Energy Research & Social Science 23 (2017) 97–112

Contents lists available at ScienceDirect

Energy Research & Social Science

journal homepage: www.elsevier.com/locate/erss

Original research article

Energy use behaviors in buildings: Towards an integrated conceptual

framework
Da Li, Carol C. Menassa ∗ , Aslihan Karatas
Dept. of Civil and Environmental Engineering, University of Michigan, United States

a r t i c l e i n f o a b s t r a c t

Article history: To achieve significant energy reductions in buildings, decision-makers can engage occupants in different
Received 3 July 2016 types of interventions such as information sharing, feedback and social marketing. To improve the effec-
Received in revised form tiveness of these energy saving interventions, this study develops and tests a model which is capable of
10 November 2016
identifying occupants’ energy use characteristics and the influential factors of their energy use behaviors
Accepted 10 November 2016
Available online 11 December 2016
(e.g., turning off lights when not in use). The consumer segmentation approach from social marketing
is adopted to divide occupants into different categories using three metrics: motivation, opportunity
and ability. As a result, a set of hypotheses and corresponding measures are identified to study the
Keywords:
Energy use characteristic effect of influential factors on occupants’ energy use characteristics and intentional energy use behav-
Occupants’ behavior iors. The occupants are then clustered into five main segments that take into account how the occupants
MOA model will respond to interventions. In the case study, a survey is designed to test the hypotheses and their
Structural equation modeling validity using descriptive statistical analysis and structural equation modeling. The proposed frame-
work is expected to provide decision-makers with useful information to design effective energy saving
interventions to reduce overall energy consumption in buildings.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction to enable voluntary behavior changes, such as: information distri-

bution outlets (e.g., posters, videos, brochures) [12–14]; feedback
The energy consumption of existing buildings in the United (e.g., comparing current energy use with historical use that pro-
States accounts for 41% of the national energy use [1], and more vides consumers with personalized evaluation and a means to
than 80% of this energy is consumed during the occupancy phase monitor progress) [3,15–17]; peer-comparison (e.g., allowing occu-
of the building [2]. Several research studies emphasized the sig- pants to acknowledge their energy consumption compared to their
nificant role that occupants can play to reduce energy use and peers) [18–20]; incentives (e.g., monetary incentives); and pledging
achieve significant energy savings in buildings (e.g., [3–6]). Other campaigns to encourage energy conservation behaviors [21–24].
studies highlighted the need to carefully analyze occupants’ energy On the other hand, involuntary behavior changes can include penal-
use characteristics, and corresponding effect on their energy use ties that consist of negative consequences, which discourage an
behaviors to achieve the anticipated energy reductions [7,8]. unfavorable behavior; and technological tools and systems (e.g.,
One approach is to incorporate occupancy-focused interven- installing energy-efficient insulation materials) that solve prob-
tions in the building’s energy management program [9,10]. These lems without any human involvement (e.g., occupancy and light
intervention strategies can be designed to invoke either voluntary sensors) [25,26]. In general, interventions at education and persua-
or involuntary behavior changes in occupants’ energy consump- sion level usually deal with posters, emails, or small amount cash
tion (see Fig. 1). Voluntary behavior changes can include education incentives which can be implemented at low cost. Penalties can
to teach and create awareness about benefits of a particular incur extra administrative cost of regulations and sanctions while
behavior [11]; and persuasion that offers reinforcing incen- technology intervention may require retrofit of existing buildings
tives/consequences to invite voluntary behavior changes at low or installing new equipment. The economic and environmental
economic costs. Prior studies have explored different approaches costs associated with these intervention methods increase across
the spectrum from voluntary to involuntary methods.
Most of the studies listed above evaluated the effectiveness of
∗ Corresponding author. the interventions methods on a selected sample of building occu-
E-mail addresses: [email protected] (D. Li), [email protected] pants, and typically report different success rates. However, the
(C.C. Menassa), [email protected] (A. Karatas).

http://dx.doi.org/10.1016/j.erss.2016.11.008
2214-6296/© 2016 Elsevier Ltd. All rights reserved.
98 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Fig. 1. Multi-Level Building Energy Use Intervention Strategies.

main limitations of these studies are as follows. First, none of these a positive impact on consumers’ own behavior and emphasized
studies investigated what occupancy characteristics lead to the suc- the need for providing feedback among energy consumers. Hine
cess of these interventions as measured by initial and sustained et al. [33] synthesized 128 studies on pro-environmental behav-
energy use behaviors over time. This is an important gap since sev- iors and identified several influential variables such as attitudes,
eral studies have shown that there is often a “rebound effect” where knowledge of issues, verbal commitment. Hong [31] showed that
occupants’ energy consumption tends to increase after the inter- occupants who are proactive in saving energy consume up to 50%
vention is removed [20,27,28]. Second, the existing studies assume less energy compared to occupants who do not care about energy
that occupants react in the same way to intervention strategies. use. Moreover, Klein et al. [3] highlighted that occupant engage-
None of them addresses the different characteristics of occupants ment in building energy reduction strategies is critical and can be
that play a key mediation role to effectively achieve the interven- achieved through interventions of informed feedback and sugges-
tion objectives. For example, occupants with extreme energy use tions. They investigated the impact of occupant preferences and
behaviors will react differently to interventions as compared to behaviors on building energy use and occupants’ comfort level, and
others with moderate energy use patterns. Education might be suf- achieved 12% reduction in energy consumption and a 5% improve-
ficient to reduce the energy use of the latter but might need to be ment in occupants’ comfort of a test bed building.
supplemented with interventions from higher levels shown in Fig. 1 Despite the significant contributions of these studies to analyz-
to affect the extreme energy users [29]. Therefore, decision-makers ing the impact of various energy use behaviors of occupants on
are challenged with designing and delivering these interventions the building’s energy consumption, they didn’t provide a concep-
effectively and efficiently due to high uncertainty in predictions of tual framework that can be implemented by decision-makers to
occupants’ energy use characteristics, their corresponding behav- fundamentally understand: what major influential factors affect
iors and the resulting energy savings. This uncertainty has mainly occupants’ energy use characteristics in a particular building,
emerged due to the different characteristics of occupants which and how the latter influence their energy use behavior. Such
vary based on several factors including occupants’ different indi- an approach can provide decision-makers (e.g., facility man-
vidual energy consumption rates (i.e., energy use intensity), and agers) with information to better design and deliver effective
dynamic changes in their energy consumption behaviors over- occupancy-focused interventions tailored to occupants’ energy use
time [29–31]. To overcome these challenges, there is a need to characteristics. To address this research gap, this study extends
understand which influential factors affect occupants’ energy use the literature on occupants’ behavior and focuses on develop-
characteristics and their behaviors in buildings. These influential ing an integrated conceptual framework which can be adopted in
factors play an important role in determining which type of inter- different types of buildings, locations and weather zones to deter-
ventions will be more effective and help guide the implementation mine the influential factors of occupants’ energy use behavior. The
process of interventions. work proposed in this paper addresses several research questions
A number of research studies have noticed the importance identified by Sovacool [37]. First, it presents a conceptual frame-
of various characteristics and behaviors of occupants on build- work that incorporates human-centered research methods such as
ing energy consumption and investigated them using qualitative survey with quantitative approaches including statistical analysis,
and quantitative approaches [3,20,29–36]. In these studies, occu- and agent based simulation to understand the role of occupants
pants’ energy use characteristics are widely described to include in building energy use. Second, this study adopts an interdisci-
occupants’ energy use intensity, occupants’ attitudes and actions plinary approach from social science and applies it in the domain
they perform (or do not perform) to influence the level of energy of energy to identify the influential factors of energy use behavior.
consumption; as well as, occupants’ connections in a given social The compatibility of the interdisciplinary method in buildings is
network and how those might affect their attitudes and actions. demonstrated through a case study. Third, the work presented in
For example, Azar and Menassa [29] focus on extreme energy use this paper draws insights from social psychology to measure occu-
behaviors, and their impact on the overall energy consumption lev- pant’s energy use characteristics and behavioral change using three
els that reflect the dynamic aspects of occupancy. The results show constructs: motivation, opportunity and ability. Finally, this study
that extreme energy users have an important negative impact on illustrates what type of information and feedback should be used
moderate peers, which can contribute to revoke the benefits of in the intervention strategies to promote energy saving behaviors
energy saving interventions. These extreme occupants are charac- based on the results of a given building.
terized by a very narrow energy variability (Var), which represents
the occupants with strong energy use habits and are harder to 2. Objective
influence than occupants with more flexible energy use behav-
ior. Peschiera et al. [20] assessed the behavioral impact of building This paper presents a framework that is capable of identify-
occupants when provided with personal energy use data contex- ing influential factors of occupants’ energy use characteristics and
tualized with different social frames of reference. This study found their corresponding energy use behaviors in buildings to enhance
that occupants’ energy use behaviors can be changed by occupancy- the effectiveness of energy reduction strategies. Therefore, the four
focused interventions such as providing peer network utilization main objectives of this study are to: (1) identify major influential
and electricity use information, which resulted in a reduction of up components of occupants’ energy use characteristics; (2) develop
to 17% in buildings energy consumption. Similarly, Pieters et al. [32] a model that establishes the relationship between identified occu-
argued that other households’ pro-environmental behaviors have pants’ energy use characteristics and their energy behaviors; (3)
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 99

illustrate the capabilities of the developed model through a case Fishbein et al. [59] identified strong intention, no environmental
study of an existing building; and (4) discuss the potential use constraints, and necessary skills as the prerequisites to generate a
of the results in an agent based simulation environment [9,29] to behavior. Michie et al. [60] reviewed existing literatures of behav-
determine energy reduction potential from peer pressure. ior change interventions and established a Behavior Change Wheel
(BCW) framework to guide the design of interventions. The BCW
3. Background aims to strengthen the identified weak components in motiva-
tion, opportunity, and capability to enhance the effectiveness of
To achieve and improve energy reductions in building sec- behavior change interventions. In Fogg’s behavior model [61], an
tor, several studies proposed technical solutions to reduce energy effective persuasive intervention should possess three driving fac-
demand by different building systems (e.g., lighting) [38,39]. How- tors: motivation, ability and trigger in order to elicit a targeted
ever, these solutions are characterized by highly uncertain energy behavior. The behavior is only performed when people have suf-
reduction benefits [30,40–42]; relatively high initial costs [43–45]; ficient motivation, required ability and appropriate triggers. Some
lack of information about the existing building systems mak- of these models and principles have been examined and introduced
ing integration of new technologies a daunting task [46,47]; and to building domain to design and implement energy conservation
low taxes and energy prices [48,49], to name a few. In addition, strategies [63–67]. For example, in order to assess the compre-
decision-makers who focus only on technological solutions often hensiveness of BCW framework in the domain of energy behavior,
ignore the significant impact that occupants have on the energy Wilson and Marselle [63] mapped BCW framework with four
use in buildings [30,50,51]. Due to high energy saving potential, low energy-behavior change guidance documents. The result indicates
or even no initial cost, and applicability to both new and existing energy-relevant behavioral determinants can be generalized into
buildings, many researchers are focusing on studying and modeling motivation, opportunity and ability (MOA) categories and the most
occupants’ behaviors, and investigating methods that effectively frequently occurring determinants in the documents are reflective
engage occupants in the energy reduction strategies. motivation, psychological capability, physical and social opportu-
Azar and Menassa [30] conducted an analysis using energy nities. Other studies like Petkov et al. [64] and Bang et al. [66]
simulation to study the impact of nine occupancy related actions incorporated the persuasive techniques discussed in Fogg’s model
(e.g., after-hours equipment and light use) on energy use in com- and developed applications which aim to raise awareness of energy
mercial buildings of different sizes and located in 10 different related issues and encourage energy efficient lifestyle.
weather climates across the US. The results from individual impacts On the other hand, researchers also applied social psychology
are found to be as high as 30%. The combined effect of some approaches for understanding and promoting pro-environmental
of these actions increases the building’s energy demand by over behaviors [68–73], such as Defra 4Es (enable, encourage, engage,
50%. In a subsequent study, Azar and Menassa [52] illustrated that exemplify) framework which aimed at promoting sustainable
the way occupants use energy in the building and the methods behaviors in accordance with social marketing principles [68],
that facility managers operate the building highly influence the the needs-opportunities-abilities (NOAs) framework for analyz-
energy consumption levels. They proposed a framework to collect ing the determinants of consumer’s environmental behaviors [69],
occupancy-focused data and develop energy analysis approach to taxonomy of behavior change techniques in published interven-
investigate human-focused energy conservation opportunities in tion descriptions and manuals [70]. Specifically, ThØgersen [71]
commercial buildings. A case study highlighted inefficiencies in adopted the MOA model to analyze consumer behaviors and sug-
the operation of the heating, ventilation, and air conditioning sys- gested opportunity and ability as the facilitating conditions of
tem (HVAC) increase energy consumption level by as much as 40%. motivation.
Other studies such as Chen et al. [53]; Jain et al. [54]; Moezzi et al. However, the aforementioned models and frameworks only
[4]; Sanchez et al. [5]; Webber et al. [55] support these results, and used generic behavioral models developed in the field of psy-
emphasize that significant energy reductions can only be achieved chology. None of them demonstrate how to measure occupants’
if building occupants are engaged in the process. These studies characteristics in energy use domain and analyze their impact on
have significant contributions to the body of knowledge on ana- occupants’ behaviors. Therefore, to fill this gap and identify the
lyzing the impact of various energy use behaviors of occupants influential factors of occupants’ energy use behavior in a partic-
on building energy consumption. However, none of these studies ular building, we propose to use the adapted MOA model which
acknowledged how occupants’ energy use behaviors are affected maps all the possible factors with motivation, opportunity and
by their energy use characteristics. All occupants in these studies ability defined in the energy use domain. To achieve this, we con-
are assumed to react in the same way to the proposed interven- duct a thorough survey which measures all the factors that may
tion implying that they have the same energy use characteristics affect occupant’s energy use behavior and then adopt the MOA
resulting in short lived benefits to the interventions implemented model which has been widely applied in consumer science to study
[11]. the influence of motivation, opportunity, and ability (MOA) mea-
In order to measure occupants’ energy use characteristics and sures on purchasing behaviors. This research approach draws on
corresponding behaviors, this study adopts a multidisciplinary the analogy between energy intervention strategies and marketing
method combining social science and marketing to develop a model approaches in consumer science. More specifically, the segmen-
that allows decision-makers to understand the building occupants’ tation of consumers based on their MOA measures is adopted
energy use characteristics prior to implementing the intervention. for this research. This model has traditionally been used to ana-
The model is based on the consumer segmentation approach from lyze consumers’ attention and comprehension processes to brand
consumer and social marketing [56–58]. The applications of the information, and other significant factors motivating consumers to
model in social science and marketing field, and its adoption to purchase certain products on a regular basis [74–80].
energy use characteristic context are discussed in detail in the fol- Several studies emphasized that customers’ MOA characteristics
lowing sections. play a significant role in their purchasing behaviors. For exam-
ple, Hastak et al. [81] highlighted the importance of MOA model
4. Methodology to determine the communication effectiveness of ads in consumer
research. This study found that consumers with loyalty to a partic-
Several studies have proposed behavior models to understand ular product have higher MOA levels that help facilitate adoption
the factors that influence human behavior [34,59–62]. For example, of this product. MacInnis et al. [77] studied MOA model to investi-
100 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Table 1
Application of the MOA Model to the Study of Energy Use Characteristics.

Psychological definition MOA model in consumer science Proposed framework with adapted
[60,83,84,92] [77,79,80,85], etc. MOA model in energy efficiency

Motivation is defined as the brain Motivation is defined as consumers’ Motivation measures an occupant’s
processes that energize and direct interest, desire and readiness to needs, goals and values (self-related
human behavior toward goals engage themselves in processing the knowledge), and the level of
brand information in an ad and their involvement with energy use
willingness to perform purchasing information, whether occupants’ are
behavior. E.g., high motivation concern about their personal energy
consumer is willing to pay his/her consumption and looking for ways to
attention to the advertised message save energy from various sources
Opportunity is defined as external Opportunity is defined as the Opportunity measures an occupant’s
factors (lie outside the individual) environment factors (e.g., distraction, availability and accessibility to the
that make the behavior possible or lack of conditions, limited exposure energy saving information and energy
prompt it time) which affect consumer’s control system, as well as some
attention to brand information in an environmental and interpersonal
ad. E.g., Consumers who are unfamiliar factors that may affect occupants
with a brand may choose other processing these information in his/her
competitive products environment (both physical and social
opportunity)
Ability is defined as necessary Ability is defined as consumers’ Ability is a knowledge-based measure
psychological and physical resources, skills, or proficiencies in which affects how an occupant
capabilities to make an outcome interpreting brand information and interpret and process the energy
happen performing the purchasing behavior. saving information. Ability implicates
E.g., high ability consumer is able to an occupant’s prior knowledge about
search desired items online and energy use, its impact and
compared with different brands consequences, as well as knowledge
about possible saving strategies

gate the influential factors such as the extent of brand information highlighted that motivation independently influences consumers’
processing from ads. Their model proposed that consumers’ MOA preventive health behaviors, and moderates the impact of ability
levels have major impact on the brand information processing stage and opportunity on adopting desired behaviors through encour-
during and/or after exposure to advertisements. Additionally, Bigné aging consumers to put their knowledge, skills, or resources into
et al. [82] implemented MOA model to analyze the major drivers practice.
of online airline ticket purchases, and identify the effectiveness of In this research, we define motivation as an occupant’s readi-
perceived channel benefits for consumers in using the Internet to ness, willingness, interest and desire to process energy saving
purchase airline tickets. This study reported that MOA model helps information provided through any of the intervention strategies
to explain 55% of the variations in adopting desired behaviors, and shown in Fig. 1, and subsequently adopt the stipulated energy sav-
is an effective methodology to predict consumers’ ticket purchasing ing behaviors. Thus, the motivation (M) level of occupants measures
intentions. their perceived personal relevance and the level of involvements
In this study, we draw on the analogy between MOA char- with the information presented in the energy reduction strategies
acteristics of customers to process brand information and pick (see Table 2). For example, occupants with high motivation level
up certain products in marketing, and MOA levels of building can volunteer to attend workshops or receive emails about energy
occupants to interpret energy reduction intervention strategies saving tips. As shown in Table 2, motivation metrics are divided
and adopt energy saving behaviors (e.g., adjusting the thermostat into internal and external stimuli indicating that energy related
to lower temperatures when leaving the office in heating sea- intervention strategies should focus on improving these metrics.
son). The psychological definition of MOA characteristics, the MOA For example, improving awareness of energy use implications and
model in consumer science and its adapted interpretation in energy desire to receive additional information and guidelines on how to
efficiency are detailed in Table 1 and additionally described in sub- reduce one’s own impact.
sequent paragraphs.

4.1. Introduction to the motivation, opportunity and ability 4.1.2. Opportunity (O)
characteristics The psychological definition of opportunity is the external fac-
tors that make the behavior possible or prompt it [60,83]. Similarly,
4.1.1. Motivation (M) in consumer science it is defined as environment factors (e.g., expo-
The psychological definition of motivation is the brain processes sure time to ads) that are not in the control of consumers to enable
that energize and direct human behavior toward goals [60,83,84]. desired actions [77,79,82,90] (See Table 1). Opportunity level is
More specifically, in consumer science motivation is defined as directly related to the immediate environment of the people and
a goal-directed arousal to engage consumers in desired behavior how that affects the availability, accessibility, and time allocated
to process brand information in the advertisement and perform for comprehension of the brand information [79,80]. Other studies
purchasing behavior [77,79,80,85–87] (See Table 1). Motivation described opportunity level as the extent to which circumstances
measures perceived personal relevance of people and their level evidenced during advertisement exposure are favorable for brand
of involvement and interest with a particularly provided piece of processing [77,79,91]. For example, Govindaraju et al. [91] stud-
information [86–88]. For example, Parra-Lopez et al. [89] analyzed ied key drivers for physicians to adopt electronic medical records
key factors in human intentions to use social media to organize (EMR), and extended “opportunity” context to physicians’ “access
holiday travel before and during the holiday itself. This study high- to EMR system” and “access to information” which refer to the
lighted that consumer’s motivation is based on functional, social physicians’ opportunity to get in contact with any information
and hedonic benefits of social media. Moorman and Matulich [85] media and other sources.
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 101

Table 2
Metrics for Measuring Occupants’ MOA for Energy Conservation.

Metrics of Construct Measures

Measures of Motivation (M) Assess self-awareness about importance of Energy Use Knowledge
Self-related Knowledge Measure desire to receive Energy Use Knowledge
(Internal Stimuli) Detect norms of avoiding Energy Use Knowledge (e.g., not interested in attending workshops or receiving
emails)
• Needs
• Goals
• Values
Energy Use Knowledge
(External Stimuli)

• Level of Energy Use

• Impact and Consequences

Measures of Opportunity (O) Determine number of times:

Ease of implementation Accessible and easy to use controls (e.g., thermostat, lighting, shading)
Amount of Information Attend awareness seminars
Information Format Read information on general advertisement boards (self-reported)
Modality Read emails (ask for response with a blank email)
Rate of Exposure to Discuss with peers (self-reported)
Information
Measures of Ability (A) Measure extent conservation strategies are used (e.g., estimate number of times consciously turn lights
Energy Use Prior Knowledge off at end of the day)
Measure subjective knowledge of energy use relative to average person
• Impact Measure actual knowledge (i.e., factual information):
• Consequences
• Conservation Strategies • Terminology
• Possible impacts/consequences
• Criteria to evaluate impacts/consequences
• Perceived effectiveness of intervention strategy to reduce impacts/consequences

In this research, opportunity (O), refers to the surrounding retrieved in a given situation (e.g., occupants turn lights off before
environmental factors influencing occupants’ attention and com- leaving their offices).
prehension processes in adopting energy saving behaviors. As
shown in Table 2, when occupants have easily accessible building
4.2. Framework for measuring MOA levels of building occupants
controls, opportunity metrics can be improved during an inter-
vention by focusing on the amount, format, modality and rate
Previous studies (e.g., medical field, ticket purchasing website)
of exposure to the information. Thus, opportunity measures how
found out that motivation is directly associated with most behav-
favorable conditions and limited time of exposure affect the occu-
iors [78,82]. However, opportunity and ability affect behaviors only
pant’s attention to information being presented in the energy
when motivation is present, which means these two characteris-
intervention strategies.
tics moderate the impact of motivation on behaviors. Therefore, in
this research we propose motivation as a precondition of successful
implementation of energy saving behaviors with opportunity and
ability as moderating factors, and design the framework described
4.1.3. Ability (A)
in following section accordingly. The proposed framework in this
The psychological definition of ability is the necessary capa-
study is developed by stating a set of research hypotheses and their
bilities (psychological and physical) to make an outcome happen
relevant measures to investigate occupants’ energy use character-
[60,83,92]. In consumer science, ability definition is modified to
istics through assessing their MOA level on adopting energy saving
focus more on psychological aspects and it is defined as consumers’
behaviors. These hypotheses are designed based on the extended
self-perception knowledge capacity of the brand information, and
context of MOA levels of occupants in energy use characteristics,
how they interpret this information to create new knowledge struc-
and incorporated with a set of measures that is identified based on
tures [77,80,82,89] (See Table 1). Ability level is largely dependent
a comprehensive literature review, as shown in Fig. 2. These mea-
on the consumer’s prior knowledge about brand information typi-
sures are utilized to demonstrate the link among each motivation,
cally acquired through experience; as well as, the consumer’s skills
opportunity, and ability levels of occupants, their related research
in interpreting brand information in an advertisement [79,80]. For
hypotheses, and their intended energy use behaviors.
example, Bigné et al. [82] defined ability level as individuals’ per-
ception of their capacity to search for information about flights on
the Internet and to carry out online purchases of airline tickets. 4.2.1. Identifying measures for occupants’ motivation
Results show that customers’ Internet ability positively influences characteristic
their intentions to purchase online airline tickets. As mentioned earlier, motivation level (M) refers to a particular
In this research, ability (A) level of occupants measures a given occupant’s perceived personal relevance in terms of needs, goals
occupant’s proficiencies in interpreting energy use knowledge. This and values, and the level of involvement with the information (e.g.,
ability (A) (see Table 2) is largely dependent on the occupant’s external stimuli) presented in the energy intervention strategy.
prior knowledge about energy use, its impact and consequences, Therefore, occupants’ concern and desire of energy conservation
as well as knowledge about possible conservation strategies. The are investigated as the measure of their motivation. Motivation
type and quality of this pre-existing knowledge will in turn deter- level independently affects occupants’ energy use behaviors, how
mine if energy use information can be cognitively and immediately often they look for ways to conserve energy at their offices, and
102 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Fig. 2. Measures for MOA Level of Occupants’ Energy-Use Behavior.

which factors are important for them to conserve energy, as shown H3. When energy conservation motivation is high, occupants who
in Fig. 2. Accordingly, the related hypothesis is stated as follows: are satisfied with their indoor environment will perform more
energy conservation behaviors.
H1. Occupants with higher energy conservation motivation levels
H4. When energy conservation motivation is high, occupants with
will perform more energy conservation behaviors than occupants
more exposure to information and peer-pressure about the envi-
with lower energy conservation motivation levels.
ronmental impacts of their energy use behaviors will perform more
energy conservation behaviors.

4.2.2. Identifying measures for occupants’ opportunity

4.2.3. Identifying measures for occupants’ ability characteristic
characteristic
In this research, ability level measures each occupant’s profi-
In this research, occupants’ opportunity (O) level is directly
ciency in interpreting energy use knowledge. This ability is largely
related to their immediate environment and how that affects the
dependent on two major factors: occupant’s perception on their
availability and accessibility for comprehension of the energy use
energy consumption level, and occupants’ prior knowledge about
knowledge. Adopting Moorman and Matulich [85] approach, we
energy use facts. A set of studies has shown that people need to
assume that occupants’ opportunity level moderates the effect of
have sufficient ability (e.g., self-efficacy) before they can actively
motivation on energy saving behaviors. For example, highly moti-
care enough to take environmentally responsible actions that bene-
vated occupants are not able to present energy saving behaviors
fit others [94–97]. These studies highlighted that occupants’ ability
in their offices such as adjusting thermostat control when leaving
(A) level also moderates the effect of motivation (M) on adopting
the office, if they do not have any control on their room’s thermo-
certain behaviors. Abrahamse and Steg [98] concluded that occu-
stat. Studies have shown that people have more tendency to make
pants with higher perceived energy conservation knowledge (e.g.,
permanent changes in their energy use behaviors when they have
“I know how to reduce cooling load in summer”) and knowledge
available resources and if the new behaviors are easy and conve-
on energy consumption facts (e.g., what a kWh unit means) are
nient to perform [93]. Therefore, opportunity level of occupants in
more likely to save energy than occupants with lower energy use
this study is measured by investigating (see Fig. 2): (i) availabil-
knowledge. On the other hand, previous studies show that higher
ity of energy conservation control systems (e.g., if the occupant
ability levels may reduce consumers’ acquisition of information if
has any individual control on the indoor climate control system);
they feel less need for more information [78]. Therefore, it is also
(ii) office environment satisfaction level (e.g., overall quality of
important to ensure that knowledgeable occupants are also highly
artificial lighting in the office); and (iii) exposure to information
motivated to maintain their energy conservation behaviors. Based
and peer-pressure about environmental concerns through inter-
on Abrahamse and Steg [98], occupant’s knowledge-based ability
ventions (e.g., having energy conservation information available
level is measured by (i) perceived energy conservation knowledge
for occupants on bulletin boards and often discuss environmental
(we assume their self-claimed knowledge is correct), and (ii) level
conservation strategies with colleagues). Accordingly, the corre-
of knowledge on energy consumption facts, as shown in Fig. 2.
sponding hypotheses are stated as follows:
Accordingly, related hypotheses are stated as follows:

H2. When energy conservation motivation is high, occupants with H5. When energy conservation motivation is high, occupants
higher control over their indoor environment will perform more with higher perceived energy conservation knowledge will perform
energy conservation behaviors. more energy conservation behaviors.
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 103

Fig. 3. Framework Implementation.

H6. When energy conservation motivation is high, occupants with and peer-pressure among co-workers (or neighbors) (e.g., my close
better knowledge of energy consumption facts will perform more friends in the company always use strategies to conserve energy).
energy conservation behaviors. Finally, to test and validate ability level of occupants, the last two
hypotheses, Hypothesis 5 (i.e., H5), and Hypothesis 6 (i.e., H6), are
tested through measures presented in Fig. 2. H5 studies if the occu-
4.3. Framework implementation
pants’ perceived energy conservation knowledge level would affect
their energy conservation behaviors when they are motivated to
The proposed framework is implemented in four phases (as
conserve energy. H5 is tested using the measure of occupants’ self-
shown in Fig. 3): (1) survey phase that involved designing an online
assessed knowledge level (e.g., I know methods to reduce heating
survey to test the stated hypotheses; (2) reliability analysis phase
load in my office). On the other hand, H6 determines if occupants’
that analyzes results from survey for integrating the measures of
knowledge level of energy consumption facts would have impact
stated hypotheses in the same construct; (3) structural equation
on energy conservation behaviors when energy motivation is high.
modeling (SEM) phase that investigates the relations among moti-
The measures for H6 are designed as six test questions asking occu-
vation, opportunity, ability, and intentional energy use behaviors
pants about their knowledge on energy consumption facts (e.g.,
of occupants; and (4) data output phase that proposes energy use
which lighting choice saves the most energy assuming the same
reduction strategies for the building occupants given their energy
delivered amount of light?).
use characteristics and predicts the expected energy savings. The
following paragraphs provide detailed explanation of each phase
4.3.2. Reliability analysis phase
of the research.
Reliability analysis needs to be conducted where the survey data
is analyzed to check the internal consistency of multiple measures.
4.3.1. Survey design phase Through reliability analysis, highly correlated questions are com-
An online survey is developed to collect data to test the stated bined as one composite question to reduce multicollinearity and
hypotheses, and identify occupants’ motivation, opportunity and make the multi-item integration intensity scale to be unidimen-
ability level and their corresponding behaviors. The survey is sional [99,100]. In this research and as illustrated in the Case Study
designed to be flexible for use both in residential and commercial section, the results for the reliability analysis are assessed based on
building settings. To test and validate Hypothesis 1 (i.e., H1), two the recommended threshold of: Cronbach ␣ with a value of ␣ = 0.70
measures were identified as: (1) occupants’ concern level on their [101], and inter-item correlation with a value of 0.30 [102]. Ques-
personal energy consumption (e.g., how often are you concerned tions with Cronbach ␣ greater than 0.70 and inter-item correlation
about your personal energy consumption at your office?), and (2) greater than 0.30 are combined into a single construct. For example,
factors deciding whether to conserve energy (e.g., how important plug load control can combine four correlated equipment questions
are the following factors to you in deciding whether to conserve (i.e., printer, fax, fridge and microwave oven) into one measure of
energy?). opportunity (O).
Hypothesis 2 (i.e., H2), Hypothesis 3 (i.e., H3), and Hypothesis 4
(i.e., H3) are tested through measures presented in Fig. 2. H2 mea- 4.3.3. Structural equation modeling phase
sures if occupants’ energy use control levels have an impact on Structural equation modeling (SEM) phase is implemented to
their energy conservation behaviors when they have motivation to test and validate the stated hypotheses. SEM is a confirmatory mul-
conserve energy. In the survey, measures of H2 are identified as tivariate analysis methodology for hypotheses testing, which has a
the availability of lighting control, availability of thermostat con- capability of constructing variables that may explain major part
trol, and availability of office equipment (e.g., printer) plug load of the unobserved heterogeneity in the model [103]. In social sci-
control for the occupants. H3 predicts that office (or residence) ence and marketing field, several research studies implemented
environment satisfaction levels have an effect on occupants’ energy SEM models to investigate the influential factors on consumers’
conservation behaviors in the presence of energy conservation characteristics and their behaviors [82,91,104–107]. Childers et al.
motivation. The measures of H3 are designed as occupants’ sat- [106] studied the dominant factors in customers’ online retail shop-
isfaction with lighting quality level (e.g., how would you describe ping behavior using SEM. Results indicated that hedonic aspects of
the quality of artificial lighting in your typical work area?), thermal the new media play at least an equal role as instrumental aspects
comfort level, and indoor air quality level in their offices. H4 studies to predict online attitudes. Van der Heijden et al. [107] used the
whether occupants exposed to environmental conservation infor- same method to explore factors that influence consumer’s pur-
mation changes their energy conservation behaviors, when they chasing behavior at an electronic commerce website and found out
are also motivated to conserve energy. This hypothesis is tested a strong positive relation exists between attitude towards online
using the measures of occupants’ exposure level to environmental purchasing and corresponding purchasing intention. This research
concern through company’s (or landlord) actions (e.g., my company also suggested that perceived risk and ease of use are antecedents
provides all employers with strategies to help us conserve energy), of attitude towards online purchasing.
104 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Table 3
Measures for Occupants’ MOA through Reliability Analysis.

Measures that can be combined into single construct

Construct Highly Correlated Measures Cronbach ␣ Single Construct

Opportunity Availability of the office equipment as follows: 0.812 Plug load

(Control systems)
(1) printer
(2) fax
(3) fridge
(4) microwave oven

Opportunity My company performs the actions as follows: 0.868 Company’s actions

(Expose to information)
(1) taking energy conservation very seriously
(2) encouraging all staff to conserve energy
(3) providing all staff with strategies to conserve energy

Opportunity I would feel comfortable explaining to (1) my close 0.839 Peer pressure
(Expose to peer pressure) friends; (2) other colleagues in the company how they can
conserve energy.
My (3) close friends; (4) other colleagues in the company
always use strategies to conserve energy.
Motivation (1) How concerned are you about your personal energy 0.760 Motivation
consumption at your office?
(2) How often do you look for ways to conserve energy at
your office?
(3) How important are the following factors to you in
deciding whether to conserve energy? (e.g., it is
morally the right thing to do)

Ability I know methods to reduce the power load as follows: 0.861 Perceived energy conservation
(Perceived energy conservation knowledge
knowledge) (1) lighting
(2) cooling
(3) heating
(4) plug load

Measures that exist individually

MOA Category Measure Description Single Construct

Opportunity Availability of the lighting control Lighting control

(Control systems) Availability of the thermostat control Thermostat control

Opportunity How do you describe the quality of artificial light in your Lighting quality
(Environment work area?
satisfaction) How do you describe the quality of thermal comfort in Thermal comfort
your work area?
How do you describe the indoor air quality in your work Indoor air quality
area?

Ability(Knowledge on energy consumption facts) Questions aiming to test occupants’ knowledge about Knowledge on energy consumption facts
energy consumption facts (e.g., Which of the following
energy resources is not renewable?)

Measures that are removed due to non-applicability

MOA Category Measure Description Removed Construct

Opportunity Availability of the task lighting Control of task lighting

(Control systems) Availability of the space heater Control of space heater
Availability of the coffee machine Control of coffee machine

Opportunity How do you describe the quality of natural lighting in your Quality level of natural lighting
(Environment work area?
satisfaction) How do you describe the glare comfort in your work area? Quality of glare comfort

In this research, SEM methodology is implemented using Stata 4.3.4. Results and energy implication phase
[114] where exogenous variables (e.g., quality of artificial lighting This phase helps decision-makers analyze the identified influ-
in working zone) are computed as factors affecting endogenous ential factors on occupants’ energy use characteristics, and their
outcome (i.e., intended energy use behavior of occupants like turn- behaviors to design effective occupancy-focused intervention
ing off the monitor when not in use). A detailed description of the strategies. The proposed motivation/opportunity/ability (MOA)
setup of SEM model and how the resulting data are interpreted is model can monitor the effectiveness of the intervention strategies
explained in the Case Study Section.
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 105

Fig. 4. Occupants’ (a) Motivation Level; (b) Opportunity Level; (c) Ability Level.

(e.g., effect of peer pressure, resulted energy savings). By comparing formance of the facility, and comfort of the building occupants. This
occupants’ MOA level at different stages throughout the interven- BAS system provides occupants with different levels of occupancy
tion, decision-makers can identify the weak component of MOA in control over the building environment. For example, some occu-
occupants’ characteristics and modify their interventions accord- pants especially in single occupancy rooms are able to control the
ingly over time. thermostat settings, shading and lighting. On the other hand, occu-
To achieve this objective, a k-means clustering analysis is con- pants in multi-occupancy rooms do not have much control, which
ducted using Matlab to organize occupants in groups with similar may lead to various energy use behaviors among the occupants in
MOA characteristics [108]. According to the individual MOA level the same building.
and the MOA distribution within the building clusters, an occupant The demographic information of the respondents is as follows:
can be categorized as prone, mildly unable, unable, mildly resis- (1) 65% are female and 35% are male; (2) 9% are between the age
tant or resistant to change the energy use behavior. In this study, of 20–30 years, 43% are between the age of 30–49, and 48% are
the results of the MOA model are integrated with the Agent-Based older than 49 years; and (3) 8% have high school degree; 26% have
Model (ABM) which studies how occupants interact and respond to college level degree; 66% have a graduate level degree of Master’s
energy saving interventions to provide an initial prediction of the (e.g., MA, MS, MENG) and/or PhD (e.g., PhD, EdC). The survey results
effectiveness of the chosen intervention. Details about this model were then reviewed for completeness resulting in a total of 130 (i.e.,
and its functionalities can be found in Azar and Menassa [9] and 73% of the total surveys filled out) fully completed surveys which
Azar and Menassa [29]. were subsequently used for the reliability and SEM analysis.
At the agent initialization phase, each agent (or occupant in our Through reliability analysis, some highly correlated measures
context) is associated with two variables: (1) energy intensity (EI) in the survey are combined into simple constructs (see Table 3) to
in kWh/person, which defines the energy consumption level of an determine occupants’ MOA level. As shown in Table 3, the Cronbach
agent, and (2) energy variability (Var), which refers to an agent’s coefficient ␣ for some listed measures is greater than the recom-
openness to adopt new energy use characteristics [29]. Data about mended cut-off value of 0.7 and thus can be combined into a single
the occupants’ energy intensity is obtained from building energy construct. Other measures with low Cronbach ␣ (e.g., availability
use data or from historical databases such as the Commercial Build- of the lighting control) can’t be combined together and are used
ing Energy Consumption Survey (CBECS) published by the Energy directly in the SEM model. Some measures are reported as “not
Information Administration [109]. The average energy consump- applicable” by all the respondents (e.g., availability of the space
tion can be treated as the median EI, and occupants who are well heater) and thus are removed from the SEM model.
performed in energy use tend to have a low EI and vice-versa. Occu- The distribution of occupants’ MOA level and behaviors are
pants’ MOA characteristics (prone, mildly unable, unable, mildly shown in Figs. 4 and 5 respectively. Fig. 4 shows that around 30% of
resistant, resistant) can be mapped with the Var parameter, where the occupants in this office building (36 out of 130) are highly moti-
occupants who are prone to change behaviors tend to have a high vated, and around 14% of them (18 out of 130) possess high ability
Var (i.e., flexible habits) while those with resistant MOA charac- level. However, the opportunity level is relatively low (77 out of
teristics tend to have a very low Var (i.e., rigid habits). As a result, 130 possess low or medium-low opportunity level) compared to
the agents which are initialized by MOA level can reflect the real the other two characteristics, with only 5 occupants having a very
occupancy characteristics in the building. The implications of this high opportunity level. This might be a direct result of the limited
approach and its ability to predict energy savings given an inter- ability placed on the occupants by the existing BAS, which reduces
vention strategy are illustrated in the Case Study Section. the occupants’ potential to contribute to energy saving.
Fig. 5 shows the number of occupants who perform well for
different energy use behaviors considered in the study. Occupants
5. Case study report several poor energy use behaviors, such as Behavior 4 (i.e.,
B4: turn off the monitor when not in use), Behavior 5 (i.e., B5: turn
A total of 177 occupants in a 32-story building in Chicago, IL, off the computer when not in use), Behavior 6 (i.e., B6: adjust shades
responded to the survey which was distributed online. The case to reduce glare), Behavior 7 (i.e., B7: adjust shades to increase day
study building is a multifunctional building of a university which lighting), Behavior 8 (i.e., B8: adjust shades to reduce heat from
contains classrooms, student dorms, administrative offices, etc. the sun), and Behavior 9 (i.e., B9: turn off the light when there is
Participants are mainly from offices located on floors 6–10 repre- enough daylight). For B4, B5 and B9, occupants fail to show better
senting administrative staff in a typical office environment. In total, energy use behaviors even though most of them have control to
205 people routinely occupy the surveyed floors (survey response the equipment (lighting and computer), that is they have a high
rate is 86%). The case study building is equipped with a build- opportunity to do these actions. The low action rate of B6, B7 and
ing automation system (BAS) with a centralized monitoring and B8 is because many occupants (52 out of 130) reported that they
control of building environment to maintain the operational per-
106 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Fig. 5. Number of Occupants Who Perform Better Energy Use Behaviors.

have no control to the shading system, which indicates a strong squared error of approximation (RMSEA) is also used to measure
correlation between opportunity and behavior response. the misfit/fit of the SEM. The RMSEA of the developed model is 0.05,
In SEM, models are illustrated in a path diagram. The boxes rep- which is equal to the ideal standard of “less than or equal to 0.05”
resent observed variables (or indicators) that are observed in the [111,112]. Also, the Normalized Chi-Squared value of the results
data. These variables are actually what are measured from par- is X 2 /df = 0.80, which is a reasonably good indicator of model fit
ticular questions on the survey (e.g., lighting quality). Circles are [113]. Therefore, the model shown in Fig. 6 was accepted as the best
unobserved latent variables that are measured by a number of fit model for this particular building.
observed variables (e.g., occupant’s exposure to information is mea- The results from SEM analysis are presented in Table 4. As
sured by company’s action and peer pressure). Single head arrows, shown in Table 4, ˇ is the coefficient of observed variable and
called paths, connect variables in the path diagram. When a path latent variable. For each change of one unit in observed variable, the
points from one variable to another, it means that the first vari- latent variable will change ˇ units while holding all other observed
able affects the second (e.g., the arrow between motivation and variables constant [114]. By examining the p value, statistically sig-
behaviors represents that motivation is a predictor of behaviors). nificant coefficient is highlighted with a label (“*” for p < 0.1, “**”
According to the measures for MOA level discussed in Fig. 2, an for p < 0.05). The subsequent subsections explain the implication of
SEM model is constructed to test the proposed hypotheses (see these results for the case study building.
Fig. 6). For example, to test H2 “occupants with higher control over
their indoor environment will perform more energy conservation 5.1. Test of the hypothesized relationships
behaviors when motivation is high”. Latent variable “Control level”
is indicated by three observed variables: lighting, thermostat, and H1 predicts that energy conservation motivation will posi-
plug load control. Then “Control level”, Motivation and Behaviors tively affect all of the intended energy conservation behaviors. The
are connected using paths to represent their proposed relationship. results show that occupants’ motivation level has positive effect
The best four occupants’ behaviors shown in Fig. 5 (i.e., Behavior on some energy use behaviors like turning off the office lights
1–4) are selected to represent the occupants’ intentional energy ˇ = 0.34, p < 0.05 , and computer monitors when not in use
 
use behaviors. However, the above-mentioned structure of SEM ˇ = 0.24, p < 0.05 . However, there is no significant correlation
model depends on the data collected from the case study build- (abbreviated as “ns”) between occupants’ motivation to conserve
ing. Several structures are evaluated in Stata to accept the model energy and occupants’ energy conservation behaviors when the
(See Fig. 6) which best fits the data. The SEM model might look office is:(i) too chilly/cold
 (e.g., wearing a thick clothes, using space
different in other buildings (e.g., M, O, and A are the distinct pre- heater) ˇ = −0.02, ns , and (ii) too warm/hot (e.g., wearing a thin
cursors of energy use behaviors and are parallel to each other) but  
layer of clothing, using an electric fan) ˇ = −0.03, ns . These find-
the framework presented in this paper can be adopted to achieve
ings indicate a mixed support for H1.
the best fit model. In this study, model fit statistics obtained from
H2 looks at whether energy use control levels affect energy
survey results indicate that R2 = 0.174. However, because people
conservation behaviors in the presence of energy conservation
are fairly unpredictable, it is difficult to only use R2 to evaluate the
motivation. The results indicate that having thermostat control to
model fit to the data [110]. Therefore, the widely adopted root mean
adjust the indoor climate conditions is negatively and significantly
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 107

Fig. 6. SEM Model for the Case Study Building.

Table 4
Results from Structural Equation Modeling (SEM) Analysis.

Outcome Coeff. (ˇ)

Behavior
Motivation → Behavior (H1)
Motivation Level → Turning off the office room lights when not in use 0.34**
Motivation Level → Turning off the office monitor when not in use 0.24**
Motivation Level → Energy conservation behavior when the office is too chilly/cold −0.02
Motivation Level → Energy conservation behavior when the office is too warm/hot −0.03

Motivation
Opportunity → Motivation
H2 → H1
Control on Lighting → Motivation Level −0.11
Control on Thermostat Settings → Motivation Level −0.32*
Control on Office Equipment Plug Load → Motivation Level 0.13
H3 → H1
Indoor Lighting Comfort Level → Motivation Level −0.06
Thermal Comfort Level → Motivation Level 0.16*
Indoor Air Quality Level → Motivation Level −0.09
H4 → H1
Exposed to Information → Motivation Level −0.01
Peer Pressure → Motivation Level 0.19*

Ability → Motivation
H5 → H1
Perceived Self-Knowledge on Energy Conservation → Motivation Level 0.23**
H6 → H1
Knowledge on Energy Use Facts → Motivation Level 0.16
*
Statistically significant (p < 0.1).
**
Statistically significant (p < 0.05).

 
correlated with occupants’ motivation ˇ = −0.32, p < 0.1 . This H3 determines if higher office environment satisfaction levels
result indicates that for this case study building, even if some would affect occupants’ energy conservation behaviors in the pres-
occupants have no control on thermostat, they still have high ence of energy conservation motivation. For example, occupants
motivations to conserve energy. Moreover, there is no significant with higher thermal comfort level tend to perform more energy
correlation between occupants control  on office equipment’s
 plug saving behaviors than occupants with lower thermal comfort level
load (e.g., printer, microwave oven) ˇ = 0.13, ns and on light- when their motivation is high. The results indicate that there is
 
ing ˇ = −0.11, ns , and occupants’ motivation to conserve energy.  between occupants’
no significant correlation  motivation and their
lighting comfort level ˇ = −0.06, ns , and indoor air quality level
These results do not support H2.  
ˇ = −0.09, ns . However, occupants’ thermal comfort is posi-
108 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Table 5

tively and significantly 
correlated with occupants’ motivation to
Centroid of Each Category for the Case Study Building.
conserve energy ˇ = 0.16, p < 0.1 . This finding can be evaluated
as providing higher thermal comfort level in the office will increase Number of Occupants Occupancy Characteristics M O A
the motivation level of occupants to conserve energy. Accordingly, 29 Prone 80 51 61
these results present a mixed support for H3. 30 Mildly Unable 56 43 32
H4 studies if higher exposure to ambient conservation infor- 29 Unable 53 40 66
mation will positively affect energy conservation behaviors in the 37 Mildly Resistant 52 61 53
5 Resistant 20 30 21
presence of energy conservation motivation. This hypothesis is
tested under two conditions, occupants’ exposure to information
through (i) company’s energy saving actions, and (ii) peer-pressure
from co-workers. The results show that peer-pressure is positively et al. [12], Dolan and Metcalfe [16], and Hayes and Cone [13]
 correlated with
and significantly  occupants’ motivation to con- stated that information distribution of energy consumption facts
serve energy ˇ = 0.19, p < 0.1 . If an occupant observes his/her and reduction guidelines do not appear to effectively influence
co-workers always use strategies to conserve energy, he/she will occupants. Moreover, He et al. [18]; Marans and Edelstein [14]
also be highly motivated to conserve energy. Direct exposure demonstrated that feedback and peer-comparison are the most
to energy conservation-related information, however, does not effective education methods to influence occupants’ energy con-
have any effect  on occupants’ energy conservation motivation servation behaviors. Klein et al. [3] highlighted that occupant
ˇ = −0.01, ns . These results indicate a mixed support for H4. engagement in building energy reduction strategies is critical
H5 determines if the perceived energy conservation knowledge and can be achieved through informed feedback and suggestions.
has a positive impact on energy conservation behaviors in the Peschiera and Taylor [19] and Peschiera et al. [20] presented that
presence of energy conservation motivation. The results indicate energy use feedback interventions which monitor and report build-
that there is a positive and significant correlation between occu- ing energy use of peers are very effective in promoting energy

pants perceived knowledge on energy conservation and occupants’ reduction strategies. Additionally, the finding of thermal comfort
motivation to conserve energy ˇ = 0.23, p < 0.05 . This result as an influential factor of energy saving behaviors also conforms to
indicates that higher perceived energy conservation knowledge Maslow’s hierarchy of needs [115]. Occupants are willing to take
results to higher motivation of occupants. These findings indicate energy saving actions and make contributions to the environment
support for H5. (self-actualization) only when their fundamental needs (e.g., phys-
H6 determines if higher knowledge level of energy consumption iological needs of thermal comfort) have been satisfied.
facts would lead to more energy saving behaviors, such that more However, it should be noted that these results highly depend
knowledgeable occupants would perform more energy conserva- on the data collected from the case study building. The major influ-
tion behaviors than less knowledgeable occupants when energy ential factors identified in this study may not remain the same for
conservation motivation is high. The results indicate that there other buildings. For the non-influential factors, further investiga-

is no significant correlation between occupants’ knowledge 
level tions in the workplace can be conducted to identify the reason why
on energy consumption facts and their motivation ˇ = 0.16, ns . these factors don’t show a strong impact on energy saving behav-
Therefore, these results do not support H6. iors. It might seem counterintuitive that people with control of
The results from SEM analysis which used to test six hypotheses lighting and office plug loads fail to perform more energy saving
stated in this study present several strategies to improve the moti- behaviors. However, there can be several reasons leading to such
vation, opportunity and ability level of occupants in the chosen situations. For example, some office equipment is public appliances
building, and accordingly their energy use behaviors for adopting (e.g., printer, microwave oven) which are not managed by a single
energy reduction strategies. person. It is also likely that occupants may not turn off the light in
a multi-occupancy room even if there is enough daylight because
5.2. Discussion of SEM analysis results the switch is next to an overbearing colleague who can be annoyed
when bothered. Moreover, company’s eco-promotion may fail if
Based on the results obtained from the previous section, higher the posters are not noticeable to the employees, or the proposal is
occupants’ motivation level (e.g., looking for ways to conserve difficult to perform.
energy in the office) can lead to some energy saving behaviors.
Therefore, occupancy interventions aiming at better occupants’
behaviors can focus on improving occupants’ motivation level. 5.3. Energy savings implications
Moreover, we identified two factors of opportunity characteristic
and one factor of ability characteristic which demonstrate strong This section explains how the results obtained from the MOA
correlations with occupants’ motivation from the list of surveyed and SEM analyses can be used to predict potential energy sav-
factors as shown in Table 3. These three factors are the major influ- ings from chosen intervention strategies. First, k-means clustering
ential factors of occupants’ energy saving behavior in this particular analysis is conducted to group the occupants in five categories (as
building. Therefore, to promote energy saving behaviors, occu- shown in Fig. 7) based on each occupant’s MOA level measured from
pancy interventions can be designed to: (1) provide occupants with on a scale of 0–100. The centroid of each cluster is calculated (see
thermal comfort conditions; (2) increase the peer-pressure among Table 5) and then mapped with occupant’s MOA characteristics.
the workers, and (3) enhance occupants’ self-assessed knowledge For example, occupants in prone category have the highest moti-
on energy conservation to promote their motivation level. For vation (80) and relatively high opportunity (51) and ability level
example, facility managers of the analyzed case study building (61) while occupants in resistant category have the lowest moti-
can install thermostat in each office room to improve the indoor vation (20), opportunity (30) and ability level (21). More details
thermal comfort, and also send regular emails (e.g., weekly) to occu- about clustering occupants can be found in Karatas et al. [116].
pants presenting their personal energy usage and a comparison of This supports the findings as discussed in the SEM analysis, a higher
the energy consumption among colleagues as well as some tips of opportunity and ability level will promote an occupant’s motiva-
daily energy saving behaviors. tion to conserve energy. For this case study building, the number of
These findings from the case study analysis also support the occupants in each category (from prone to resistant) is 29, 30, 29,
research studies conducted by [3,12–14,16,18–20]. Agha-Hossein 37 and 5 correspondingly.
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 109

Fig. 7. K-means Clustering Analysis for Occupants in the Case Study Building.

As shown in Fig. 6, peer pressure is one of methods that is shown Physical capability is excluded from the measurements (E.g., light
to improve opportunity level of the occupants and thus having a switches are set too high for a wheelchair user to reach it). However,
positive influence on their motivation level. Therefore, peer pres- through the evaluation of behavior change wheel (BCW) framework
sure functionality in the Agent-Based Model [9,29] is selected to in the domain of energy behavior, Wilson and Marselle [63] found
study the potential impact of this intervention on reducing the physical capability to be the least frequently occurring determinant
occupant’s energy consumption in the building. The occupants are in behavior change guidance documents. They argue that physi-
assumed to form a single small world network, and their energy cal capability might be more common in health field compared to
intensity (EI) is initialized using log normal distribution (␮ = 1.626, energy behavior domain, which requires further investigation. Sec-
␴ = 0.875, min = 0.272), emulating the energy consumption of a typ- ond, when evaluating ability level, occupants’ perceived knowledge
ical US office buildings based on CBECS data (2003). However, the about possible conservation strategies is assumed correct, which
energy variability (Var) and corresponding number of people are might not always be true. Occupants’ wrong assumption about
established based on the MOA characteristics from this study. Occu- energy conservation actions might turn out to be counterproduc-
pants in prone category have the largest Var, followed by mildly tive. Third, the proposed framework mainly focuses on individual’s
unable category, unable category, etc. During the simulation, occu- MOA level to determine energy use characteristics. However, other
pant’s EI and Var change due to presence of peer pressure. When social factors (e.g., education, religion, economic status, lifestyle)
the model reaches convergence, the result shows that 95 out of may also collectively affect an occupant’s energy behavior, which
130 occupants have reduced their EI. The average EI decreases from requires research studies to use more generic metrics in the future.
6.93 to 4.19 kWh/m2/person/year. Fig. 8 shows the original number Fourth, motivation level is treated as an overall measure of occu-
of occupants in each of the five categories and the corresponding pant’s concern and desire of energy conservation. However, in
number of occupants in each category who reduced their energy reality people may have a mixed level of motivation for different
consumption. These results support the conclusions from the SEM behaviors (e.g., people may be more motivated to turn off the light
analysis that peer pressure in this case can enhance the opportunity than to adjust the thermostat). Fifth, the proposed framework is
level of the occupants, which directly improves their motivation only tested in commercial buildings. In residential properties, util-
level and thus their energy use behavior. However, if the building ity bill can become a key factor of energy use. Future studies will
has a large number of people in each cluster (or in a multi-tenanted take household energy bill into consideration when investigating
building), different interventions can be considered for each cluster. the influential factors in residential buildings.
The choice of multi-interventions is beyond the scope of this paper
and is part of our future research efforts to map building occupancy 7. Conclusion and policy implications
clusters obtained based on MOA to different interventions.
This study developed a framework which is capable of iden-
6. Limitations tifying the influential factors on building occupants’ energy use
characteristics and their relevant behaviors to adopt energy
This research has a number of limitations. First, as pointed out by reduction strategies. To achieve this, the study proposed an
several studies [60–62], the behavioral determinant ability includes approach from social sciences which adopts an analogy of motiva-
both psychological and physical capability or effort to engage tion/opportunity/ability (MOA) model and marketing field. Based
human in a targeted behavior. In our MOA model, ability is mainly on this analogy, the MOA definitions are adapted to context of
a knowledge-based measure which affects occupant’s interpreta- energy use in buildings, and a set of research hypotheses and a
tion, comprehension, and reasoning about energy use information. corresponding survey are developed to evaluate the MOA influen-
110 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

Fig. 8. Distribution of Occupants Who Reduced Energy Consumption in Each Category.

Fig. 9. Steps for Designing Occupancy-Focused Intervention Strategies.

tial factors. The developed model is implemented using structural approaches such as education (see Fig. 9). For buildings which show
equation modeling (SEM), a widely used approach in behavioral diverse occupants’ characteristics, multi-level interventions can be
sciences, to investigate the direct impact of occupants’ energy use adopted to account for different groups of occupants. Although
motivation level, and moderating effect of their opportunity and the implementation and testing of an intervention design in this
ability levels on energy use behaviors. Agent based modeling (ABM) manner are outside the scope of this paper, the results from the
is then used to estimate the potential impact of the chosen inter- agent based simulation analysis illustrate that energy savings are
vention on energy reduction. The results obtained from the case possible when the right type of intervention is designed. This
study building demonstrate the model’s capability for achieving approach of designing and implementing interventions based on
the objectives described in this paper. occupancy characteristics is bound to improve the effectiveness of
It is worth noting that the major contribution of this study is the occupancy-focused interventions and providing sustainable energy
flexible evaluation framework, which may vary from case to case reductions in buildings. In particular, policy makers can implement
due to different building types, weather zones, occupancy states this methodology to obtain an understanding of the population
(multi-tenanted vs single company) to name a few. This frame- perspective in city blocks, urban environments or residential neigh-
work can be adopted and modified according to the actual situation borhoods. They can then use the results to determine which of the
of a given building. Decision-makers can first analyze the influ- M, O or A needs to be supplemented or addressed by the interven-
ential factors and MOA levels relevant to their building, and then tion method or policy to achieve large scale and effective energy
use the resulting information to design occupancy-focused energy reductions.
reduction intervention strategies tailored to occupants’ charac- Future research by the authors will focus on collecting energy
teristics in their building. For example, if a building has a high use data from building automation systems to correlate actual lev-
percentage of occupants who are identified as “prone” to change els of energy use to intentional occupancy behaviors. This will allow
energy use behaviors after distributing the survey and analyz- the research team to subsequently be able to design and test inter-
ing the data following the approach proposed in this study, then vention strategies under different scenarios.
the chosen intervention strategies can focus on knowledge-based
 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112 111

Acknowledgements [24] D.D. Whitsett, H.C. Justus, E. Steiner, K. Duffy, Persistence of Energy
Efficiency Behaviors over Time: Evidence from a Community-Based
Program, 2013.
The authors would like to acknowledge the financial support [25] A. Ruzelli, Proceedings of the 2nd ACM Workshop on Embedded Sensing
for this research received from the US National Science Foundation Systems for Energy-Efficiency in Building. ACM (2010) 93, 1878431.
(NSF) CBET 1407908 and 1349921. Any opinions and findings in [26] N.H. Wong, D. Cheong, H. Yan, J. Soh, C. Ong, A. Sia, The effects of rooftop
garden on energy consumption of a commercial building in Singapore,
this paper are those of the authors and do not necessarily represent Energy Build. 35 (4) (2003) 353–364.
those of NSF. [27] I.M. Azevedo, Consumer end-use energy efficiency and rebound effects,
Annu. Rev. Environ. Resour. 39 (2014) 393–418.
[28] L.A. Greening, D.L. Greene, C. Difiglio, Energy efficiency and
consumption—the rebound effect—a survey? Energy Policy 28 (6) (2000)
References 389–401.
[29] E. Azar, C.C. Menassa, Evaluating the impact of extreme energy use behavior
[1] US EIA, How Much Energy Is Consumed in Residential and Commercial on occupancy interventions in commercial buildings, Energy Build. 97
Buildings in the United States? 2014, http://www.eia.gov/tools/faqs/faq. (2015) 205–218.
cfm?id=86&t=1 (31 December 2014). [30] E. Azar, C.C. Menassa, A comprehensive analysis of the impact of occupancy
[2] UNEP, Common Carbon Metric for Measuring Energy Use & Reporting parameters in energy simulation of office buildings, Energy Build. 55 (2012)
Greenhouse Gas Emissions from Building Operations, 2010, http://www. 841–853.
unep.org/sbci/pdfs/UNEPSBCICarbonMetric.pdf (31 December 2010). [31] T. Hong, Occupant behavior: impact on energy use of private offices, in:
[3] L. Klein, J.-Y. Kwak, G. Kavulya, F. Jazizadeh, B. Becerik-Gerber, P. ASim 2012-1st Asia Conference of International Building Performance
Varakantham, M. Tambe, Coordinating occupant behavior for building Simulation Association, Shanghai, China, 2014, 11/25/12–11/27/12.
energy and comfort management using multi-agent systems Autom. Constr. [32] Rik Pieters, et al., Consumers’ attributions of proenvironmental behavior,
22 (2012) 525–536. motivation, and ability to self and others, J. Public Policy Mark. (1998)
[4] M. Moezzi, M. Iyer, L. Lutzenhiser, J. Woods, Behavioral assumptions in 215–225.
energy efficiency potential studies, in: Prepared for the California Institute [33] Jody M. Hines, Harold R. Hungerford, Audrey N. Tomera, Analysis and
for Energy and Environment (CIEE), 2009, May, Oakland, Calif. synthesis of research on responsible environmental behavior: a
[5] M. Sanchez, C. Webber, R. Brown, J. Busch, M. Pinckard, J. Roberson, Space meta-analysis, J. Environ. Educ. 18 (2) (1987) 1–8.
Heaters, Computers, Cell Phone Chargers: How Plugged in Are Commercial [34] T. Hong, S. D’Oca, W.J. Turner, S.C. Taylor-Lange, An ontology to represent
Buildings? Lawrence Berkeley National Laboratory, 2007. energy-related occupant behavior in buildings. Part I: introduction to the
[6] J.-J. Wang, Y.-Y. Jing, C.-F. Zhang, J.-H. Zhao, Review on multi-criteria DNAs framework, Build. Environ. 92 (2015) 764–777.
decision analysis aid in sustainable energy decision-making Renew. Sustain. [35] O.T. Masoso, L.J. Grobler, The dark side of occupants’ behaviour on building
Energy Rev. 13 (9) (2009) 2263–2278. energy use, Energy Build. 42 (2) (2010) 173–177.
[7] M. Schweiker, M. Shukuya, Comparative effects of building envelope [36] Wouter Poortinga, Linda Steg, Charles Vlek, Values, environmental concern,
improvements and occupant behavioural changes on the exergy and environmental behavior a study into household energy use, Environ.
consumption for heating and cooling, Energy Policy 38 (6) (2010) Behav. 36 (1) (2004) 70–93.
2976–2986. [37] Benjamin K. Sovacool, What are we doing here: analyzing fifteen years of
[8] C. Von Borgstede, M. Andersson, F. Johnsson, Public attitudes to climate energy scholarship and proposing a social science research agenda, Energy
change and carbon mitigation—implications for energy-associated Res. Soc. Sci. 1 (2014) 1–29.
behaviours, Energy Policy 57 (2013) 182–193. [38] M.A. Abraham, N. Nguyen, Green engineering: defining the principles,
[9] E. Azar, C. Menassa, A framework to evaluate energy saving potential from Environ. Prog. 22 (4) (2003) 233–236.
occupancy interventions in typical US commercial buildings, J. Comput. Civil [39] T.F. Allen, in: C.J. Kibert, Jan Sendzimir, G.B. Guy (Eds.), Applying the
Eng. – Spec. Issue Comput. Approaches Underst. Reduce Energy Consump. Principles of Ecological Emergence to Building Design and Construction,
Built Environ. ASCE 28 (1) (2014) 63–78. Spon, 2002, London, 2004, pp. 108–126.
[10] A.R. Carrico, M. Riemer, Motivating energy conservation in the workplace: [40] E. Azar, C. Menassa, An agent-based approach to model the effect of
an evaluation of the use of group-level feedback and peer education? J. occupants’ energy use characteristics in commercial buildings, Am. Soc. Civil
Environ. Psychol. 31 (1) (2011) 1–13. Eng. (ASCE) (2011) 536–543.
[11] W. Abrahamse, L. Steg, C. Vlek, T. Rothengatter, A review of intervention [41] A. Entrop, H. Brouwers, A. Reinders, Evaluation of energy performance
studies aimed at household energy conservation, J. Environ. Psychol. 25 indicators and financial aspects of energy saving techniques in residential
(2005) 273–291. real estate? Energy Build. 42 (5) (2010) 618–629.
[12] M. Agha-Hossein, R. Tetlow, M. Hadi, S. El-Jouzi, A. Elmualim, J. Ellis, M. [42] B. Poel, G. van Cruchten, C.A. Balaras, Energy performance assessment of
Williams, Providing persuasive feedback through interactive posters to existing dwellings, Energy Build. 39 (4) (2007) 393–403.
motivate energy-saving behaviours, Intell. Build. Int. (2014) 1–20, [43] F. Nemry, A. Uihlein, C.M. Colodel, C. Wetzel, A. Braune, B. Wittstock, I.
Ahead-of-print. Hasan, J. Kreißig, N. Gallon, S. Niemeier, Options to reduce the
[13] S.C. Hayes, J.D. Cone, Reducing residential electrical energy use: payments, environmental impacts of residential buildings in the European
information, and feedback1, J. Appl. Behav. Anal. 10 (3) (1977) 425–435. Union—potential and costs, Energy Build. 42 (7) (2010) 976–984.
[14] R.W. Marans, J.Y. Edelstein, The human dimension of energy conservation [44] J.H. Scofield, Do LEED-certified buildings save energy? Not really. . ., Energy
and sustainability: a case study of the University of Michigan’s energy Build. 41 (12) (2009) 1386–1390.
conservation program? Int. J. Sustain. High. Educ. 11 (1) (2010) 6–18. [45] J. Yudelson, Greening Existing Buildings, McGraw-Hill, New York, 2010.
[15] Stephanie N. Timm, Brian M. Deal, Effective or ephemeral? The role of [46] S.J. Bosch, A.R. Pearce, Sustainability in public facilities: analysis of guidance
energy information dashboards in changing occupant energy behaviors, documents? J. Perform. Constr. Facil. 17 (1) (2003) 9–18.
Energy Res. Soc. Sci. 19 (2016) 11–20. [47] Y.-K. Juan, P. Gao, J. Wang, A hybrid decision support system for sustainable
[16] P. Dolan, R. Metcalfe, Neighbors, Knowledge, and Nuggets: Two Natural office building renovation and energy performance improvement, Energy
Field Experiments on the Role of Incentives on Energy Conservation, 2013. Build. 42 (3) (2010) 290–297.
[17] J.H. Van Houwelingen, W.F. Van Raaij, The effect of goal-setting and daily [48] S.M. Beheiry, W.K. Chong, C.T. Haas, Examining the business impact of
electronic feedback on in-home energy use, J. Consum. Res. (1989) 98–105. owner commitment to sustainability? J. Constr. Eng. Manag. 132 (4) (2006)
[18] H.A. He, S. Greenberg, E.M. Huang, One size does not fit all: applying the 384–392.
transtheoretical model to energy feedback technology design, in: [49] H.M. Bernstein, M.A. Russo, Green Building Retrofit and Renovation: Rapidly
Proceedings of the SIGCHI Conference on Human Factors in Computing Expanding Market Opportunities Through Existing Buildings, Smart Market
Systems, ACM, 2010, pp. 927–936. Report, 2009.
[19] G. Peschiera, J.E. Taylor, The impact of peer network position on electricity [50] A. Mahdavi, C. Pröglhöf, Toward empirically-based models of people’s
consumption in building occupant networks utilizing energy feedback presence and actions in buildings, Proc., Proceedings of Building Simulation
systems, Energy Build. 49 (2012) 584–590. (2009) 537–544.
[20] G. Peschiera, J.E. Taylor, J.A. Siegel, Response–relapse patterns of building [51] J. Tanimoto, A. Hagishima, Total Utility Demand Prediction Based on
occupant electricity consumption following exposure to personal, Probabilistically Generated Behaviroal Schedules ff Actural Inhabitants,
contextualized and occupant peer network utilization data, Energy Build. 42 2009.
(8) (2010) 1329–1336. [52] E. Azar, C.C. Menassa, A comprehensive framework to quantify energy
[21] M.A. Alahmad, P.G. Wheeler, A. Schwer, J. Eiden, A. Brumbaugh, A savings potential from improved operations of commercial building stocks,
comparative study of three feedback devices for residential real-time energy Energy Policy 67 (2014) 459–472.
monitoring, IEEE Trans. Ind. Electron. 59 (4) (2012) 2002–2013. [53] J. Chen, J.E. Taylor, H.H. Wei, Modeling building occupant network energy
[22] T.E. Boyce, E.S. Geller, Encouraging college students to support consumption decision-making: the interplay between network structure
pro-environment behavior effects of direct versus indirect rewards, Environ. and conservation, Energy Build. 47 (2012) 515–524.
Behav. 33 (1) (2001) 107–125. [54] R.K. Jain, R. Gulbinas, J.E. Taylor, P.J. Culligan, Can social influence drive
[23] M.J. Handgraaf, M.A. Van Lidth de Jeude, K.C. Appelt, Public praise vs: private energy savings? Detecting the impact of social influence on the energy
pay: effects of rewards on energy conservation in the workplace, Ecol. Econ. consumption behavior of networked users exposed to normative
86 (2013) 86–92. eco-feedback, Energy Build. 66 (2013) 119–127.
112 D. Li et al. / Energy Research & Social Science 23 (2017) 97–112

[55] C.A. Webber, J.A. Roberson, M.C. McWhinney, R.E. Brown, M.J. Pinckard, J.F. [85] C. Moorman, E. Matulich, A model of consumers’ preventive health
Busch, After-hours power status of office equipment in the USA, Energy 31 behaviors: the role of health motivation and health ability, J. Consum. Res.
(14) (2006) 2823–2838. (1993) 208–228.
[56] L. Wolff, H. Massett, E. Maibachc, D. Weberd, S. Hassmillere, R. [86] L. Steg, C. Vlek, Encouraging pro-environmental behaviour: an integrative
Mockenhaupte, Validating a health consumer segmentation model: review and research agenda, J. Environ. Psychol. 29 (3) (2009) 309–317.
behavioral and attitudinal differences in disease prevention-related [87] J.L. Zaichkowsky, Measuring the involvement construct, J. Consum. Res.
practices, J. Health Commun.: Int. Perspect. 15 (2) (2010) 167–188. (1985) 341–352.
[57] L. Sudbury, P. Simock, A multivariate segmentation model of senior [88] M.L. Richins, P.H. Bloch, After the new wears off: the temporal context of
consumers, J. Consum. Mark. 26 (4) (2009) 251–262. product involvement, J. Consum. Res. (1986) 280–285.
[58] B. Bayus, R. Mehta, A segmentation model for the targeted marketing of [89] E. Parra-Lopez, D. Gutierrez-Tano, R.J. Diaz-Armas, J. Bulchand-Gidumal,
consumer durables, J. Mark. Res. 32 (4) (1995) 463–469. Travellers 2.0: motivation, opportunity and ability to use social media, in:
[59] Martin Fishbein, et al., Factors Influencing Behavior and Behavior Change, Social Media in Travel, Tourism and Hospitality, 2012, pp. 171–185.
2001, pp. 1–17. [90] K. Hallahan, Enhancing motivation, ability, and opportunity to process
[60] Susan Michie, Maartje M. van Stalen, Robert West, The behaviour change public relations messages, Public Relat. Rev. 26 (4) (2001) 463–480.
wheel: a new method for characterising and designing behaviour change [91] R. Govindaraju, A.F. Hadining, D.R. Chandra, Physicians’ adoption of
interventions, Implement. Sci. 6 (1) (2011) 1. electronic medical records: model development using
[61] Brian J. Fogg, A behavior model for persuasive design, Proceedings of the 4th ability-motivation-opportunity framework, in: Information and
International Conference on Persuasive Technology. ACM (2009). Communication Technology, Springer, 2013, pp. 41–49.
[62] Icek Ajzen, Attitudes, Personality, and Behavior, McGraw-Hill Education, UK, [92] Melvin Blumberg, Charles D. Pringle, The missing opportunity in
2005. organizational research: some implications for a theory of work
[63] Caroline Wilson, Melissa R. Marselle, Insights from psychology about the performance, Acad. Manag. Rev. 7 (4) (1982) 560–569.
design and implementation of energy interventions using the behaviour [93] A.H. McMakin, E.L. Malone, R.E. Lundgren, Motivating residents to conserve
change wheel, Energy Res. Soc. Sci. 19 (2016) 177–191. energy without financial incentives, Environ. Behav. 34 (6) (2002) 848–863.
[64] Petromil Petkov, et al., Personalised eco-feedback as a design technique for [94] M. Eccles, J. Grimshaw, A. Walker, M. Johnston, N. Pitts, Changing the
motivating energy saving behaviour at home, Proceedings of the 7th Nordic behavior of healthcare professionals: the use of theory in promoting the
Conference on Human-Computer Interaction: Making Sense Through uptake of research findings, J. Clin. Epidemiol. 58 (2) (2005) 107–112.
Design. ACM (2012). [95] E.S. Geller, Evaluating energy conservation programs: is verbal report
[65] D.V. Geelen, Stimulating energy efficiency in households: comparison of the enough? J. Consum. Res. (1981) 331–335.
Livinggreen. eu methods to theory, in: Livinggreen Scientific Conference, [96] B.G. Southwell, A. Torres, Connecting interpersonal and mass
Delft, The Netherlands, 19 April 2013, Delft University of Technology, 2013. communication: science news exposure, perceived ability to understand
[66] Magnus Bang, Carin Torstensson, Cecilia Katzeff, The powerhhouse: a science, and conversation, Commun. Monogr. 73 (3) (2006) 334–350.
persuasive computer game designed to raise awareness of domestic energy [97] C. Wilson, H. Dowlatabadi, Models of decision making and residential
consumption, in: International Conference on Persuasive Technology, energy use, Annu. Rev. Environ. Resour. 32 (2007) 169–203.
Springer, Berlin, Heidelberg, 2006. [98] W. Abrahamse, L. Steg, How do socio-demographic and psychological
[67] Sam C. Staddon, et al., Intervening to change behaviour and save energy in factors relate to households’ direct and indirect energy use and savings? J.
the workplace: a systematic review of available evidence, Energy Res. Soc. Econ. Psychol. 30 (5) (2009) 711–720.
Sci. 17 (2016) 30–51. [99] G. Ray, W.A. Muhanna, J.B. Barney, Information technology and the
[68] DEFRA, A Framework for Pro-Environmental Behaviors? 2008 https://www. performance of the customer service process: a resource-based analysis,
gov.uk/government/uploads/system/uploads/attachment data/file/69277/ MIS Q. (2005) 625–652.
pb13574-behaviours-report-080110.pdf. [100] E.D. Rosenzweig, A.V. Roth, J.W. Dean Jr., The influence of an integration
[69] Charles Vlek, Essential psychology for environmental policy making Int. J. strategy on competitive capabilities and business performance: an
Psychol. 35 (2) (2000) 153–167. exploratory study of consumer products manufacturers, J. Oper. Manage. 21
[70] Charles Abraham, Susan Michie, A taxonomy of behavior change techniques (4) (2003) 437–456.
used in interventions, Health Psychol. 27 (3) (2008) 379. [101] J. Nunnally, Psychometric Methods, McGraw-Hill, New York, 1978.
[71] John ThØgersen, Understanding of consumer behaviour as a prerequisite for [102] J.F. Hair, Multivariate Data Analysis, 2010.
environmental protection, J. Consum. Policy 18 (4) (1995) 345–385. [103] K.A. Bollen, J.S. Long, Testing Structural Equation Models, Sage, 1993.
[72] Janet Stephenson, et al., Energy cultures: a framework for understanding [104] T. Al-Maghrabi, C. Dennis, S. Vaux Halliday, Antecedents of continuance
energy behaviours, Energy Policy 38 (10) (2010) 6120–6129. intentions towards e-shopping: the case of Saudi Arabia, J. Enterp. Inf.
[73] Elizabeth Shove, Beyond the ABC: climate change policy and theories of Manag. 24 (1) (2011) 85–111.
social change, Environ. Plan. A 42 (6) (2010) 1273–1285. [105] G.C. Bruner, A. Kumar, Explaining consumer acceptance of handheld
[74] M.J. Polonsky, W. Binney, J. Hall, Developing better public policy to motivate internet devices, J. Bus. Res. 58 (5) (2005) 553–558.
responsible environmental behavior – an examination of managers’ [106] T.L. Childers, C.L. Carr, J. Peck, S. Carson, Hedonic and utilitarian motivations
attitudes and perceptions towards controlling introduced species, J. for online retail shopping behavior, J. Retail. 77 (4) (2002) 511–535.
Nonprofit Public Sect. Mark. 12 (1) (2004) 93–107. [107] H. Van der Heijden, T. Verhagen, M. Creemers, Understanding online
[75] H. Buurma, Public policy marketing: marketing exchange in the public purchase intentions: contributions from technology and trust perspectives,
sector, Eur. J. Mark. 35 (11/12) (2001) 1287–1302. Eur. J. Inf. Syst. 12 (1) (2003) 41–48.
[76] K.A. Machleit, T.J. Madden, C.T. Allen, Measuring and modeling brand [108] J.B. MacQueen, Some methods for classification and analysis of multivariate
interest as an alternative ad effect with familiar brands, Adv. Consum. Res. observations, Berkeley, University of California Press, in: Proceedings of 5-th
17 (1) (1990). Berkeley Symposium on Mathematical Statistics and Probability, 1, 1967,
[77] D.J. MacInnis, C. Moorman, B.J. Jaworski, Enhancing and measuring pp. 281–297.
consumers’ motivation, opportunity, and ability to process brand [109] US Energy Information Administration, Commercial Buildings Energy
information from ads, J. Mark. (1991) 32–53. Consumption Survey, 2003.
[78] C. Moorman, The effects of stimulus and consumer characteristics on the [110] Minitab, How to Interpret a Regression Model with Low R-squared and Low
utilization of nutrition information, J. Consum. Res. (1990) 362–374. P Values? 2014 http://blog.minitab.com/blog/adventures-in-statistics/how-
[79] M.L. Rothschild, Carrots, sticks, and promises: a conceptual framework for to-interpret-a-regression-model-with-low-r-squared-and-low-p-values.
the management of public health and social issue behaviors, J. Mark. (1999) [111] R.C. MacCallum, M.W. Browne, H.M. Sugawara, Power analysis and
24–37. determination of sample size for covariance structure modeling Psychol.
[80] R.L. Celsi, J.C. Olson, The role of involvement in attention and Methods 1 (2) (1996) 130.
comprehension processes, J. Consum. Res. (1988) 210–224. [112] A.C. Acock, Discovering Structural Equation Modeling Using Stata, Stata
[81] M. Hastak, M.B. Mazis, L.A. Morris, The role of consumer surveys in public Press books, 2013.
policy decision making, J. Public Policy Mark. 20 (2) (2001) 170–185. [113] R.B. Kline, D.A. Santor, Principles & practice of structural equation
[82] E. Bigné, B. Hernández, C. Ruiz, L. Andreu, How motivation, opportunity and modelling, Can. Psychol. 40 (4) (1999) 381.
ability can drive online airline ticket purchases, J. Air Transp. Manag. 16 (6) [114] Stata, Stata Structural Equation Modeling Reference Manual (Release 13),
(2010) 346–349. 2013 http://www.stata.com/manuals13/sem.pdf&gt.
[83] Thomas W. Gruen, Talai Osmonbekov, Andrew J. Czaplewski, How [115] Abraham Maslow, Maslow’s Hierarchy of Needs, Salenger Incorporated,
e-communities extend the concept of exchange in marketing: an 1987.
application of the motivation, opportunity, ability (MOA) theory, in: [116] Aslihan Karatas, Allisandra Stoiko, Caro C. Menassa, Framework for selecting
Marketing Theory 5.1, 2005, pp. 33–49. occupancy-focused energy interventions in buildings, Build. Res. Inf. 44
[84] W.D. Hoyer, D. MacInnis, Consumer Behavior, Houghton Mifflin, Boston, MA, (5–6) (2016) 535–551.
1997.