a, perceived ease of reducing evening use; (3) testing H2b and H2c, perceived vs. actual
understanding of rates; and (4) control variables.
-400 -200 0 200 400
0 2 4 6 8 10
Perceived savings (Likert rating)
Decrease in bills ($) Fitted values
-20 -10 0 10 20
0 2 4 6 8 10
Perceived savings (Likert rating)
Decrease in on-peak use (kWh, daily average) Fitted values
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
69

4.4.2.1 Perceived savings vs. actual decreases in bills and on-peak use
We hypothesize that perceived savings will be the strongest predictor of TOU acceptance
(H1a) compared to decrease in bills (H1b) and on-peak use (H2c); all of these hypotheses are
supported by the findings. Standardized coefficients in Table 4.4 show large effects of
perceived savings on both rate satisfaction and intent to remain on TOU (Model 1: β=0.70,
p= 0.000; Model 2: β=0.74, p=0.000; Model 3: β=0.69, p=0.000); these are larger than those
of all other variables in the model, including decrease in bills (Model 1: β=0.03, p= 0.03;
Model 2: β=0.04, p=0.01; Model 3: β=0.04, p=0.003) and on-peak usage (Model 1: β=-0.00,
p= 0.98; Model 2: β=-0.02, p=0.07; Model 3: β=-0.02, p=0.05). The effect size of perceived
savings on TOU acceptance is consistently larger than that of bill decreases and usage
decreases, which do not explain any notable portion of the variance within the models (Model
1: perceived savings

=0.461, change in usage

=0.000, change in bills

=0.001; Model
2: perceived savings

=0.477, change in usage

=0.001, change in bills

=0.002; Model
3: perceived savings

=0.317, change in use

=0.001, change in bill

=0.002). The
effect size of perceived savings is also consistently far larger than that of any other predictors
in the model, all of which have only small or negligible effect sizes.
4.4.2.2 Perceived ease of reducing use
As shown in Figure 4.2, we hypothesize that higher scores on perceived ease of reducing
evening use predict TOU acceptance (H2a). H2a is supported by regression coefficients,
though effect sizes are very small. Perceived ease of reducing evening use has a positive and
significant effect on both rate satisfaction (Model 1: β=0.05, p=0.000) and intent to remain on
TOU (Model 2: β=0.04, p=0.002; Model 3: β=0.04, p=0.002). Perceived ease of reducing
evening use has a negligible impact in terms of effect size (Model 1:

=0.005; Model 2:

=0.002; Model 3:

=0.004).
4.4.2.3 Perceived vs. actual understanding
Additionally, in Model 3, we test hypotheses H2b and H2c that both perceived and actual
understanding are positive predictors of intent to remain on TOU. The effect sizes of both
perceived and actual understanding are small to negligible (Model 3: perceived understanding

=0.007, actual understanding

=0.005). However, regression coefficients reveal a more
interesting story – while both perceived and actual understanding are significant predictors of
intent to remain on TOU, perceived understanding has a positive effect on intent (Model 3:
β=0.08, p=0.000) while actual understanding has a negative effect (Model 3: β=-0.05,
p=0.000). A supplementary analysis shows that there is no correlation between perceived and
actual understanding of rates (Spearman’s ρ=0.02, p=0.28).
Dissertation by Lee V. White
70

Table 4.4: Models of TOU acceptance
(1) (2) (3)
Satisfaction with
current rate plan
Intent to remain
on TOU
Intent to remain on TOU
(evaluating understanding)
Perceived savings 0.70***
(0.01)
0.74***
(0.01)
0.69***
(0.02)
Perceived ease of reducing
evening use
0.05***
(0.01)
0.04**
(0.01)
0.03**
(0.01)
Decrease in bills ($) 0.03*
(0.00)
0.04**
(0.00)
0.04**
(0.00)
Perceived understanding
of rate

0.08***
(0.02)
Decrease in on-peak use
(kWh, daily average)
-0.00
(0.01)
-0.02
(0.02)
-0.02
(0.02)
Actual understanding of
peak times

-0.05***
(0.07)
Control rate plan 0.00
(.)

TOU plan 1 -0.02*
(0.04)
0.00
(.)
0.00
(.)
TOU plan 2 -0.05***
(0.05)
0.02
(0.06)
0.02
(0.06)
On-peak use at baseline
(kWh, daily average)
-0.10***
(0.00)
0.05***
(0.01)
0.05***
(0.01)
Education 0.03***
(0.04)
0.00
(0.06)
0.01
(0.06)
Home owned 0.01
(0.04)
0.01
(0.06)
0.01
(0.06)
Low-income customer 0.06***
(0.04)
0.02
(0.06)
0.01
(0.06)
R
2
0.58 0.57 0.58
N 8945 5249 5249
Standardized beta coefficients; Huber-White robust standard errors in parentheses
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

4.4.2.4 Rate type: Control vs. TOU
Households on the control rate plan report significantly higher rate satisfaction than those
enrolled in either TOU rate 1 or 2 (Model 1). Given that TOU rate 1 had twice as long an on-
peak period as TOU rate 2, TOU rate 1 could have been associated with lower intent to
remain on TOU. In Model 1, rate satisfaction is slightly lower for those on TOU rate 2
compared to TOU rate 1 with the control as baseline. However, there is no difference in
intent to remain on TOU based on TOU group (Models 2 and 3).
4.4.2.5 Control variables
Models 1-3 include a number of control variables based on prior literature. Higher on-peak
use at baseline negatively predicts rate satisfaction (Model 1), but positively impacts intent to
remain on TOU (Models 2 and 3). Of the demographic variables included, education and
income are positive and significant predictors of rate satisfaction (Model 1) while
homeownership is not, and no demographics are significant predictors of intent to remain on
TOU (Models 2 and 3). Effect sizes of demographic variables are negligible (Model 1:
education

=0.001, home ownership

=0.000, income

=0.006; Model 2: education
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
71

=0.000, home ownership

=0.000, income

=0.001; Model 3: education

=0.000,
home ownership

=0.000, income

=0.000).
4.4.3 Mediation Analysis
The models in Table 4.4 tell us that perceived savings is by far the strongest predictor of
TOU acceptance, including intent to remain on TOU, but offer limited insight into what kinds
of policy interventions might strengthen intent to adopt TOU. For this second step, we
consider whether perceived savings may act as a mediator for factors that could be targeted
by policies to strengthen intent to remain on TOU. The primary value that perceived
understanding of rates, actual understanding of rates, and perceived ease of shifting evening
use will have to a household is monetary (i.e., perceived or actual savings). Thus, the impact
on TOU acceptance of higher perceived ease of reducing evening use, perceived
understanding of rates, and/or actual understanding of rates, is expected to be mediated by
perceived savings (H3a, H3b, and H3c, respectively).

If mediation occurs, then policy mechanisms requiring provision of technologies or measures
such as in-home displays (IHDs), utility-automated thermostat adjustments, or simplified rate
plans may be recommended as a way to strengthen TOU acceptance. IHDs may impact rate
understanding by increasing the visibility and cognitive accessibility of energy use, and
experience with IHDs facilitates consumer learning that may improve household decision-
making during on-peak events (Jessoe & Rapson, 2014). Utility actions such as automating
thermostat adjustments during on-peak times or introducing simplified rate plans may
strengthen perceived ease of reducing evening use, and/or rate understanding, by reducing the
need for households to interact with energy-using devices or remember complex rate
structures. People view automated on-peak heating system adjustments favorably compared
to static TOU (Fell, Shipworth, Huebner, & Elwell, 2015).

We use mediation analysis to examine the indirect effects of perceived ease of reducing
evening use, perceived understanding of TOU rate, and actual understanding of TOU rate on
intent to remain on TOU (Figure 4.4). Prior studies using mediation analyses have
inconsistently reported on whether data met statistical assumptions of normality. For the
present study, we use Cameron and Trivedi’s (2009) omnibus test to determine whether our
data meet normality and heteroscedasticity assumptions. Results indicate that skewness and
heteroskedasticity cannot be ruled out for any of the three mediation models (p < 0.000). The
bootstrapping method (Hayes, 2013) is therefore not suitable, as it is not equipped to handle
these issues (Yuan & MacKinnon, 2014). Therefore, a robust mediation method using median
regression is used as has been done in prior work (Sintov, Geislar, & White, 2017; Yuan &
MacKinnon, 2014).
Three mediation models are examined, with perceived savings the mediator in all models and
intent to remain on TOU the dependent variable in all models; the independent variables
examined are (1) perceived ease of reducing evening use, (2) perceived understanding of
TOU rate, and (3) actual understanding of TOU rate. The paths in Figure 4.4 show the
indirect effect (path through α and β) of each independent variable on the dependent variable
(intent to remain on TOU) through the mediator (perceived savings), the direct effect (τ) of
each independent variable on the dependent variable, and the reduced effect of each
independent variable on the dependent variable when the mediator is added to the model (τ’).
Intent to remain on TOU is used here as the dependent variable rather than rate satisfaction,
as intent to remain on TOU is expected to be a slightly stronger representation of TOU
Dissertation by Lee V. White
72

acceptance; mediation models were also run using rate satisfaction as a dependent variable
and the pattern of results was comparable (Section 9.2: Chapter 3, Appendix 2).
As shown in Figure 4.4, all three of the independent variables are completely mediated by
perceived savings, with indirect effects that are significant at the 99% level and direct effects
of the independent variables dropping to zero when the mediator is included in models. Thus,
it is possible that policy mechanisms requiring provision of measures such as IHDs and
simplified rate plans may indirectly impact intent to remain on TOU, through the mediating
variable of perceived savings. However, unlike perceived ease of shifting evening use and
perceived rate understanding, actual understanding of TOU rates has a negative effect on
perceived savings, and hence a negative indirect effect on intent to remain on TOU. This
indicates that further avenues need to be explored to understand why correctly recalling on-
peak hours is associated with lower perceived savings, so that household concerns can be
addressed in future pilots and studies.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
73

Figure 4.4a: Mediation of perceived ease of reducing evening use by perceived savings using
median regression. Indirect effect = 0.667 (SE=0.017), and for 99% confidence interval
LL=0.623, UL=0.711

Figure 4.4b: Mediation of perceived understanding by perceived savings using median
regression. Indirect effect = 0.97 (SE=0.015), and for 99% confidence interval LL=0.933,
UL=1.007

Figure 4.4c: Mediation of actual understanding by perceived savings using median
regression. Indirect effect = -0.667 (SE=0.116), and for 99% confidence interval LL=-0.965,
UL=-0.369

4.5 Discussion
As hypothesized, perceived savings has a far greater impact on TOU acceptance than actual
decreases in bills and on-peak usage during the pilot period. Perceived savings also has only a
weak association with actual reductions in bills and electricity use. There are several factors
that may feed into this greater importance of perceived savings, and its lack of relation to
actual bill decreases. People within households generally have poor energy bill literacy and
Dissertation by Lee V. White
74

are unlikely to remember their electricity bills from prior or even current years (Brounen et
al., 2013), thus would have needed to consult archival data to determine true spending
changes. Further, household residents generally have weak understanding of energy use at
given times during the day, though provision of real-time energy feedback (e.g., through in-
home displays or smart phone applications) can generate learning effects in this area
(Faruqui, Sergici, & Sharif, 2010; Lynham, Nitta, Saijo, & Tarui, 2016). Residents may also
suffer from optimism bias (Dejoy, 1989) and perceive that they have greater ability to shift
their electricity use away from on-peak times (and thus more control over their electricity
bills) than they truly do. Energy-using appliances with which households interact more
frequently are more cognitively accessible and are viewed as more energy intensive than
activities with infrequent interaction (Schley & DeKay, 2015), but because appliances such as
lighting use less energy than cooling, people likely overestimate the amount of their energy
use that can be shifted through highly visible actions.
Our finding that perceived savings is the most powerful predictor of TOU acceptance raises
concerns that residents may enroll in TOU on the basis of perceived savings without actually
seeing any savings. This suggests that policies designed to support household DSR should
include support for utilities to provide participating households with mechanisms for
accessible feedback, such as comparing their bills and usage timing to previous years, to
reduce the likelihood of people falsely perceiving savings. Mediation analysis reveals that
measures intending to strengthen both perceived understanding of rates and perceived ease of
reducing evening use could be supported by policy to strengthen perceived savings, in turn
strengthening intent to remain on TOU. These measures may include, for example, IHDs and
clearer messaging regarding on-peak times, which additionally have the potential to increase
household ability to manage electricity spending.
The negative impact of actual understanding of rates on TOU acceptance suggests that people
with greater awareness of the length and timing of expensive on-peak rates view TOU more
negatively. This may be due to the current utility pilot placing on-peak rates in the evening,
when households often perform highly visible energy-using activities such as cooking,
watching television, and using lighting; the high cognitive accessibility of these evening
activities means residents likely perceive high evening energy consumption (Schley &
DeKay, 2015). Future TOU programs may benefit from different rate structures and/or
offering households more options to reduce usage during on-peak times without making
people feel the need to sacrifice planned activities (Hargreaves, Nye, & Burgess, 2010).
Measures such as policy support of IHDs or automated thermostat adjustment may
additionally bring perceived and actual understanding of rates into closer alignment,
particularly if these technologies incorporate visual cues regarding on-peak times and usage
(Hargreaves et al., 2010).

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
75

5 References
Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic Control Methods for
Comparative Case Studies: Estimating the Effect of California’s Tobacco Control
Program. Journal of the American Statistical Association, 105(490), 493–505.
http://doi.org/10.1198/jasa.2009.ap08746
Abadie, A., Diamond, A., & Hainmueller, J. (2015). Comparative Politics and the Synthetic
Control Method. American Journal of Political Science, 59(2), 495–510.
http://doi.org/10.1111/ajps.12116
Abadie, A., & Gardeazabal, J. (2003). The Economic Costs of Conflict: A Case Study of the
Basque Country. American Economic Review, 93(1), 113–132.
http://doi.org/10.1257/000282803321455188
Abrahamse, W., Steg, L., Vlek, C., & Rothengatter, T. (2005). A review of intervention
studies aimed at household energy conservation. Journal of Environmental Psychology,
25(3), 273–291. http://doi.org/10.1016/j.jenvp.2005.08.002
Achtnicht, M. (2012). German car buyers’ willingness to pay to reduce CO2 emissions.
Climatic Change, 113(3–4), 679–697. http://doi.org/10.1007/s10584-011-0362-8
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human
Decision Processes, 50(2), 179–211. http://doi.org/10.1016/0749-5978(91)90020-T
Alexander, B. R. (2010). Dynamic Pricing? Not So Fast! A Residential Consumer
Perspective. The Electricity Journal, 23(6), 39–49.
http://doi.org/10.1016/j.tej.2010.05.014
Allen, M. R., Frame, D. J., Huntingford, C., Jones, C. D., Lowe, J. A., Meinshausen, M., &
Meinshausen, N. (2009). Warming caused by cumulative carbon emissions towards the
trillionth tonne. Nature, 458(7242), 1163–6. http://doi.org/10.1038/nature08019
Asensio, O. I., & Delmas, M. A. (2016). The dynamics of behavior change: Evidence from
energy conservation. Journal of Economic Behavior & Organization, 126.
http://doi.org/10.1016/j.jebo.2016.03.012
Attari, S. Z. (2014). Perceptions of water use. Proceedings of the National Academy of
Sciences of the United States of America, 111(14), 5129–34.
http://doi.org/10.1073/pnas.1316402111
Attari, S. Z., DeKay, M. L., Davidson, C. I., & Bruine de Bruin, W. (2010). Public
perceptions of energy consumption and savings. Proceedings of the National Academy
of Sciences of the United States of America, 107(37), 16054–9.
http://doi.org/10.1073/pnas.1001509107
Axsen, J., & Kurani, K. S. (2013). Hybrid, plug-in hybrid, or electric-What do car buyers
want? Energy Policy, 61. http://doi.org/10.1016/j.enpol.2013.05.122
Axsen, J., TyreeHageman, J., & Lentz, A. (2012). Lifestyle practices and pro-environmental
technology. Ecological Economics, 82, 64–74.
http://doi.org/10.1016/j.ecolecon.2012.07.013
Barbarossa, C., Beckmann, S. C., De Pelsmacker, P., Moons, I., & Gwozdz, W. (2015). A
self-identity based model of electric car adoption intention: A cross-cultural comparative
study. Journal of Environmental Psychology, 42, 149–160.
Dissertation by Lee V. White
76

http://doi.org/10.1016/j.jenvp.2015.04.001
Baron, R. M., & Kenny, D. a. (1986). The moderator-mediator variable distinction in social
psychological research: conceptual, strategic, and statistical considerations. Journal of
Personality and Social Psychology, 51(6), 1173–1182. http://doi.org/10.1037/0022-
3514.51.6.1173
Barth, M., Jugert, P., & Fritsche, I. (2016). Still underdetected - Social norms and collective
efficacy predict the acceptance of electric vehicles in Germany. Transportation
Research Part F: Traffic Psychology and Behaviour, 37.
http://doi.org/10.1016/j.trf.2015.11.011
Betsill, M. ., & Rabe, B. . (2009). Climate change and multi-level governance: the evolving
state and local roles. In D. A. Mazmanian & M. E. Kraft (Eds.), Towards sustainable
communities.
Brounen, D., Kok, N., & Quigley, J. M. (2013). Energy literacy, awareness, and conservation
behavior of residential households. Energy Economics, 38, 42–50.
http://doi.org/10.1016/j.eneco.2013.02.008
Brucker, T., Bashmakov, I. A., Mulugetta, Y., Chum, H., de la Vega Navarro, A., Edmonds,
J., … Zhang, X. (2014). Energy Systems. In O. Edenhofer, R. Pichs-Madruga, Y.
Sokona, S. Farahani, S. Kadner, K. Seyboth, … J. C. Minx (Eds.), Climate Change
2014: Mitigation of Climate Change. Contribution of Working Group III to the Fifth
Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge,
United Kingdom and New York, NY, USA: Cambridge University Press. Retrieved
from http://www.ipcc.ch/pdf/assessment-report/ar5/wg3/ipcc_wg3_ar5_chapter7.pdf
Brugge, D., Durant, J. L., & Rioux, C. (2007). Near-highway pollutants in motor vehicle
exhaust: a review of epidemiologic evidence of cardiac and pulmonary health risks.
Environmental Health : A Global Access Science Source, 6(1), 23.
http://doi.org/10.1186/1476-069X-6-23
Buchanan, K., Banks, N., Preston, I., & Russo, R. (2016). The British public’s perception of
the UK smart metering initiative: Threats and opportunities. Energy Policy, 91.
http://doi.org/10.1016/j.enpol.2016.01.003
Buchanan, K., Russo, R., & Anderson, B. (2015). The question of energy reduction: The
problem(s) with feedback. Energy Policy, 77, 89–96.
http://doi.org/10.1016/j.enpol.2014.12.008
Bulkeley, H. (2010). Cities and the Governing of Climate Change. Annual Review of
Environment and Resources, 35(1), 229–253. http://doi.org/10.1146/annurev-environ-
072809-101747
Bulkeley, H., & Betsill, M. M. (2013). Revisiting the urban politics of climate change.
Environmental Politics, 22(1), 136–154. http://doi.org/10.1080/09644016.2013.755797
Burgess, M., King, N., Harris, M., & Lewis, E. (2013). Electric vehicle drivers’ reported
interactions with the public: Driving stereotype change? Transportation Research Part
F: Traffic Psychology and Behaviour, 17, 33–44.
http://doi.org/10.1016/j.trf.2012.09.003
Burkhardt, J., Wiser, R., Darghouth, N., Dong, C. G., & Huneycutt, J. (2015). Exploring the
impact of permitting and local regulatory processes on residential solar prices in the
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
77

United States. Energy Policy, 78, 102–112. http://doi.org/10.1016/j.enpol.2014.12.020
California New Car Dealers Association. (2014). California Auto Outlook. Retrieved from
http://www.cncda.org/CMS/Pubs/Cal_Covering_3Q_14.pdf
California Secretary of State. (2015). Voter Registration Statistics. Retrieved July 10, 2016,
from http://www.sos.ca.gov/elections/voter-registration/voter-registration-statistics
Cameron, A. C., & Trivedi, P. K. (2009). Microeconometrics using Stata. College Station,
Tex.: Stata Press.
Caperello, N. D., & Kurani, K. S. (2012). Households’ Stories of Their Encounters With a
Plug-In Hybrid Electric Vehicle. Environment and Behavior, 44(4), 493–508.
http://doi.org/10.1177/0013916511402057
Carley, S., Krause, R. M., Lane, B. W., & Graham, J. D. (2013). Intent to purchase a plug-in
electric vehicle: A survey of early impressions in large US cites. Transportation
Research Part D: Transport and Environment, 18(1), 39–45.
http://doi.org/10.1016/j.trd.2012.09.007
Caves, D. W., Herriges, J. A., & Kuester, K. A. (1989). Load Shifting Under Voluntary
Residential Time-of-Use Rates. The Energy Journal, 10, 83–99.
http://doi.org/10.2307/41322373
CEC. (2015). California Electric Utility Service Areas Map. Retrieved January 16, 2016,
from http://www.energy.ca.gov/maps/serviceareas/Electric_Utility_Service_Areas.html
CEC, & CPUC. (2016). History of Solar Rebates in California. Retrieved April 9, 2016, from
http://www.gosolarcalifornia.ca.gov/solar_basics/rebate_history.php
City of San Diego and the California Center for Sustainable Energy, S. (2009). Barriers and
Solutions: A Detailed Analysis of Solar Photovoltaics in San Diego.
Corfee, K., Cullen, J., Graham, S., Fry, N. R., Davis, B., Meyer, A., … Bloch, C. (2014).
California Solar Initiative: Third‐Party Ownership Market Impact Study. San Francisco.
CPUC. (2016). Net Energy Metering (NEM). Retrieved March 16, 2016, from
http://cpuc.ca.gov/General.aspx?id=3800
CPUC. (2018). Residential Rate Reform / R.12-06-013. Retrieved January 12, 2018, from
http://www.cpuc.ca.gov/General.aspx?id=12154
CSE. (2016). AB 2188 Implementation.
Darby, S. J. (2012). Metering: EU policy and implications for fuel poor households. Energy
Policy, 49, 98–106. http://doi.org/10.1016/j.enpol.2011.11.065
Davidson, C., Drury, E., Lopez, A., Elmore, R., & Margolis, R. (2014). Modeling
photovoltaic diffusion: an analysis of geospatial datasets. Environmental Research
Letters, 9(7), 74009. http://doi.org/10.1088/1748-9326/9/7/074009
Dejoy, D. M. (1989). The optimism bias and traffic accident risk perception. Accident
Analysis & Prevention, 21(4), 333–340. http://doi.org/10.1016/0001-4575(89)90024-9
Delmas, M. A., Fischlein, M., & Asensio, O. I. (2013). Information strategies and energy
conservation behavior: A meta-analysis of experimental studies from 1975 to 2012.
Energy Policy, 61, 729–739. http://doi.org/10.1016/j.enpol.2013.05.109
Dissertation by Lee V. White
78

Dietz, T., Gardner, G. T., Gilligan, J., Stern, P. C., & Vandenbergh, M. P. (2009). Household
actions can provide a behavioral wedge to rapidly reduce US carbon emissions.
Proceedings of the National Academy of Sciences of the United States of America,
106(44), 18452–6. http://doi.org/10.1073/pnas.0908738106
Dillon, L. (2018, January 26). California will put 5 million electric cars on the road by 2030,
Gov. Jerry Brown says. Los Angeles Times. Retrieved from
http://www.latimes.com/politics/essential/la-pol-ca-essential-politics-updates-gov-
brown-california-will-put-5-1516999162-htmlstory.html
Donaldson, C., Thomas, R., & Torgerson, D. J. (1997). Validity of open-ended and payment
scale approaches to eliciting willingness to pay. Applied Economics, 29(1), 79–84.
http://doi.org/10.1080/000368497327425
Dong, C., & Wiser, R. (2013). The impact of city-level permitting processes on residential
photovoltaic installation prices and development times: An empirical analysis of solar
systems in California cities. Energy Policy, 63, 531–542.
http://doi.org/10.1016/j.enpol.2013.08.054
Drury, E., Miller, M., Macal, C. M., Graziano, D. J., Heimiller, D., Ozik, J., & Perry IV, T.
D. (2012). The transformation of southern California’s residential photovoltaics market
through third-party ownership. Energy Policy, 42, 681–690.
http://doi.org/10.1016/j.enpol.2011.12.047
Dumortier, J., Siddiki, S., Carley, S., Cisney, J., Krause, R. M., Lane, B. W., … Graham, J.
D. (2015). Effects of providing total cost of ownership information on consumers’ intent
to purchase a hybrid or plug-in electric vehicle. Transportation Research Part A: Policy
and Practice, 72, 71–86. http://doi.org/10.1016/j.tra.2014.12.005
Dütschke, E., & Paetz, A.-G. (2013). Dynamic electricity pricing—Which programs do
consumers prefer? Energy Policy, 59, 226–234.
http://doi.org/10.1016/j.enpol.2013.03.025
Egbue, O., & Long, S. (2012). Barriers to widespread adoption of electric vehicles: An
analysis of consumer attitudes and perceptions. Energy Policy, 48, 717–729.
http://doi.org/10.1016/j.enpol.2012.06.009
EIA. (2014). California leads the nation in the adoption of electric vehicles. Retrieved from
http://www.eia.gov/todayinenergy/detail.cfm?id=19131
EIA. (2016). Retail sales of electricity, monthly. Retrieved April 13, 2016, from
http://www.eia.gov/electricity/data/browser/#/topic/5?agg=0,1&geo=g&linechart=ELEC
.SALES.US-ALL.M~ELEC.SALES.US-RES.M~ELEC.SALES.US-
COM.M~ELEC.SALES.US-IND.M&columnchart=ELEC.SALES.US-
ALL.M~ELEC.SALES.US-RES.M~ELEC.SALES.US-COM.M~ELEC.SALES.US-
IND.M&map=E
EIA. (2017a). Air conditioning and other appliances increase residential electricity use in the
summer - Today in Energy - U.S. Energy Information Administration (EIA). Retrieved
December 15, 2017, from https://www.eia.gov/todayinenergy/detail.php?id=31312
EIA. (2017b). U.S. households’ heating equipment choices are diverse and vary by climate
region - Today in Energy - U.S. Energy Information Administration (EIA). Retrieved
December 15, 2017, from
https://www.eia.gov/todayinenergy/detail.php?id=30672&src=‹ Consumption
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
79

Residential Energy Consumption Survey (RECS)-f2
Ericson, T. (2011). Households’ self-selection of dynamic electricity tariffs. Applied Energy,
88(7), 2541–2547. http://doi.org/10.1016/j.apenergy.2011.01.024
Fabrigar, L. R., & Wegener, D. T. (2011). Exploratory Factor Analysis. Oxford University
Press.
Faruqui, A., & George, S. (2005). Quantifying Customer Response to Dynamic Pricing. The
Electricity Journal, 18(4), 53–63. http://doi.org/10.1016/j.tej.2005.04.005
Faruqui, A., & Malko, J. R. (1983). The residential demand for electricity by time-of-use: A
survey of twelve experiments with peak load pricing. Energy, 8(10), 781–795.
http://doi.org/10.1016/0360-5442(83)90052-X
Faruqui, A., & Palmer, J. (2011). Dynamic Pricing of Electricity and its Discontents. SSRN
Electronic Journal. http://doi.org/10.2139/ssrn.1908963
Faruqui, A., & Sergici, S. (2010). Household response to dynamic pricing of electricity: a
survey of 15 experiments. Journal of Regulatory Economics, 38(2), 193–225.
http://doi.org/10.1007/s11149-010-9127-y
Faruqui, A., Sergici, S., & Sharif, A. (2010). The impact of informational feedback on energy
consumption—A survey of the experimental evidence. Energy, 35(4), 1598–1608.
http://doi.org/10.1016/j.energy.2009.07.042
Feiock, R. C., Krause, R. M., Hawkins, C. V., & Curley, C. (2013). The Integrated City
Sustainability Database. Urban Affairs Review, 50(4), 577–589.
http://doi.org/10.1177/1078087413515176
Fell, M. J., Shipworth, D., Huebner, G. M., & Elwell, C. A. (2015). Public acceptability of
domestic demand-side response in Great Britain: The role of automation and direct load
control. Energy Research & Social Science, 9, 72–84.
http://doi.org/10.1016/j.erss.2015.08.023
Fiorino, D. J. (2014). Sustainable cities and governance: what are the connections? In D. A.
Mazmanian & H. Blanco (Eds.), Elgar Companion to Sustainable Cities: Strategies,
Methods and Outlook (pp. 413–433). Edward Elgar Publishing Limited.
Franke, T., & Krems, J. F. (2013). Interacting with limited mobility resources: Psychological
range levels in electric vehicle use. Transportation Research Part A: Policy and
Practice, 48, 109–122. http://doi.org/10.1016/j.tra.2012.10.010
Freimuth, V. S., & Hovick, S. R. (2012). Cognitive and Emotional Health Risk Perceptions
Among People Living in Poverty. Journal of Health Communication, 17(3), 303–318.
http://doi.org/10.1080/10810730.2011.626505
Friedman, B., Ardani, K., Feldman, D., Citron, R., Margolis, R., & Zuboy, J. (2013).
Benchmarking Non-Hardware Balance-of-System (Soft) Costs for U.S. Photovoltaic
Systems, Using a Bottom-Up Approach and Installer Survey – Second Edition.
Geels, F. W. (2004). From sectoral systems of innovation to socio-technical systems.
Research Policy, 33(6–7), 897–920. http://doi.org/10.1016/j.respol.2004.01.015
Gino, F., Sharek, Z., & Moore, D. A. (2011). Keeping the illusion of control under control:
Ceilings, floors, and imperfect calibration. Organizational Behavior and Human
Dissertation by Lee V. White
80

Decision Processes, 114(2), 104–114. http://doi.org/10.1016/j.obhdp.2010.10.002
Graham-Rowe, E., Gardner, B., Abraham, C., Skippon, S., Dittmar, H., Hutchins, R., &
Stannard, J. (2012). Mainstream consumers driving plug-in battery-electric and plug-in
hybrid electric cars: A qualitative analysis of responses and evaluations. Transportation
Research Part A: Policy and Practice, 46(1), 140–153.
http://doi.org/10.1016/j.tra.2011.09.008
Grewal, R., Mehta, R., & Kardes, F. R. (2000). The role of the social-identity function of
attitudes in consumer innovativeness and opinion leadership. Journal of Economic
Psychology, 21(3), 233–252. http://doi.org/10.1016/S0167-4870(00)00003-9
Griskevicius, V., Tybur, J. M., & Van den Bergh, B. (2010). Going green to be seen: status,
reputation, and conspicuous conservation. Journal of Personality and Social
Psychology, 98(3), 392–404. http://doi.org/10.1037/a0017346
Hahnel, U. J. J., Ortmann, C., Korcaj, L., & Spada, H. (2014). What is green worth to you?
Activating environmental values lowers price sensitivity towards electric vehicles.
Journal of Environmental Psychology, 40, 306–319.
http://doi.org/10.1016/j.jenvp.2014.08.002
Hargreaves, T., Nye, M., & Burgess, J. (2010). Making energy visible: A qualitative field
study of how householders interact with feedback from smart energy monitors. Energy
Policy, 38(10), 6111–6119. http://doi.org/10.1016/j.enpol.2010.05.068
Hargreaves, T., Nye, M., & Burgess, J. (2013). Keeping energy visible? Exploring how
householders interact with feedback from smart energy monitors in the longer term.
Energy Policy, 52, 126–134. http://doi.org/10.1016/j.enpol.2012.03.027
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process
analysis. New York: The Guilford Press.
Heffner, R. R., Kurani, K. S., & Turrentine, T. S. (2007). Symbolism in California’s early
market for hybrid electric vehicles. Transportation Research Part D: Transport and
Environment, 12(6), 396–413. http://doi.org/10.1016/j.trd.2007.04.003
Hidrue, M. K., Parsons, G. R., Kempton, W., & Gardner, M. P. (2011). Willingness to pay
for electric vehicles and their attributes. Resource and Energy Economics, 33(3), 686–
705. http://doi.org/10.1016/j.reseneeco.2011.02.002
Hoppmann, J., Huenteler, J., & Girod, B. (2014). Compulsive policy-making—The evolution
of the German feed-in tariff system for solar photovoltaic power. Research Policy,
43(8), 1422–1441. http://doi.org/10.1016/j.respol.2014.01.014
ICMA. (2015). Solar Case Study: Lancaster, California. Retrieved from
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uac
t=8&ved=0ahUKEwiUl-
GOicPPAhWI7IMKHTDCA3cQFggeMAA&url=http%3A%2F%2Ficma.org%2FDocu
ments%2FDocument%2FDocument%2F307197&usg=AFQjCNEa3UjjSYInYc0t8rN8k
USCfD5yVw&sig2=tl-tKlklvvdHXUiNPv
IEA. (2011). Technology Roadmap: Smart grids.
IEA. (2012). EV City casebook: A look at the global electric vehicle movement, 1–75.
IEA. (2016). Global EV Outlook 2016. Retrieved from
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
81

https://www.iea.org/publications/freepublications/publication/Global_EV_Outlook_201
6.pdf
IEA. (2017). Electricity Information: Overview. Retrieved from
http://www.iea.org/publications/freepublications/publication/ElectricityInformation2017
Overview.pdf
IPCC. (2013). Summary for Policymakers. In T. F. Stocker, D. Qin, G.-K. Plattner, M.
Tignor, S. K. Allen, J. Boschung, … P. M. Midgley (Eds.), Climate Change 2013: The
Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report
of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and
New York, NY, USA: Cambridge University Press. Retrieved from
https://www.ipcc.ch/pdf/assessment-report/ar5/wg1/WG1AR5_SPM_FINAL.pdf
IPCC. (2014). Summary for policymakers. In C. B. Field, V. R. Barros, D. J. Dokken, K. J.
Mach, M. D. Mastrandrea, T. E. Bilir, … L. L. White (Eds.), Climate Change 2014:
Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects.
Contribution of Working Group II to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change (pp. 1–32). Cambridge, United Kingdom
and New York, NY, USA,: Cambridge University Press. Retrieved from
https://www.ipcc.ch/pdf/assessment-report/ar5/wg2/ar5_wgII_spm_en.pdf
IREC. (2013). Simplifying the Solar Permitting Process: Residential Solar Permitting Best
Practices Explained. Retrieved from http://www.irecusa.org/publications/residential-
solar-permitting-best-practices-explained/
Iwata, K., Katayama, H., & Arimura, T. H. (2015). Do households misperceive the benefits
of energy-saving actions? Evidence from a Japanese household survey. Energy for
Sustainable Development, 25, 27–33. http://doi.org/10.1016/J.ESD.2014.12.005
Jackman, M., & Lorde, T. (2014). Why buy when we can pirate? The role of intentions and
willingness to pay in predicting piracy behavior. International Journal of Social
Economics, 41(9), 801–819. http://doi.org/10.1108/IJSE-04-2013-0104
Jensen, A. F., Cherchi, E., & de Dios Ortúzar, J. (2014). A long panel survey to elicit
variation in preferences and attitudes in the choice of electric vehicles. Transportation,
41(5), 973–993. http://doi.org/10.1007/s11116-014-9517-6
Jensen, A. F., Cherchi, E., & Mabit, S. L. (2013). On the stability of preferences and attitudes
before and after experiencing an electric vehicle. Transportation Research Part D:
Transport and Environment, 25, 24–32. http://doi.org/10.1016/j.trd.2013.07.006
Jessoe, K., & Rapson, D. (2014). Knowledge is (Less) Power: Experimental Evidence from
Residential Energy Use. American Economic Review, 104(4), 1417–1438.
http://doi.org/10.1257/aer.104.4.1417
Jochem, P., Babrowski, S., & Fichtner, W. (2015). Assessing CO2 emissions of electric
vehicles in Germany in 2030. Transportation Research Part A: Policy and Practice, 78,
68–83. http://doi.org/10.1016/j.tra.2015.05.007
Kaatz, J., & Anders, S. (2015). AB 2188: Implementation of the Solar Rights Act at the Local
Level (Rev. ed.). Retrieved from
http://energycenter.org/sites/default/files/docs/nav/policy/research-and-
reports/AB_2188_Implementation_Guide.pdf
Dissertation by Lee V. White
82

Kaatz, J., & Anders, S. J. (2014). Residential and Commercial Property Assessed Clean
Energy (PACE) Financing in California Rooftop Solar Challenge Areas. Retrieved from
http://energycenter.org/sites/default/files/docs/nav/policy/research-and-reports/PACE in
California.pdf
Kastner, I., & Stern, P. C. (2015). Examining the decision-making processes behind
household energy investments: A review. Energy Research & Social Science, 10, 72–89.
http://doi.org/10.1016/j.erss.2015.07.008
Keizer, K., Lindenberg, S., Steg, L., Cialdini, R., Reno, R., Kallgren, C., … Ellickson, R.
(2013). The Importance of Demonstratively Restoring Order. PLoS ONE, 8(6), e65137.
http://doi.org/10.1371/journal.pone.0065137
Kirkpatrick, A. J., & Bennear, L. S. (2014). Promoting clean energy investment: An
empirical analysis of property assessed clean energy. Journal of Environmental
Economics and Management, 68(2), 357–375. http://doi.org/10.1016/j.jeem.2014.05.001
Klos, M., Erickson, J., Bryant, E., & Ringhof, S. L. (2008). Communicating Thermostats for
Residential Time-of-Use Rates: They Do Make a Difference.
Krupa, J. S., Rizzo, D. M., Eppstein, M. J., Brad Lanute, D., Gaalema, D. E., Lakkaraju, K.,
& Warrender, C. E. (2014). Analysis of a consumer survey on plug-in hybrid electric
vehicles. Transportation Research Part A: Policy and Practice, 64, 14–31.
http://doi.org/10.1016/j.tra.2014.02.019
Kurani, K. S., Turrentine, T. S., & Heffner, R. R. (2006). Narrative Self-Identity and Societal
Goals : Automotive Fuel Economy and Global Warming Policy, 217–238.
Kurani, K. S., Turrentine, T., & Sperling, D. (1996). Testing electric vehicle demand in
“hybrid households” using a reflexive survey. Transportation Research Part D:
Transport and Environment, 1(2), 131–150. http://doi.org/10.1016/S1361-
9209(96)00007-7
Kwan, C. L. (2012). Influence of local environmental, social, economic and political
variables on the spatial distribution of residential solar PV arrays across the United
States. Energy Policy, 47, 332–344. http://doi.org/10.1016/j.enpol.2012.04.074
Lawrence Berkeley National Lab. (2016). Tracking the Sun Public Data File. Retrieved
August 24, 2016, from https://openpv.nrel.gov/assets/data/openpv_tts_data.zip
Li, H., & Yi, H. (2014). Multilevel governance and deployment of solar PV panels in U.S.
cities. Energy Policy, 69, 19–27. http://doi.org/10.1016/j.enpol.2014.03.006
Lynham, J., Nitta, K., Saijo, T., & Tarui, N. (2016). Why does real-time information reduce
energy consumption? Energy Economics, 54, 173–181.
http://doi.org/10.1016/j.eneco.2015.11.007
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A
comparison of methods to test mediation and other intervening variable effects.
Psychological Methods, 7(1), 83–104. http://doi.org/10.1037/1082-989X.7.1.83
Matsukawa, I. (2001). Household Response to Optional Peak-Load Pricing of Electricity.
Journal of Regulatory Economics, 20(3), 249–267.
http://doi.org/10.1023/A:1011115025920
Meinshausen, M., Meinshausen, N., Hare, W., Raper, S. C. B., Frieler, K., Knutti, R., …
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
83

Allen, M. R. (2009). Greenhouse-gas emission targets for limiting global warming to 2
degrees C. Nature, 458(7242), 1158–1162. http://doi.org/10.1038/nature08017
Mendonça, M., Jacobs, D., & Sovacool, B. (2010). Powering the Green Economcy: The feed-
in tariff handbook. London: Earthscan.
Menz, F. C., & Vachon, S. (2006). The effectiveness of different policy regimes for
promoting wind power: Experiences from the states. Energy Policy, 34(14), 1786–1796.
http://doi.org/10.1016/j.enpol.2004.12.018
Miles, J., & Shevlin, M. (2001). Applying Regression and Correlation: A Guide for Students
and Researchers. London: Sage.
Millard-Ball, A. (2012). Do city climate plans reduce emissions? Journal of Urban
Economics, 71(3), 289–311. http://doi.org/10.1016/J.JUE.2011.12.004
Miller, M. J. (2014). Did “Pookie” get a green-collar job? : a critical case study on the East
Bay Green Corridor’s employment goals, activities, and impacts. Massachusetts
Institute of Technology. Retrieved from https://dspace.mit.edu/handle/1721.1/90213
Mills, C., & Newick, K. (2008). Solar Electric Permit Fees in Northern California: A
Comparative Study.
Mills, C., Newick, K., Stewart, J., & Compeán, T. W. (2009). Solar Electric Permit Fees in
Southern California: A Comparative Report.
Moons, I., & De Pelsmacker, P. (2012). Emotions as determinants of electric car usage
intention. Journal of Marketing Management, 28(3–4), 195–237.
http://doi.org/10.1080/0267257X.2012.659007
Mostafa Baladi, S., Herriges, J. A., & Sweeney, T. J. (1998). Residential response to
voluntary time-of-use electricity rates. Resource and Energy Economics, 20(3), 225–
244. http://doi.org/10.1016/S0928-7655(97)00025-0
Mountain, D. C., & Lawson, E. L. (1995). Some initial evidence of Canadian responsiveness
to time-of-use electricity rates: Detailed daily and monthly analysis. Resource and
Energy Economics, 17(2), 189–212. http://doi.org/10.1016/0928-7655(94)00025-F
Nayum, A., & Klöckner, C. A. (2014). A comprehensive socio-psychological approach to car
type choice. Journal of Environmental Psychology, 40, 401–411.
http://doi.org/10.1016/j.jenvp.2014.10.001
Neter, J., Wasserman, W., & Kutner, M. H. (1989). Applied Linear Regression Models.
Homewood: Irwin.
Nicolson, M., Huebner, G., & Shipworth, D. (2017). Are consumers willing to switch to
smart time of use electricity tariffs? The importance of loss-aversion and electric vehicle
ownership. Energy Research & Social Science, 23, 82–96.
http://doi.org/10.1016/j.erss.2016.12.001
Nixon, H., & Saphores, J.-D. (2011). Understanding Household Preferences For Alternative-
Fuel Vehicle Technologies. MTI Report.
Noppers, E. H., Keizer, K., Bolderdijk, J. W., & Steg, L. (2014). The adoption of sustainable
innovations: Driven by symbolic and environmental motives. Global Environmental
Change, 25, 52–62. http://doi.org/10.1016/j.gloenvcha.2014.01.012
Dissertation by Lee V. White
84

Noppers, E., Keizer, K., Bockarjova, M., & Steg, L. (2015). The adoption of sustainable
innovations: the role of instrumental, environmental, and symbolic attributes for earlier
and later adopters. Journal of Environmental Psychology.
http://doi.org/10.1016/j.jenvp.2015.09.002
NREL. (2011). Sacramento, California: Solar in Action, Solar America Cities, Energy
Efficiency and Renewable Energy. Retrieved from
https://energy.gov/sites/prod/files/2014/01/f7/51057_sacramento.pdf
NREL. (2012). Renewable Electricity Futures Study. Retrieved from
http://www.nrel.gov/analysis/re_futures
Osborne, J. ., & Costello, A. B. (2009). Best practices in exploratory factor analysis: four
recommendations for getting the most from your analysis. Pan Pacific Management
Review, 12(2), 131–146.
Ouellette, J. A., & Wood, W. (1998). Habit and intention in everyday life: The multiple
processes by which past behavior predicts future behavior. Psychological Bulletin,
124(1), 54–74.
Pacific Energy Center. (2006). The Pacific Energy Center’s Guide to: California Climate
Zones and Bioclimatic Design.
Pearre, N. S., Kempton, W., Guensler, R. L., & Elango, V. V. (2011). Electric vehicles: How
much range is required for a day’s driving? Transportation Research Part C: Emerging
Technologies, 19(6), 1171–1184. http://doi.org/10.1016/j.trc.2010.12.010
Peters, A., & Dütschke, E. (2014). How do Consumers Perceive Electric Vehicles? A
Comparison of German Consumer Groups. Journal of Environmental Policy &
Planning, 16(3), 359–377. http://doi.org/10.1080/1523908X.2013.879037
Pincetl, S., Graham, R., Murphy, S., & Sivaraman, D. (2016). Analysis of High-Resolution
Utility Data for Understanding Energy Use in Urban Systems: The Case of Los Angeles,
California. Journal of Industrial Ecology, 20(1), 166–178.
http://doi.org/10.1111/jiec.12299
Pitt, D. (2008). Taking the Red Tape out of Green Power.
Plötz, P., Schneider, U., Globisch, J., & Dütschke, E. (2014). Who will buy electric vehicles?
Identifying early adopters in Germany. Transportation Research Part A: Policy and
Practice, 67, 96–109. http://doi.org/10.1016/j.tra.2014.06.006
Pritoni, M., Meier, A. K., Aragon, C., Perry, D., & Peffer, T. (2015). Energy efficiency and
the misuse of programmable thermostats: The effectiveness of crowdsourcing for
understanding household behavior. Energy Research & Social Science, 8, 190–197.
http://doi.org/10.1016/j.erss.2015.06.002
Procter, R. J. (2013). Integrating Time-Differentiated Rates, Demand Response, and Smart
Grid to Manage Power System Costs. The Electricity Journal, 26(3), 50–60.
http://doi.org/10.1016/j.tej.2013.02.017
Rai, V., Reeves, D. C., & Margolis, R. (2016). Overcoming barriers and uncertainties in the
adoption of residential solar PV. Renewable Energy, 89, 498–505.
http://doi.org/10.1016/j.renene.2015.11.080
Reiss, P. C., & White, M. W. (2005). Household Electricity Demand, Revisited. Review of
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
85

Economic Studies, 72(3), 853–883. http://doi.org/10.1111/0034-6527.00354
Reiss, P., & White, M. (2003). Demand and Pricing in Electricity Markets: Evidence from
San Diego During California’s Energy Crisis. Cambridge, MA. Retrieved from
http://www.nber.org/papers/w9986.pdf
REN21. (2014). REN21 - Renewables Global Status Report. Retrieved December 6, 2014,
from http://www.ren21.net/ren21activities/globalstatusreport.aspx
Rezvani, Z., Jansson, J., & Bodin, J. (2015). Advances in consumer electric vehicle adoption
research: A review and research agenda. Transportation Research Part D: Transport
and Environment, 34, 122–136. http://doi.org/10.1016/j.trd.2014.10.010
Robinson, S. A., & Rai, V. (2015). Determinants of spatio-temporal patterns of energy
technology adoption: An agent-based modeling approach. Applied Energy, 151, 273–
284. http://doi.org/10.1016/j.apenergy.2015.04.071
Rose, A., Stevens, B., Edmonds, J., & Wise, M. (1998). International Equity and
Differentiation in Global Warming Policy. Environmental and Resource Economics,
12(1), 25–51. http://doi.org/10.1023/A:1008262407777
Rosen, A. M. (2015). The Wrong Solution at the Right Time: The Failure of the Kyoto
Protocol on Climate Change. Politics & Policy, 43(1), 30–58.
http://doi.org/10.1111/polp.12105
Rowlands, I. H., & Furst, I. M. (2011). The cost impacts of a mandatory move to time-of-use
pricing on residential customers: an Ontario (Canada) case-study. Energy Efficiency,
4(4), 571–585. http://doi.org/10.1007/s12053-011-9113-y
Schelly, C. (2014). Residential solar electricity adoption: What motivates, and what matters?
A case study of early adopters. Energy Research & Social Science, 2, 183–191.
http://doi.org/10.1016/j.erss.2014.01.001
Schley, D. R., & DeKay, M. L. (2015). Cognitive accessibility in judgments of household
energy consumption. Journal of Environmental Psychology, 43, 30–41.
http://doi.org/10.1016/j.jenvp.2015.05.004
Schuitema, G., Anable, J., Skippon, S., & Kinnear, N. (2013). The role of instrumental,
hedonic and symbolic attributes in the intention to adopt electric vehicles.
Transportation Research Part A: Policy and Practice, 48, 39–49.
http://doi.org/10.1016/j.tra.2012.10.004
Schwartz, S. H. (1977). Normative Influences on Altruism. Advances in Experimental Social
Psychology, 10, 221–279. http://doi.org/10.1016/S0065-2601(08)60358-5
Seel, J., Barbose, G. L., & Wiser, R. H. (2014). An analysis of residential PV system price
differences between the United States and Germany. Energy Policy, 69, 216–226.
http://doi.org/10.1016/j.enpol.2014.02.022
Seto, C. K., Dhakal, S., Bigio., A., Blanco, H., Delgado, G. C., Dewar, D., … Ramaswami,
A. (2014). Human Settlements, Infrastructure and Spatial Planning. In T. Z. andJ. C. M.
Edenhofer, O., R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A.
Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C.
von Stechow (Ed.), In: Climate Change 2014: Mitigation of Climate Change.
Contribution of Working Group III to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New
Dissertation by Lee V. White
86

York, NY, USA: Cambridge University Press.
Sexton, S. E., & Sexton, A. L. (2014). Conspicuous conservation: The Prius halo and
willingness to pay for environmental bona fides. Journal of Environmental Economics
and Management, 67(3), 303–317. http://doi.org/10.1016/j.jeem.2013.11.004
SF Environment. (2012). San Francisco’s Streamlined Solar Permitting. Retrieved from
http://sfenvironment.org/sites/default/files/fliers/files/san_franciscos_streamlined_solar_
permitting.pdf
Sheldon, T. L., & DeShazo, J. R. (2017). How does the presence of HOV lanes affect plug-in
electric vehicle adoption in California? A generalized propensity score approach.
Journal of Environmental Economics and Management.
http://doi.org/10.1016/j.jeem.2017.05.002
Sierzchula, W., Bakker, S., Maat, K., & Van Wee, B. (2014). The influence of financial
incentives and other socio-economic factors on electric vehicle adoption. Energy Policy,
68, 183–194. http://doi.org/10.1016/j.enpol.2014.01.043
Sintov, N., Geislar, S., & White, L. V. (2017). Cognitive Accessibility as a New Factor in
Proenvironmental Spillover: Results From a Field Study of Household Food Waste
Management. Environment and Behavior, 1391651773563.
http://doi.org/10.1177/0013916517735638
Skippon, S., & Garwood, M. (2011). Responses to battery electric vehicles: UK consumer
attitudes and attributions of symbolic meaning following direct experience to reduce
psychological distance. Transportation Research Part D: Transport and Environment,
16(7), 525–531. http://doi.org/10.1016/j.trd.2011.05.005
Sperling, D., & Eggert, A. (2014). California’s climate and energy policy for transportation.
Energy Strategy Reviews, 5, 88–94. http://doi.org/10.1016/j.esr.2014.10.001
Steg, L. (2005). Car use: Lust and must. Instrumental, symbolic and affective motives for car
use. Transportation Research Part A: Policy and Practice, 39(2–3 SPEC. ISS.), 147–
162. http://doi.org/10.1016/j.tra.2004.07.001
Steg, L., Bolderdijk, J. W., Keizer, K., & Perlaviciute, G. (2014). An Integrated Framework
for Encouraging Pro-environmental Behaviour: The role of values, situational factors
and goals. Journal of Environmental Psychology, 38, 104–115.
http://doi.org/10.1016/j.jenvp.2014.01.002
Stern, P. C. (1999). Information, Incentives, and Proenvironmental Consumer Behavior.
Journal of Consumer Policy, 22, 461–478.
Stern, P. C. (2000). Toward a coherent theory of environmentally significant behavior.
Journal of Social Issues, 56(3), 407–424. Retrieved from
http://www.scopus.com/inward/record.url?eid=2-s2.0-
0034473324&partnerID=tZOtx3y1
Stern, P. C. (2014). Individual and household interactions with energy systems: Toward
integrated understanding. Energy Research & Social Science, 1, 41–48.
http://doi.org/10.1016/j.erss.2014.03.003
SunShot and North Carolina Solar Center. (2013). Harmonizing Solar PV Permitting and
Interconnection in Santa Clara, CA. Retrieved from https://nccleantech.ncsu.edu/solar-
center-releases-new-case-study-harmonizing-solar-pv-permitting-and-interconnection/
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
87

Tan, K. M., Ramachandaramurthy, V. K., & Yong, J. Y. (2016). Integration of electric
vehicles in smart grid: A review on vehicle to grid technologies and optimization
techniques. Renewable and Sustainable Energy Reviews, 53, 720–732.
http://doi.org/10.1016/j.rser.2015.09.012
Taufik, D., Bolderdijk, J. W., & Steg, L. (2016). Going green? The relative importance of
feelings over calculation in driving environmental intent in the Netherlands and the
United States. Energy Research & Social Science, 22, 52–62.
http://doi.org/10.1016/j.erss.2016.08.012
Todd, A., Cappers, P., & Goldman, C. (2013). Residential Customer Enrollment in Time-
based Rate and Enabling Technology Programs: Smart Grid Investment Grant
Consumer Behavior Study Analysis.
Tofighi, D., & MacKinnon, D. P. (2011). RMediation: An R package for mediation analysis
confidence intervals. Behavior Research Methods, 43(3), 692–700.
http://doi.org/10.3758/s13428-011-0076-x
Toftum, J. (2010). Central automatic control or distributed occupant control for better indoor
environment quality in the future. Building and Environment, 45(1), 23–28.
http://doi.org/10.1016/j.buildenv.2009.03.011
Train, K. E., McFadden, D. L., & Goett, A. A. (1987). Consumer Attitudes and Voluntary
Rate Schedules for Public Utilities. The Review of Economics and Statistics, 69(3), 383.
http://doi.org/10.2307/1925525
Train, K., & Mehrez, G. (1994). Optional Time-of-Use Prices for Electricity: Econometric
Analysis of Surplus and Pareto Impacts. The RAND Journal of Economics, 25(2), 263.
http://doi.org/10.2307/2555830
Treadwell, T., Gishri, T., Wheeland, J., Anders, S., & Kaatz, J. (2012). Streamlining Solar
Standards and Processes Southern California Rooftop Solar Challenge.
Turrentine, T. S., & Kurani, K. S. (2007). Car buyers and fuel economy? Energy Policy,
35(2), 1213–1223. http://doi.org/10.1016/j.enpol.2006.03.005
U.S. Census Bureau. (2013). 2009-2013 American Community Survey 5-Year Estimates.
Retrieved October 7, 2015, from http://quickfacts.census.gov/qfd/states/06/06037lk.html
U.S. Department of Energy. (2015a). Electric Vehicle Charging Station Locations. Retrieved
March 17, 2015, from http://www.afdc.energy.gov/fuels/electricity_locations.html
U.S. Department of Energy. (2015b). The eGallon: How Much Cheaper Is It to Drive on
Electricity? Retrieved December 11, 2015, from http://energy.gov/articles/egallon-how-
much-cheaper-it-drive-electricity
U.S. Environmental Protection Agency. (2015a). U.S. Transportation Sector Greenhouse Gas
Emissions 1990-2011. Retrieved from
http://www.epa.gov/otaq/climate/documents/420f13033a.pdf
U.S. Environmental Protection Agency. (2015b). U.S. Transportation Sector Greenhouse
Gas Emissions 1990-2012. Retrieved from
http://www3.epa.gov/otaq/climate/documents/420f15002.pdf
Van der Werff, E., & Steg, L. (2015). One model to predict them all: Predicting energy
behaviours with the norm activation model. Energy Research & Social Science, 6, 8–14.
Dissertation by Lee V. White
88

http://doi.org/10.1016/j.erss.2014.11.002
Van der Werff, E., Steg, L., & Keizer, K. (2013a). It is a moral issue: The relationship
between environmental self-identity, obligation-based intrinsic motivation and pro-
environmental behaviour. Global Environmental Change, 23(5), 1258–1265.
http://doi.org/10.1016/j.gloenvcha.2013.07.018
Van der Werff, E., Steg, L., & Keizer, K. (2013b). The value of environmental self-identity:
The relationship between biospheric values, environmental self-identity and
environmental preferences, intentions and behaviour. Journal of Environmental
Psychology, 34, 55–63. http://doi.org/10.1016/j.jenvp.2012.12.006
Van der Werff, E., Steg, L., & Keizer, K. (2014). I Am What I Am, by Looking Past the
Present: The Influence of Biospheric Values and Past Behavior on Environmental Self-
Identity. Environment and Behavior, 46(5), 626–657.
http://doi.org/10.1177/0013916512475209
Vergis, S., & Chen, B. (2015). Comparison of plug-in electric vehicle adoption in the United
States: A state by state approach. Research in Transportation Economics.
http://doi.org/10.1016/j.retrec.2015.10.003
Wang, X., Hawkins, C. V., Lebredo, N., & Berman, E. M. (2012). Capacity to Sustain
Sustainability: A Study of U.S. Cities. Public Administration Review, n/a-n/a.
http://doi.org/10.1111/j.1540-6210.2012.02566.x
White, H. (1980). A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct
Test for Heteroskedasticity. Econometrica, 48(4), 817–838.
http://doi.org/10.2307/1912934
Whitmarsh, L., & O’Neill, S. (2010). Green identity, green living? The role of pro-
environmental self-identity in determining consistency across diverse pro-environmental
behaviours. Journal of Environmental Psychology, 30(3), 305–314.
http://doi.org/10.1016/j.jenvp.2010.01.003
Wiser, R., Namovicz, C., Gielecki, M., & Smith, R. (2007). The Experience with Renewable
Portfolio Standards in the United States. The Electricity Journal, 20(4), 8–20.
http://doi.org/10.1016/j.tej.2007.03.009
World Health Organization. (2005). Health effects of transport-related air pollution. ….
Retrieved from
http://www.bvsde.paho.org/bvsacd/cd63/e86650/conten.pdf%5Cnpapers2://publication/
uuid/0896BDA1-6025-41A0-A320-420490564B66
Yin, H., & Powers, N. (2010). Do state renewable portfolio standards promote in-state
renewable generationʔ. Energy Policy, 38(2), 1140–1149.
http://doi.org/10.1016/j.enpol.2009.10.067
Yuan, Y., & MacKinnon, D. P. (2014). Robust mediation analysis based on median
regression. Psychological Methods, 19(1), 1–20. http://doi.org/10.1037/a0033820
Zhang, K., & Batterman, S. (2013). Air pollution and health risks due to vehicle traffic.
Science of the Total Environment, 450–451, 307–316.
http://doi.org/10.1016/j.scitotenv.2013.01.074
Zwick, W. R., & Velicer, W. F. (1986). Comparison of five rules for determining the number
of components to retain. Psychological Bulletin, 99(3), 432–442.
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
89

Dissertation by Lee V. White
90

6 Chapter 1 Article Appendices
6.1 Chapter 1, Appendix 1: Fixed Effects Analysis of PV adoption rates
This appendix supplements the analyses in Chapter 1, and provides a table of coefficients
from fixed effects model of factor impacts on PV installation rates, using cluster robust
standard errors clustered at city level. All models included fixed effects dummies for year and
city.
Table 6.1: Fixed effects model of factor impacts on PV installation rates, using cluster robust
standard errors clustered at city level. All models included fixed effects dummies for year and
city.
W/capita PV
installed
annually
(main sample)
Number of PV
systems
installed/1000
people annually
(main sample)
W/capita PV
installed
annually
(sub-sample)
Number of PV
systems
installed/1000
people annually
(sub-sample)
Streamlined
permitting
0.03
(1.62)
0.04
(0.33)
1.26
(2.21)
0.27
(0.45)
Permitting fees
<$400
0.63
(1.47)
0.13
(0.33)
PACE -0.03
(1.58)
0.07
(0.34)
Electricity price
(log)
-5.51
(4.63)
-1.16
(0.88)
-0.18
(7.13)
-0.31
(1.38)
Financial incentives
(log)
-0.75
(0.68)
-0.09
(0.15)
-0.14
(1.49)
0.03
(0.33)
Share democratic -0.03
(0.66)
0.02
(0.14)
0.72
(1.35)
0.17
(0.27)
Age (sq) 0.01*
(0.00)
0.00**
(0.00)
0.01*
(0.00)
0.00*
(0.00)
Share with income >
$100,000
-0.13
(0.15)
-0.02
(0.03)
-0.16
(0.34)
-0.02
(0.07)
Share with income
$100,000 to $50,000
-0.20+
(0.11)
-0.04+
(0.02)
-0.41**
(0.14)
-0.08**
(0.03)
Education -0.11
(0.11)
-0.02
(0.02)
-0.21
(0.19)
-0.05
(0.04)
R
2
0.53 0.53 0.53 0.53
N 476 476 207 207
Standard errors in parentheses
+
p < 0.10,
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
91

7 Chapter 1, Supplementary Appendix: Synthetic Control Analysis of PV
adoption rates
This appendix supplements the analyses in Chapter 1, and describes the synthetic control
analyses undertaken during development of the paper. These synthetic control analyses were
not included in the paper as submitted for review (i.e., Chapter 1), due to their inconclusive
results.
7.1 Method
Synthetic control methods are used as an additional analysis strategy that can better account
for potential differences in these trends, and that is both highly suitable for analyzing small
sample sizes with limited treated units and was designed to examine aggregate units and
outcomes such as the present sample. This relatively new empirical approach, pioneered by
Abadie and others (Abadie, Diamond, & Hainmueller, 2010, 2015; Abadie & Gardeazabal,
2003), weights several “donor” units to construct a single “synthetic control” that closely
follows the trend of a single treated unit pre-treatment – and which can theoretically represent
the trend that the treated unit would have continued along in absence of treatment.
When constructing the donor pool for a synthetic control to be drawn from, care needs to be
taken that the cities in the donor pool are not wildly different from the treated city. If cities
are too different, then this may introduce issues of interpolation bias or overfitting (Abadie et
al., 2015). Hence, very large cities such as San Francisco cannot be readily examined with
synthetic control methods, because there are not enough sufficiently similar cities within
California to create a reasonable donor pool. Cities with populations of 65,000 to 650,000
were included in the synthetic control model so that extremely large cities such as Los
Angeles were systematically excluded, and these were then weighted based on demographic
and other factors for each city including median age, median income, share of population
over age 25 with a bachelor’s degree or higher, and insolation rates. Electricity prices,
financial incentives, and demographic voter share were not included in the constraints, as
they largely vary at the utility and county level and this may have skewed the weighting of
the synthetic control.
Synthetic control methods don’t allow tests for statistical significance in the same way that
traditional regression models do, but a method using placebo tests has been developed to
produce an indication of statistical significance analogous to that provided by t-stats in the
absence of these standard tools. Placebo testing rests on the premise that there should not be
similar magnitudes of estimated differences between the synthetic control and unit of interest
if we substitute a “unit of interest” where no intervention took place (Abadie et al., 2015).
Placebo tests can be performed to examine all donor units as if they were “treated” to test for
estimations showing differences between “treated” and “control” – if there is no policy effect
found when the “treated” unit does not actually receive treatment, this indicates that any
findings for the true treated unit are unlikely to have arisen solely due to chance.
7.2 Results
Synthetic control analyses were run for six cities that both met streamlined permitting criteria
and did so near the middle of the panel years observed. Figure 2 shows the results of placebo
analysis, with the treated unit shown in black and the placebo tests shown in light grey. These
synthetic control tests are unable to reject the null hypothesis that implementing streamlined
permitting has no effect on residential PV installation rates. The treated unit effects always
Dissertation by Lee V. White
92

fall within the envelope of placebo test effects, indicating that any effects observed due to
treatment could have been random. Further, the effect sizes of the treated units are close to or
below zero in most cases, which does not support a reject of the null hypothesis. The poor
match pre-treatment severely limits any conclusions that can be drawn from the analyses; in
particular, this poor pre-treatment match indicates that the synthetic control analyses are
unlikely to be able to fully incorporate omitted variables, so bias may still be present due to
unobserved differences between cities.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
93

Figure 7.1: Synthetic control methods showing placebo analyses of effect sizes (treated unit
in black, placebo units in light grey)

Dissertation by Lee V. White
94

8 Chapter 2 Article Appendices
This appendix includes the appendices as presented in the journal article published in
Transportation Research Part A (with table and figure numbering updated for consistency
with main dissertation document).
8.1 Chapter 2, Appendix 1: Descriptive statistics for EV adoption survey sample
Table 8.1: Descriptive statistics for key variables among main sample (246 < N < 355
respondents with available demographic data) and sub-sample (N = 224 respondents will full
data on demographic variables)
Characteristic Sample (%)
a
Sample (%)
N = 224
Los Angeles County
b

(%)
Ethnicity
African American
Asian / Asian American
Caucasian
Latino
Native American / Pacific Islander
Multiracial
Other
(N = 346)
2.3
11.9
67.6
10.7
0.6
4.9
2.0

1.8
12.1
67.4
11.2
0.5
5.4
1.8

8.1
13.7
27.5
47.9
0.2
2.1
0.2
Educational Attainment (>= 25 yrs)
Less than High School
High school diploma
Some college / Associate’s
degree
4-year college degree
Graduate / professional degree
High School diploma or higher
Bachelor’s degree or higher
(N = 353)
1.1
4.3
22.1
37.7
34.8
98.9
72.5

0.9
3.13
18.8
38.4
38.8
99.0
77.2

23.4
20.5
26.5
19.4
10.2
76.6
29.7
Annual household income
<$25,000
$25,001-$50,000
$50,001-$75,000
$75,001-$100,000
>$100,000
(N = 294)
9.9
13.6
20.4
13.6
42.5

8.9
15.2
21.9
12.5
41.5

Median household
income: $55,909
Home ownership rate (N = 355) 60.8 53.1 46.9
Home type
Single Family Home
Apartment/Condo
Duplex, Triplex
Townhouse
Mobile Home
Other
(N = 354)
52.8
35.9
5.9
4.5
0.6
0.3

46.4
42.0
6.7
4.5
0.5
0

49.7
34.3
8.0
6.5
1.5
Gender (% male) (N = 284) 51.8 50.0 49.3
Age (median) (N = 246) 53 48.0 35.1
a
Not all respondents answered all demographic questions. Available data for respondents
who answered questions is as follows: N ethnicity = 346, N education = 353, N income = 294, N home
ownership = 355, Nhome type = 354, Ngender = 284, Nage = 246.
b
(US Census Bureau, 2013).
8.2 Chapter 2, Appendix 2: Scree plot for symbolic attributes of EVs
Figure 8.1: Scree plot for symbolic attributes factor analysis; promax-rotated iterative
principal factor analysis constrained to three factors
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
95

0 1 2 3
Eigenvalues
1 2 3 4 5
Number
Scree plot of eigenvalues after factor
Dissertation by Lee V. White
96

8.3 Chapter 2, Appendix 3: Factor loadings of symbolic attributes for EVs
Table 8.2: Promax rotated factor loadings and uniqueness of symbolic attributes items based
on iterative principal factor analysis constrained to two factors
Question Factor 1:
Environmentalist
Factor 2:
Social Innovator
Uniqueness
(1) Owning an EV demonstrates to
others that I care about the
environment
0.78 0.05 0.37
(2) Changing from a gasoline-
powered vehicle to an EV will lessen
my impact on the environment
0.84 -0.03 0.32
(3) Driving an EV means that I am
doing the right thing
0.72 0.27 0.19
(4) Driving an EV means that I am a
trendsetter for environmentally
friendly technologies
0.30 0.55 0.42
(5) Driving an EV means that I am
socially responsible
-0.04 0.54 0.73
Eigenvalues 2.49 1.61

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
97

9 Chapter 3 Article Appendices
9.1 Chapter 3, Appendix 1: Models of TOU acceptance with unstandardized
coefficients

Table 9.1: Models of TOU acceptance (unstandardized coefficients)
(1) (2) (3)
Satisfaction with
current rate plan
Intent to remain on
TOU
Intent to remain on
TOU
On-peak use at baseline
(kWh, daily average)
-0.05
***

(0.00)
0.03
***

(0.01)
0.03
***

(0.01)
Decrease in on-peak use
(kWh, daily average)
-0.00
(0.01)
-0.04
(0.02)
-0.04
(0.02)
Perceived ease of reducing
evening use
0.05
***

(0.01)
0.04
**

(0.01)
0.04
**

(0.01)
Perceived savings 0.68
***

(0.01)
0.85
***

(0.01)
0.78
***

(0.02)
Decrease in bills ($) 0.00
*

(0.00)
0.00
**

(0.00)
0.00
**

(0.00)
Perceived understanding of
rate

0.10
***

(0.02)
Actual understanding of
peak times

-0.35
***

(0.07)
Control rate plan 0.00
(.)

TOU plan 1 -0.11
*

(0.04)
0.00
(.)
0.00
(.)
TOU plan 2 -0.26
***

(0.05)
0.10
(0.06)
0.09
(0.06)
Home owned 0.05
(0.04)
0.05
(0.06)
0.05
(0.06)
Education 0.13
***

(0.04)
0.02
(0.06)
0.06
(0.06)
Low-income 0.29
***

(0.04)
0.12
(0.06)
0.09
(0.06)
R
2
0.58 0.57 0.58
N 8945 5249 5249
Huber-White robust standard errors in parentheses
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

Dissertation by Lee V. White
98

9.2 Chapter 3, Appendix 2: Mediation analyses with Rate Satisfaction as the
Dependent Variable

Figure 9.1a: Mediation of perceived understanding by perceived savings using median
regression. Indirect effect = 0.798 (SE=0.016), and for 99% confidence interval interval
LL=0.759, UL=0.839

Figure 9.1b: Mediation of actual understanding by perceived savings using median
regression. Indirect effect = -0.575 (SE=0.100), and for 99% confidence interval LL=-0.833,
UL=-0.318

Figure 9.1c: Mediation of perceived ease of reducing evening use by perceived savings using
median regression. Indirect effect = 0.566 (SE=0.016), and for 99% confidence interval
LL=0.526, UL=0.607

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
99

10 Chapter 3, Supplementary Appendix: Additional predictors of TOU
acceptance included in initial analyses, to be formed into a second
journal publication

Chapter 3 originally addressed a wider set of hypotheses, and was split into two halves
following discussant feedback at the Association for Public Policy Analysis and Management
(APPAM) Conference in November 2017. This appendix provides a brief overview of the
original framework as presented at APPAM prior to the paper being split. Sections of the
original literature review that are no longer covered in the main paper are also included in this
appendix for clarification. Several variables were recoded after this initial analysis to reduce
ambiguities and address potential sources of error, and initial analyses were briefly re-run
with this updated coding for applicable variables to generate the results presented in this
section.
10.1 Initial Abstract
Maintaining a reliable electricity supply with increasing shares of intermittent and
decentralized renewable generation will require both upgraded infrastructure and greater
flexibility within electric grids. Time of Use (TOU) rates are one of the many demand-side
response measures that can facilitate this increased flexibility. Although pilot programs have
established that households reduce peak electricity demand in response to TOU on-peak
rates, it is not yet well understood which factors affect household acceptance of TOU, nor
whether these factors differ for based on household income. Building on recent findings that
perceptions of control predict intentions to adopt TOU plans, this paper creates a model of
residential customer TOU acceptance to investigate the roles of household income,
perceptions around control of spending and household activities, and ownership of smart
thermostats in predicting TOU acceptance. We use a sample of several thousand residential
electricity customers who took part in a large utility-sponsored TOU pilot in Southern
California.
We find that perceptions of control over spending and household activities are far stronger
predictors of TOU acceptance than actual change in spending or usage (based on interval
energy data and billing data). The impact of perceived spending control on TOU acceptance
is stronger among low- (vs. higher) income households, and these households are even more
likely to intend to continue on TOU if during the pilot they reported sacrificing comfort to
save money during on-peak times. Findings can inform utility program design and
implementation; given that TOU acceptance is predicted more strongly by perceptions of
control than actual achieved management of electricity bills, enhancing customer perceptions
of control (e.g., through messaging) could improve TOU acceptance.
10.2 Initial Literature Review Framing for Perceived vs. Actual Savings
The hypotheses examined in Chapter 3 are drawn from what was initial H2a-H2d, and these
initial hypotheses and associated literature review are described in the following sections.
10.2.1 Usage patterns
Households on TOU rates face higher prices for electricity consumed during on-peak times,
but can sometimes benefit financially from a shift to TOU rates without behavior change if
electricity use commonly occurs during off-peak times (Rowlands & Furst, 2011). Thus, if
Dissertation by Lee V. White
100

households typically perform electricity-intensive activities during off-peak times to begin
with, TOU would require little change in existing daily activities and would be more likely to
lead to cost savings rather than cost increases. Households are expected to evaluate whether
they would attain greater surplus under TOU rates or under standard tiered rates, and are
expected to prefer whichever rates gives the highest surplus (K. Train & Mehrez, 1994); thus,
households with usage patterns whereby heavy electricity use tends to fall during off-peak
times, or where on-peak usage is generally low, are expected to be more accepting of TOU.
Several studies have confirmed that households with usage patterns favorable to TOU rates
i.e., with less usage in on-peak times, are more likely to opt-in to TOU (Matsukawa, 2001; K.
E. Train, McFadden, & Goett, 1987; K. Train & Mehrez, 1994). However, other studies have
found no relationship between actual favorable load profiles and TOU adoption (Mostafa
Baladi et al., 1998) or between projected monetary savings and TOU adoption (Mountain &
Lawson, 1995).
H1: Households that begin with low actual usage during on-peak times will be more
accepting of TOU
10.2.2 Perceived control
Perceived control associated with TOU rates and electricity use is expected to heavily impact
acceptance of TOU; households are more likely to switch to TOU rates if they have a strong
sense that their individual actions are effective at reducing electricity use (K. E. Train et al.,
1987), and including perceived control constructs in previous models of hypothetical TOU
acceptance increased model explanatory power (Fell et al., 2015). Ajzen (1991), in the
Theory of Planned Behavior (TPB), describes perceived behavioral control as whether “the
person can decide at will whether to perform or not perform the behavior” (pp 182), and
expands that this refers to the ease or difficulty with which people perceive they can perform
a behavior. Perceived control in the context of TOU rates can affect several dimensions of
behavior performance, which were grouped by Fell et al. (2015) into two major constructs:
spending control, and a second construct comprised of autonomy, timing control, and comfort
control which will here be referred to as ‘activity control’.
10.2.2.1 Perceived Activity Control
Since TOU rates impose a price penalty for using electricity during on-peak times, there is a
corresponding penalty for a person deciding to perform an activity that uses electricity during
those times, such that following from Ajzen’s definition the person’s ability to decide at will
to perform the behavior has been lessened. This applies similarly across autonomy, timing
control, and comfort control, the three of which Fell (2015) found to be highly collinear and
to form a single construct which here will be referred to as activity control. These were
defined as “Comfort (such as being able to obtain desired thermal conditions in the home)”,
“Timing (control over when people do things, such as running appliances like dishwashers)”
and “Autonomy (a more general sense of directing events in one’s life, free of outside
influence)” (Fell et al., 2015, p2).
Following the definition of activity control as households being able to use appliances at
times that they wish to, a household is expected to be less accepting of TOU if using
appliances during off-peak vs. on-peak times creates discomfort or is difficult to integrate
with household schedules. Greater demand flexibility, i.e., the ability to adjust demand away
from TOU on-peak times, has been found to predict TOU acceptance (Caves et al., 1989;
Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
101

Ericson, 2011; K. Train & Mehrez, 1994). However, households may not be willing to
change their electricity use habits (Hargreaves et al., 2013). Some electricity uses such as
lighting and television are viewed by some households as non-negotiable, in both timing and
performance (Dütschke & Paetz, 2013; Hargreaves et al., 2013). Thus, households that
perceive it is easy to shift activities, i.e., that do not consider activity timing non-negotiable
or activity shifting difficult, are expected to be more accepting of TOU.
10.2.2.2 Perceived Spending Control
If a person on TOU rates does not change the timing of their electricity usage then they are
still able to perform their activities at will, but they will receive a price penalty if they use
large amounts of electricity during peak-rate times. This is complicated due to the relative
invisibility of electricity use and the generally poor understanding that households may have
of which activities are electricity-intensive (Attari et al., 2010; Brounen et al., 2013). A
household could perform an activity without knowing that it would be very expensive at that
time of day; following from Ajzen’s definition, they would not have truly decided at will to
incur higher electricity costs if they were not fully aware of the cost penalty for a given
activity. This is a different type of control concern compared to activity control, since the first
deals with forced changes of timing and the second deals with potentially unintentional
spending due to price penalties fixed in time. This separation of control types was also
supported in Fell (2015), where spending control was not found to be collinear with
autonomy, timing control, and comfort control.
As described above, loss of perceived spending control may occur across two main
dimensions: lack of understanding of the energy costs of different activities, and lack of
understanding about which times on-peak and off-peak TOU rates occur. Households have
been found to slightly overestimate the energy use of low-energy appliances such as lighting
and laptops, and to vastly underestimate the energy use of high-energy appliances such as
heating and cooling (Attari et al., 2010). This might lead to a perception that spending control
on TOU rates is very low, since households might e.g. curtail lighting and laptop use during
on-peak times while still using AC at high levels, which would lead to high electricity bills
regardless of attempts to curtail. Further, if households do not understand TOU rates well,
then they will have more difficulty predicting their monthly electricity bills or understanding
what times to avoid electricity intensive activities. It is expected that households with a better
understanding of TOU rate timing and function will be more accepting of TOU.
Conversely, households would perceive having greater spending control if they are able to
recall the times at which TOU on-peak vs. off-peak rates occur, and particularly so if they
perceive that they typically perform energy intensive activities during off-peak times – i.e.,
perceive that the rate is a good match to their existing schedule and provides them
opportunities to save money. However in previous studies, greater perceived spending control
was not positively associated with willingness to switch to a TOU rate (Fell et al., 2015). This
is surprising given the centrality of monetary mechanisms to TOU function, and the
expectation that TOU rate acceptance is at least in part financially motivated (K. Train &
Mehrez, 1994). The current study utilizes survey questions focused on pilot participants’
perceived ability to save money on TOU rates, whereas Fell et al. (2015) asked households
questions with a focus on feelings of being in charge of electricity spending for a plan they
had not actually experienced living with. Both the use of a sample that has experienced TOU
rates and the focus on perceived savings is expected to more strongly capture the influence of
Dissertation by Lee V. White
102

perceived spending control on TOU acceptance, and the ability to control spending and/or
save money on a TOU rate is expected to be a positive predictor of intent to stay on TOU
rates post-pilot.
10.2.2.3 Perceived vs. Actual Control
People have been found to make imperfect estimates of their level of control, such that they
may underestimate control when they have high control and overestimate control when they
have low control (Gino, Sharek, & Moore, 2011). When examining thermal comfort, the level
of perceived control is more important for occupant comfort than actual ventilation mode
(Toftum, 2010), and this disconnect between perceived control and actual conditions may
also occur for households adapting to TOU. Households may have a poor understanding of
which activities use the most electricity (Attari et al., 2010), and so may not truly know
which actions will reduce electricity charges during on-peak times. Thus, households may
perceive that they have high spending control even if their actions do not lead to lower bills
on TOU.
It is expected that perceiving high spending control will have a greater positive effect on
TOU acceptance than achieving lower bills, as many households are not aware of their
monthly energy bills (Brounen et al., 2013). It has also been found that the general population
tends to greatly underestimate the energy used by heating and cooling of homes (Attari et al.,
2010), and that many households had misconceptions about the way that household heating
systems functioned: “…over one third of the respondents did not understand the relationship
between inside temperatures and heat loss.” (Pritoni, Meier, Aragon, Perry, & Peffer, 2015).
This type of fundamental misunderstanding of energy indicates that households’ perceptions
of usage patterns may be inaccurate, which could in turn cause inaccurate perceptions about
cost savings achieved through curtailment activities during on-peak times.
If households find it easy to adjust their usage to match TOU rates then they will have high
perceived activity control – that is, since the adjustment to TOU is perceived as easy the
households will not perceive a loss of control over activity timing and comfort. Household
adjustment to TOU rates is expected to be reflected in actual usage patterns, such that if
households found it easy to adjust their activities to coincide with TOU rates a drop in
quantity of electricity used during on-peak times is expected to occur. As with perceived
spending control, it is expected that perceived activity control will more strongly predict
TOU acceptance when compared to actual changes during on-peak usage.
H2a: Households that consider TOU rates to allow high perceived activity control will be
more accepting of TOU rates, i.e., households that do not face difficulty or inconvenience in
reducing usage during peak-rate times will be more accepting of TOU
H2b: Households that consider TOU rates to allow retention of perceived spending control
will be more accepting of TOU rates
H2c: Household perceptions of spending control will have a stronger effect on acceptance of
TOU compared to observed actual observed electricity bills
H2d: Household perceptions of activity control will have a stronger effect on acceptance of
TOU compared to actual observed reduction is on-peak electricity usage

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
103

10.3 Additional Segments from Initial Literature Review
10.3.1 Income
Income is expected to be particularly significant in predicting TOU acceptance. Households
with lower incomes are more sensitive to energy prices (P. C. Reiss & White, 2005), and
TOU rates sometimes raise concerns about low-income households and other vulnerable
groups facing fuel poverty, i.e., being unable to afford sufficient energy expenditures to
maintain comfort and health within their homes without sacrificing spending in other areas
such as food (Darby, 2012; Faruqui & Palmer, 2011). However, some simulations have
indicated that low-income customers may actually save money on TOU rates even without
changing and electricity use habits, since these households may have flatter load profiles
(Faruqui & Palmer, 2011). But in contrast, a study in Ontario, Canada found that households
with lower electricity usage were more likely to face higher costs when switched to TOU
rates from a two-tier regime (Rowlands & Furst, 2011).
10.3.1.1 Discomfort due to shifting usage
Households sacrificing comfort to reduce costs is a concern, particularly since it has
previously been found that income predicts thermostat settings with wealthier households in
cold climates keeping their homes warmer at night (Brounen et al., 2013). Further, household
members may perceive heating or cooling as wasteful once made more aware of the energy
costs (Hargreaves et al., 2010). Reviews have found a mixture of positive and negative
impacts specifically for fuel-poor households across prior TOU studies, and emphasize that
this group should be considered specifically in future studies (Darby, 2012). Potential issues
with TOU rates can be seen in the literature examining the effect of household-level feedback
on energy use through IHDs, e.g.,
“I am so aware of how much I’m using that I think to myself well do I need to? Do
I need to put my light on right now? Can I still sit here in the dark and work by
candlelight? Do I need to watch [TV] tonight, you know?” p 6116 (Hargreaves et
al., 2010)
“One interviewee – a member of the control group but who took regular meter
readings and carefully monitored his electricity use, a situation heightened during
participation in the trial – expressed this potential source of anxiety in a dramatic
metaphor, referring to his wife: She could kind of feel the money seeping out
every time she had the boiler on. And to be honest beating herself up over it, you
know. ‘I can’t have it on because I’m wasting money, but I’m cold’.” p 6114
(Hargreaves et al., 2010)
H3a: Compared to those that do not have low incomes, financially-vulnerable households
that have higher perceived spending control will have even higher TOU acceptance
H3b: Compared to those that do not have low incomes, financially-vulnerable households
reporting discomfort due to engaging in rate-shifting to save money will be more accepting of
TOU rates
10.3.2 Enabling technology
Customers with enabling technology such as smart thermostats have been shown to achieve
greater reductions in demand during peak pricing events for both TOU and CPP when
Dissertation by Lee V. White
104

compared with customers who only received informational materials (Faruqui et al., 2010;
Klos, Erickson, Bryant, & Ringhof, 2008). Customers offered smart thermostats have also
been found to view TOU slightly more favorably (Klos et al., 2008). Smart thermostats may
be a useful intermediary between manual control of temperatures and full DLC of
heating/cooling by utilities. While smart thermostat capabilities and specifications vary, in
general they typically will include features such as learning when a home is unoccupied and
will not need heating or cooling, and some models have the potential to incorporate utility
price variations over time into their settings. Thus, smart thermostats may represent a
mechanism to reduce electricity use during on-peak times with very little effort required by
households.
H4: Households with a smart thermostat will have higher acceptance of TOU rates
10.3.3 Environmental concern
It is expected that households will be more accepting of TOU if they have concerns about the
impacts of their own energy use on the environment. However, the link between TOU and
environmental benefits may not be clear to all participants, and previous research has found
that participants in dynamic TOU trials did not always find it apparent that the rate would
help with integration of renewable generation capacity (Dütschke & Paetz, 2013). There may
further be a disconnect between willingness to adopt pro-environmental programs and taking
action to reduce energy consumption while on these programs; environmental ideology has
been found in some studies to have no effect on choice of thermostat settings (Brounen et al.,
2013). Greater concern about a forthcoming energy crisis was also not found to predict
willingness to switch to TOU rates (K. E. Train et al., 1987).
H5: Households that feel compelled to reduce energy use due to environmental concerns are
more likely to decide to stay on TOU post-pilot
10.4 Methodology
10.4.1 Procedures
See procedures as described in Chapter 3, section 4.2.1. Only the TOU sub-sample was
examined for the initial analysis presented here in Section 10, and the additional independent
variables included here reduce the sample size to N=4920.
10.4.2 Actual usage and costs
See billing and usage data coding described in Chapter 3, section 4.2.2.1 and 4.2.2.2,
respectively. Due to coding improvements, the initial analyses were re-run with several
variables updated from the version initially presented at APPAM in November 2017.
10.4.3 Perceptions
10.4.3.1 Perceived activity control
See procedures as described in Chapter 3, section 4.2.2.3. This variable was renamed
‘Perceived ease of reducing evening use’ in the submitted paper.
10.4.3.2 Perceived spending control
See procedures as described in Chapter 3, section 4.2.2.3. This variable was renamed
‘Perceived savings’ in the submitted paper.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
105

10.4.4 Demographics
10.4.4.1 Financial stress
Financial stress faced by each household was assessed using five questions rated on a 5-point
Likert scale. Households indicated whether they felt the following statements described their
situations, from 1 (not at all) to 5 (completely): “Since June 2016, how well does this
statement describe you and your situation? Because of my money situation, I feel like I will
never have the things I want in life”, “Since June 2016, how well does this statement describe
you and your situation? I am just getting by financially”, “Since June 2016, how well does
this statement describe you and your situation? I am concerned that the money I have won’t
last”, “How often does this statement apply to you? I have money left over at the end of the
month” (reverse coded), and “How often does this statement apply to you? My finances
control my life”. These questions were found to have a Cronbach’s alpha = 0.86, and an
iterated principal factor analysis found that all items loaded onto the first factor > 0.65 with
the Eigenvalue for Factor 1 = 2.8. The scale was taken as the mean of responses for
households that had answered at least three of the five questions.
10.4.4.2 Low-income
See procedures as described in Chapter 3, section 4.2.2.3.
10.4.4.3 Home ownership
See procedures as described in Chapter 3, section 4.2.2.3.
10.4.4.4 Environmental concern
Environmental concern was measured with a scale taken as the mean of responses for
households that had answered at least two of three questions. These three questions asked
households to rate agreement with the following statements on an 11-point Likert scale where
0 = “Do not agree at all” and 10 = “completely agree”: “I am very concerned about how my
energy use affects the environment”, “It is my responsibility to use as little energy as possible
to help the environment”, and “I feel guilty if I use too much energy”. These items had a
Cronbach’s alpha of 0.84, and when tested with iterated principal factor analysis were found
to load onto the first factor > 0.72 with an Eigenvalue for Factor 1 = 2.0.

10.4.5 Physical constraints
10.4.5.1 Enabling technology
Ownership of a smart thermostat was coded as a dichotomous variable, with 1 indicating that
respondents had claimed to own a smart thermostat. Respondents were assigned 0 if they
either stated that they owned another kind of thermostat, or did not own any kind of
thermostat.

10.4.5.2 Climate zone
There are 16 climate zones in California, which are used to set building codes based on
climactic factors including expected heating and cooling degree days. Households were
marked with the climate zone in which they reside. The average HDD and CDD for each
climate zone are taken from the reported HDD and CDD of four major cities in each climate
zone as an indicator for thermal conditions (Pacific Energy Center, 2006).
Dissertation by Lee V. White
106

10.4.5.3 Barriers to shifting usage
Households answered the question “Which of the following, if any, has kept you from
reducing or shifting your electricity usage in the afternoons and evenings?” with a ‘select all
that apply’ response including options such as “Child(ren) in household make it difficult to
change our routines”. The number of responses indicating a barrier to shifting usage was
tallied to provide a count of the number of barriers faced by each household to shifting usage.
10.4.6 Initial analyses
Ordinary Least Squares (OLS) multiple linear regression was run using the statistical analysis
software STATA to compare effect sizes of perceived control, ease of use, thermostat
ownership and other factors identified as relevant. Figure 10.1 shows the variables included,
and Table 10.1 provides summary statistics for the sample. Post-estimation tests indicated
heteroscedasticity of error terms, so Huber-White standard errors were applied to correct for
this. Non-normality was also revealed by post-estimation tests; while this is not expected to
affect the accuracy of co-efficient estimates, it indicates that tests of statistical significance
should be interpreted with caution.
Data also exhibited clustering at both extremes consistent with upper and lower censoring, so
a Tobit model was additionally run to better account for this. Use of Tobit models with
Likert-scale dependent variables is relatively rare, but prior examples exist in the literature
(Freimuth & Hovick, 2012; Jackman & Lorde, 2014). The Tobit model, presented in Table
10.3, largely replicates the pattern of results identified in the OLS models, and is provided as
a comparison to demonstrate that the upper and lower censoring present in the data did not
significantly affect OLS coefficient estimates.
Figure 10.1: Variables in main OLS regression model (grey indicates smart-meter/billing
variables, white indicates survey variables)

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
107

Table 10.1: Summary statistics (updated with post-conference variable coding as described
above)
Mean Standard Dev. Min. Max.
I'm satisfied with my
current rate plan
6.56 2.56 0.00 10.00
I'd like to stay on my new
rate
6.79 3.00 0.00 10.00
Perceived spending control 6.54 2.61 0.00 10.00
Perceived activity control 6.36 2.50 0.00 10.00
Decrease in bills ($) -22.80 52.87 -463.97 402.48
Decrease in on-peak use
(kWh, daily average)
0.23 1.82 -15.42 15.55
Actual understanding of
peak times
0.55 0.40 0.00 1.00
On-peak use at baseline
(kWh, daily average)
6.39 5.16 0.00 50.84
Education 0.50 0.50 0.00 1.00
Home owned 0.75 0.43 0.00 1.00
Low-income customer 0.39 0.49 0.00 1.00
Financial stress 2.78 0.97 1.00 5.00
Environmental concern 6.21 2.73 0.00 10.00
Number of barriers to
shifting usage
20.36 1.35 11.00 31.00
Frequency of discomfort
trying to save $
2.60 1.02 1.00 5.00
HDD 1803.29 632.96 1177.00 5056.75
CDD 2047.35 1260.85 595.75 4759.75
Owns smart thermostat 0.07 0.26 0.00 1.00
TOU plan 1.56 0.50 1.00 2.00
Observations 4920

10.4.7 Initial results
10.4.7.1 Main regression
Table 10.2 presents the results of OLS multiple regression with Huber-White standard errors
and showing standardized coefficients, while Table 10.3 provides a comparison between
unstandardized coefficients of OLS and Tobit models (confirming that these are similar in
size).
TOU acceptance was predominantly predicted by greater perceived spending control,
strongly supporting H2b, and the standardized coefficient for perceived spending control was
over 10 times the magnitude of the next largest independent variable. VIF testing did not
indicate any multi-collinearity issues between perceived spending control and other variables,
suggesting that despite the extremely high explanatory power of perceived spending control it
was likely not obscuring effects of any other variables due to correlation. Greater perceived
activity control also positively predicted TOU acceptance, supporting H2a. Perceived
spending control was a far stronger predictor of TOU acceptance when compared to observed
bill change, supporting H2c. Results likewise supported H2d, as perceived activity control
was a stronger predictor of TOU acceptance than change in use during on-peak rates, the
latter of which has a negative impact on intent to adopt TOU. Households that reported better
understanding of their TOU rates were additionally more likely to report high TOU
Dissertation by Lee V. White
108

acceptance. Overall, results support the hypothesis that perceived control is more important
for TOU acceptance than actual changes in use and costs. Households that started with lower
usage occurring during off-peak hours were additionally more likely to have high TOU
acceptance, supporting H1.
A household being low-income did not impact intent to remain on TOU. However, if low-
income households perceived that they had high spending control while on TOU rates then
they reported higher TOU acceptance compared to higher-income households, supporting
H3a. The number of barriers households faced to shifting usage unexpectedly positively
predicts TOU acceptance, in isolation while having no impact as a moderator of low-income,
i.e., H3b was not supported. Households that reported higher levels of financial stress also
reported stronger higher TOU acceptance, and compared to higher-income households the
low-income households reported higher TOU acceptance if they experienced discomfort
while trying to save money on TOU rates, supporting H3c.
Owning a smart thermostat has a small negative impact on TOU acceptance, thus H4 is not
supported. Households that strongly reported feeling environmental concern related to their
energy consumption had higher TOU acceptance, in support of H5. As would be expected,
the control variable for homeownership indicates that those who own their homes have higher
TOU acceptance. The climate variables have no statistically significant impact, and there was
no difference in TOU acceptance reported for those on TOU rate 1 vs. TOU rate 2.

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
109

Table 10.2: OLS multiple regression with Huber-White standard errors, examining impact of
each factor on stated intent to remain on TOU after pilot
9

I'd like to stay on my new rate
On-peak use at baseline (kWh, daily average) 0.06
***

(0.01)
Decrease in on-peak use (kWh, daily average) -0.03
*

(0.02)
Perceived activity control 0.04
***

(0.01)
Perceived spending control 0.71
***

(0.02)
Decrease in bills ($) 0.05
***

(0.00)
Home owned 0.02
*

(0.07)
Education 0.01
(0.07)
Low-income customer 0.04
(0.94)
Interaction of low-income and spending control 0.08
*

(0.03)
Financial stress 0.07
***

(0.04)
Environmental concern 0.04
***

(0.01)
Number of barriers to shifting usage 0.04
*

(0.03)
Interaction of low-income and barriers to
shifting usage
-0.19
(0.05)
Frequency of discomfort trying to save $ -0.01
(0.04)
Interaction of low-income and discomfort trying
to save $
0.08
**

(0.06)
HDD -0.01
(0.00)
CDD 0.01
(0.00)
Owns smart thermostat -0.02
*

(0.11)
TOU plan=1 0.00
(.)
TOU plan=2 0.02
(0.06)
R
2
0.58
N 4920
Standardized beta coefficients; Standard errors in parentheses
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

9
Updated from the table and figures presented at APPAM in Fall 2017 to use improved variable coding for
variables that were also included in Chapter 3.
Dissertation by Lee V. White
110

Table 10.3: OLS vs. Tobit multiple regression, examining impact of each factor on stated
intent to remain on TOU after pilot
OLS Tobit
I'd like to stay on my new
rate
I'd like to stay on my new
rate
On-peak use at baseline (kWh, daily
average)
0.03
***

(0.01)
0.04
***

(0.01)
Decrease in on-peak use (kWh, daily
average)
-0.05
*

(0.02)
-0.07
*

(0.03)
Perceived activity control 0.05
***

(0.01)
0.06
**

(0.02)
Perceived spending control 0.81
***

(0.02)
1.04
***

(0.03)
Decrease in bills ($) 0.00
***

(0.00)
0.00
***

(0.00)
Home owned 0.17
*

(0.07)
0.30
**

(0.10)
Education 0.08
(0.07)
0.03
(0.09)
Low-income customer 0.26
(0.94)
-0.45
(1.35)
Interaction of low-income and
spending control
0.07
*

(0.03)
0.19
***

(0.04)
Financial stress 0.22
***

(0.04)
0.34
***

(0.05)
Environmental concern 0.04
***

(0.01)
0.06
***

(0.02)
Number of barriers to shifting usage 0.08
*

(0.03)
0.09
*

(0.04)
Interaction of low-income and
barriers to shifting usage
-0.06
(0.05)
-0.07
(0.06)
Frequency of discomfort trying to
save $
-0.03
(0.04)
-0.06
(0.06)
Interaction of low-income and
discomfort trying to save $
0.15
**

(0.06)
0.24
**

(0.09)
HDD -0.00
(0.00)
-0.00
(0.00)
CDD 0.00
(0.00)
0.00
(0.00)
Owns smart thermostat -0.28
*

(0.11)
-0.30
*

(0.15)
TOU plan=1 0.00
(.)
0.00
(.)
TOU plan=2 0.09
(0.06)
0.13
(0.09)
Sigma

2.57
***

(0.05)
R
2
0.58 0.19 (pseudo-R
2
)
N 4920 4920
Standard errors in parentheses
*
p < 0.05,
**
p < 0.01,
***
p < 0.001

Household carbon footprints: how to encourage adoption of emissions-reducing behaviors and technologies
111

10.4.8 Initial Discussion
Modeling results strongly supported the hypothesis that perceived control is a more powerful
predictor of TOU acceptance when compared to measures of actual change in usage and
billing, supporting H2a-H2d. The finding that changes in billing had only a small predictive
effect compared to perceived spending control is expected, based on previous findings that
households are often unaware of their energy bills (Brounen et al., 2013). Confirming the
overwhelming importance of perceived spending control above changes in bills has
implications for policy and TOU rate design, as does the mediation of understanding by
perceived spending control – an emphasis should be placed on designing TOU in such a way
that households perceive on-peak times to be easy to remember and hence easy to adjust
usage to.
Income did not affect TOU acceptance, but when spending control was included as a
moderator financially vulnerable households with stronger perceived spending control
reported higher TOU acceptance, supporting H3a. There is cause for some concern about the
tendency of households feeling financial stress to have stronger intent to remain on TOU,
coupled with the finding in support of H3b that financially vulnerable households are more
likely to support TOU if they also report feeling discomfort during the pilot due to trying to
save money on TOU rates. It has been repeatedly noted that TOU peak-rates may
unintentionally induce households to reduce their energy consumption in ways that are
detrimental to health through not sufficiently heating or cooling homes, and it has been noted
that this needs to be considered in future research so that such an effect can be avoided
(Alexander, 2010; Buchanan, Russo, & Anderson, 2015).
This finding relating to discomfort in particular indicates that TOU on-peak times may be
pushing households to cut energy use to a level that could be detrimental, although this
depends to an extent on environmental conditions and health of household members.
Conversely, the finding that these financially vulnerable households have preferences for
electricity rates that allow them to save money by making comfort sacrifices at select times
may allow these households to shift financial resources to non-energy areas while still having
the option to cool or heat sufficiently during off-peak and super off-peak times. However, the
increased awareness of the cost of everyday activities using electricity that TOU promotes
may cause a great deal of stress for financially vulnerable households (Hargreaves et al.,
2010). In-depth qualitative investigations of these households in future may be able to further
uncover the extent to which deliberate thermal discomfort to avoid on-peak rates can become
stressful or detrimental to health in the long-term vs. providing an additional method for
managing household costs.
Results also supported H1, that households would be more accepting of TOU if they had
begun in 2014 and 2015 lower baseline use during what would become on-peak times in
2016. There is some concern that households perceiving themselves to have lower usage
during evening times would have been more likely to volunteer to take part in the pilot in the
first place, and hence households that took part in the pilot may have been motivated by the
perceived ability to save to begin with; this in turn would impact the interpretation of results
addressing H2a-H2d. However, while previous research has indicated that households are
indeed more likely to volunteer for TOU pilots if they perceive that they have existing
favorable usage patterns, it also confirmed that the usage patterns of these volunteer
households did not actually differ from those of the general population, i.e., the difference
Dissertation by Lee V. White
112

was only perceived (Mostafa Baladi et al., 1998). The design of the pilot used in the present
analysis also sought to over-enroll certain participant groups (financially vulnerable and
elderly), and offered repayment in the case where costs on TOU rates exceeded what would
have been incurred on the household’s previous rate. Finally, even in the event that
households enrolled in the pilot were more favorable toward TOU for cost-saving potential to
begin with, the insights regarding what factors predicted intent to remain on the rate
following completion of the pilot still have relevance, as they provide an indication of which
factors remained appealing to households following actual experience with the rate.
The finding (against H4) that smart thermostat ownership has no significant impact on TOU
acceptance is unexpected, particularly since enabling technology such as smart thermostats is
often considered to be a convenient method for TOU response (Faruqui & Sergici, 2010;
Klos et al., 2008). This may be partially due to the configuration of thermostats at this stage
of the TOU pilot – households with smart thermostats did not have their thermostats
automatically set up to reduce energy use during on-peak times, limiting the convenience of
these thermostats as an enabling technology. Households may have been able to teach or set
their smart thermostats to take account of on-peak vs. off-peak rate times, but since this was
not provided as a service by the utility households would have needed to invest some time in
this measure. Households reporting greater concern about the environmental impact of their
energy use were more accepting of TOU rates (supporting H5), which suggests that it will be
useful to emphasize this area in messaging efforts regarding TOU. It is possible that this link
is particularly strongly in California, where solar is highly prevalent.
When default TOU rates are rolled out in California in 2019, the best way to encourage
households to remain on these rates rather than opting out is likely to be emphasizing the
potential to save money offered by TOU rates. Future work could also investigate to a greater
extent which factors predict perceived spending control, beyond reported understanding of
TOU rates. It is not fully clear whether greater energy literacy would drive greater perceived
spending control, but if this link can be established then it will have implications for
messaging associated with TOU.
10.4.9 Initial Conclusion
Acceptance of TOU rates is driven overwhelmingly by households having greater perceived
spending control, i.e., by households perceiving that they are able to save money on TOU
rates. Further, perceived spending control predicted intent to remain on TOU above and
beyond observable changes in electricity bills received by each household. Contrary to
expectations, smart thermostat ownership did not positively predict TOU acceptance, instead
having an insignificant effect. The analysis also highlights the trade-offs that financially
vulnerable households may make on TOU rates whereby comfort is sacrificed during on-peak
times to reduce electricity bills; results indicate that households making such trade-offs are
actually more likely to intend to remain on TOU rates outside of the pilot, indicating that the
option to reduce bills by temporarily sacrificing comfort may be seen as a desirable feature of
TOU rates. The results highlight the importance of perceptions compared to actual changes in
use for predicting TOU acceptance, and further suggest that future studies should maintain a
focus on financially vulnerable groups to better understand the wider implications of trade-
offs between energy costs and thermal comfort.

Abstract (if available)

Abstract Households account for over a third of CO₂₋ₑ gas emissions in the US, and in aggregate can contribute significantly to efforts to mitigate climate change by reducing the carbon footprint associated with their daily activities. However, we do not yet fully understand what drives households to adopt new and emerging technologies to reduce emissions, nor what encourages households to adopt new behaviors that can reduce emissions. Decisions at the household level can be affected by policies at the national, state, and local level offering market‐based incentives or streamlining regulatory environments, and can simultaneously be affected by the personal values, beliefs, and self‐identities of household members. This dissertation examines household adoption of solar photovoltaic panels, electric vehicles, and new electricity use behaviors, broadly framing analyses within the attitude‐behavior‐context theory to consider both personal level drivers and contextual policy and societal drivers of adoption. ❧ I conclude that household decisions to take part in climate change mitigation efforts are powerfully driven by perceptions of individual benefits including financial savings and ability to showcase care for the environment. Local governments may be able to remove contextual barriers to solar adoption, but this could not be confirmed nor disproved in the current analysis due to limitations on available city‐level data. Adoption of electric vehicles and new electricity use behaviors was confirmed to be powerfully driven by perceptions of symbolism and savings, respectively. Overall, findings emphasize the need to include consideration of personal factors when developing policy settings, rather than relying on purely financial instruments. Findings also emphasize the need to include mechanisms to strengthen household understanding of energy usage and costs, to facilitate stronger economic evaluations by households of their own energy use decisions.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Why go green? Cities' adoption of local renewable energy policies and urban sustainability certifications

PDF

Resilient and equitable urbanism by design: insights from the collaborative process to reimagine the SF Bay Area

PDF

A time of crisis: the Australian experience and what can California learn?

PDF

Emissions markets, power markets and market power: a study of the interactions between contemporary emissions markets and deregulated electricity markets

PDF

The economic and political impacts of U.S. federal carbon emissions trading policy across households, sectors and states

PDF

Measuring the drivers of economic, energy, and environmental changes: an index decomposition analysis

PDF

Environmental effects from a large-scale adoption of electric vehicle technology in the City of Los Angeles

PDF

The long-term impact of COVID-19 on commute, employment, housing, and environment in the post-pandemic era

PDF

Collaborating to provide microfinance to caregivers of orphans and vulnerable children in Ethiopia

PDF

Children’s travel behavior in journeys to school

PDF

Intersect policing: bringing CompStat to the field level to reduce the fear and incidence of crime

PDF

Healthy mobility: untangling the relationships between the built environment, travel behavior, and environmental health

PDF

The Wenchuan earthquake recovery: civil society, institutions, and planning

PDF

Residential electricity demand in the context of urban warming: leveraging high resolution smart meter data to quantify spatial and temporal patterns…

PDF

Reinforcement learning in hybrid electric vehicles (HEVs) / electric vehicles (EVs)

PDF

Active travel, outdoor leisure, and neighborhood environment: path analysis, Los Angeles County

PDF

The flexible workplace: regional tendencies and daily travel implications

PDF

Integration of truck scheduling and routing with parking availability

Asset Metadata

Creator White, Lee Victoria (author)

Core Title Household carbon footprints: how to encourage adoption of emissions‐reducing behaviors and technologies

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Urban Planning and Development

Publication Date 03/07/2018

Defense Date 03/02/2018

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

Tag behavior,demand side response,electric vehicle,Energy,environmental,household,OAI-PMH Harvest,Planning,policy,Residential,solar,sustainability,time of use

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Mazmanian, Dan (committee chair), Blanco, Hilda (committee member), Rose, Adam (committee member), Sintov, Nicole. D. (committee member)

Creator Email leevictoriawhite@gmail.com,leewhite@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-484207

Unique identifier UC11268077

Identifier etd-WhiteLeeVi-6086.pdf (filename),usctheses-c40-484207 (legacy record id)

Legacy Identifier etd-WhiteLeeVi-6086.pdf

Dmrecord 484207

Document Type Dissertation

Rights White, Lee Victoria

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA