Energy Policy ()

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

A study of factors enhancing smart grid consumer engagement

Chan-Kook Park a,b, Hyun-Jae Kim c,n, Yang-Soo Kim b
a
Business and Technology Management, KAIST, South Korea
b
Energy Policy Research Group, Korea Energy Economics Institute, South Korea
c
Korea Energy Economics Institute, 132, Naesonsunhwan-ro, Uiwang-si, Gyeonggi-do 437-713, South Korea

H I G H L I G H T S

We examine what factors inuence electricity consumers' smart grid acceptance.

We test the smart grid technology acceptance model including the perceived risk as a main factor.

The importance of consumer education and public relations of the smart grid has been conrmed.

Another shortcut to ensure the acceptance of the smart grid is to mitigate the anxiety about the risk in the use of the smart grid.

art ic l e i nf o a b s t r a c t

Article history: It is important to ensure consumer acceptance in a smart grid since the ultimate deployment of the
Received 30 December 2013 smart grid depends on the end users' acceptance of smart grid products and services such as smart
Received in revised form meters and advanced metering services. We examine how residential consumers perceive the smart grid
10 March 2014
and what factors inuence their acceptance of the smart grid through a survey for electricity consumers
Accepted 12 March 2014
in Korea. In this study, consumers' smart grid acceptance factors, including the perceived risk, were
examined with the existing technology acceptance model suggested by Davis. This study has an
Keywords: implication that it has provided theoretical and empirical ground, based on which the policies to
Smart grid promote consumer participation in the deployment of the smart grid can be developed. Since there are
Consumer engagement
few studies on the policies from the perspective of the smart grid users, this study will contribute
Technology acceptance model
directly to the development of the strategy to ensure the acceptance of the smart grid.
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction grid is something unfamiliar, possessing acceptance restraints

A smart grid is a next-generation power grid to optimize
energy efciency with exchange of real-time information between
suppliers and consumers by integrating information and commu-
nication technologies with the existing power grid. Through
a market's various price signals, energy consumption can be
managed in a time-based way and the supply and demand of
power can be optimized. Since 2009, many countries around the
world have actively promoted various projects aimed at building a
smart grid as the primary means for green growth. The countries
have been installing advanced metering infrastructure including
smart meters, which are considered as primary infrastructure for a
smart grid, at residential buildings as well as commercial and
industrial ones. As for residential consumers, however, the smart
n
Corresponding author. Tel.: 82 31 420 2228.
E-mail addresses: green@keei.re.kr (C.-K. Park), hjkim@keei.re.kr (H.-J. Kim),
yskim12239@keei.re.kr (Y.-S. Kim).
such as cyber security threats, the possibility of electricity rate
increase, and reluctance among targets in using a new kind of
technology (Baltimoresun, 2009; Venturebeat, 2009; Earth2tech,
2010; Chicagotribune, 2011).
In order to successfully build a smart grid, it is necessary to set
key policy objectives to ensure consumer acceptance and to make
a scientic analysis of the factors affecting consumer acceptance,
since the ultimate deployment of the smart grid depends on the
end users' acceptance of smart grid products and services (Harris
Interactive, 2010; Pike Research, 2010; IEA, 2011; SGCC (2010);
IBM, 2011; EPRI, 2012; GridWise Alliance, 2013). Likewise, the
degree of consumers' participation in smart grid projects causes a
high scale of smart grid benets. In Korea, the importance of
consumer acceptance has been stressed in smart grid policies, but
it has not been treated as a key policy objective, being emphasized
mainly in the aspect of public relations.
In this study, we examine how residential consumers perceive
the smart grid and what factors inuence their acceptance of the
smart grid through a survey to electricity consumers in Korea.

http://dx.doi.org/10.1016/j.enpol.2014.03.017
0301-4215/& 2014 Elsevier Ltd. All rights reserved.

Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
2 C.-K. Park et al. / Energy Policy ()
We assume that the advanced metering infrastructure including
smart meters is considered as the main smart grid technology to
residential electricity consumers. The research result would
contribute to lay the foundation for setting policies and strategies
that will ensure smart grid acceptance.
2. Theoretical discussion
2.1. Technology acceptance model
The technology acceptance model (TAM) proposed by Davis
(1989) has been widely applied in research related to information
and communication technology acceptance and has been regarded
as a suitable model to predict and explain the acceptance of a new
technology by many researchers.
In TAM, the users' technology acceptance is said to be made up
of two important beliefs: perceived usefulness and perceived ease
of use of a new technology. Perceived usefulness is dened as the
degree of belief that to use a specic technology will improve
one's work performance. Perceived ease of use is dened as the
degree of belief that to use a particular technology will be easy
(Davis, 1989). These two beliefs inuence the attitude toward
using, with the attitude toward using and the perceived usefulness
interacting to alter users' behavioral intention to use. Actual use is
affected by a user's behavioral intention in the same way perceived
usefulness is inuenced by perceived ease of use. TAM by Davis
(1989) is employed in Fig. 1.
Later, researchers have simplied TAM by removing the atti-
tude toward using, which is similar to the behavioral intention to
use. In other words, the users intention to use a technology is
affected directly by belief variables such as perceived usefulness
and perceived ease of use (Davis, 1989; Venkatesh and Davis,
2000).
After the publication of TAM by Davis (1989), many studies
have been conducted to modify and extend TAM. A representative
extended TAM is TAM II proposed in 2000 by Davis and Venkatesh
who had presented TAM for the rst time. TAM II emphasized
perceived usefulness over TAM I. The reason TAM II specied
perceived usefulness of the two variables is because perceived
usefulness consistently had higher explanatory power on the
intention to accept information technology than the perceived
ease of use as a result of most studies of TAM. TAM II was
supposed to explain perceived usefulness through variables such
as subjective norm, image, job relevance, output quality, and result
demonstrability. Meanwhile, Venkatesh and Bala (2008) presented
TAM III by providing the basic TAM with exogenous variables such
as individual differences, system characteristics, social inuence,
and facilitating conditions.
Other studies were actively done to learn if there was a third
determinant variable other than perceived usefulness and per-
ceived ease of use. Those studies started from the criticism that
Davis' TAM (Davis, 1989) did not reect users' various viewpoints
because it limited the determinant variables affecting the
Fig. 1. Davis' TAM.
Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
intention by using only perceived usefulness and perceived ease
of use. Koufaris and Hampton-Sosa (2002), in the study on the
acceptance of the Web, maintained that users' acceptance was
determined by ve factors: perceived control, pleasure of action,
concentration, along with perceived usefulness and perceived ease
of use of the existing TAM. In the study applied to e-commerce, by
adding the variable of trust to the existing TAM based on previous
studies, Gefen et al. (2003) stressed the importance of trust as a
factor affecting intention to use.
2.2. Needs to consider perceived risk in TAM
The existing technology acceptance theory focuses on per-
ceived usefulness and perceived ease of use as factors inuencing
one's intention to accept based on the theory of reasoned action.
It has been proven by many documental researches that a wide
range of explanation in relation to the intention to accept a
technology is possible with this rational decision-making factor.
However, the acceptance of a new technology is affected by
subjective and irrational factors such as emotion and image as
well as objective and rational factors (Kim, 2009). A representative
subjective factor with respect to technology acceptance is per-
ceived risk. There are important issues that hinder the acceptance
of smart grid technologies such as cyber security threats in a smart
grid, performance reliability of smart meters, and concerns about
electromagnetic radiation. The analysis of the smart grid accep-
tance factors should include the consideration of those risk factors.
The consideration of the perceived risk in relation to the
analysis of consumer behavior was started by Bauer (1960).
Bauer's perceived risk of consumers referred to subjective risk
which was distinct from the risk of objective probability. The
perceived risk involved psychological risk which was perceived in
the process of choice such as brand choice, store choice, and way
of purchasing a particular product (Yang and Jung, 1999:120121).
Bauer regarded consumers' perceived risk as a function of two
factors: uncertainty and performance (gain or loss) (Jun et al.,
2003:22).
Cox (1967) developed the idea of Bauer (1960) to dene
perceived risk as a function of two factors: uncertainty and loss.
Cox dened perceived risk as the existence of one or more cases in
a consumers mentality among three: not knowing the purpose to
buy a specic product, not knowing which choice to meet the
purchase purpose, and not knowing that negative consequences
resulted from dissatisfaction with the purchase result. Jacoby and
Kaplan (1972) and Vincent and Zikmund (1976) also analyzed
perceived risk with two factors of uncertainty and loss (Jun et al.,
2003:2224).
Though the denition of the perceived risk was more or less
different from one another among researchers, it generally
included both the subjective uncertainty which a consumer
perceived about the results from a purchasing behavior and
subjective expectations about any loss caused by the purchase
results (Kassarjian and Robertson, 1991; Stone and Gronhaug,
1993, Jun et al., 2003:24).
The type of perceived risk was categorized in various ways by
researchers. Roselius (1971) used time loss, risk loss, ego loss, and
pecuniary loss for the types of the perceived risks in the study on
how to reduce perceived risk. Brooker (1984), in the research on
the relationship between the characteristics of each risk type and
the overall risk about spaghettis and peaches, also classied the
risk types into psychological risk, nancial risk, functional risk,
physical risk, and social risk, clarifying that these ve types of
perceived risk could explain more than 60% of the overall
perceived risk. Taylor and Todd (1995) took the concept of loss
for the perceived risk and divided social/psychological loss and
functional/economic loss in the study on the consumers' behavior
 C.-K. Park et al. / Energy Policy ()
process according to the existence of perceived risk (Yang and
Jung, 1999:121122).
The perceived risk of the smart grid can be categorized largely
into the social/psychological risk of cyber security threats, the
physical risk of concerns about the electromagnetic radiation
(EMR) of smart meters, and the functional/economic risk repre-
sented by concerns about the accuracy and reliability of the smart
meter functions
Perceived Cyber Insecurity is the key risk that smart grid
consumers experience. Consumers' concerns about power
consumption data leakage grow all over the world, acting as a
barrier to the deployment of smart grids (GlobalData, 2010; IEA,
2011; Greentech Media, 2011).
The Fear of EMR is also another risk that power consumers
perceive in relation to the smart grid. As a result of the survey of
1082 homeowners and tenants aged 1875 in the United States on
the potential vulnerabilities of smart meters by Zpryme (2011),
61% of people are concerned about EMR compared with 39% of
people who are not.
Concerns about Equipment Malfunction and Performance
Degradation are also a kind of smart grid risk which consumers
perceive. According to prior studies, the reliability of smart meters,
the basic infrastructure of the smart grid, had an effect on the
acceptance of the smart grid. In the previously mentioned survey
by Zpryme (2011) regarding questions about the accuracy of data
and the utility rate, 36% of people answered very concerned, 32%
quite concerned, 14% a little concerned, and 14% not con-
cerned. According to the survey of 1678 American consumers
conducted by BCG (2010) in December 2009, 32% of people
answered that they were worried about the accuracy of smart
meters. In fact, after the smart meter deployment projects by
Pacic Gas and Electric (PG & E) in California and by Oncor in
Texas, the United States, the electricity rates were charged more
than double, thereby causing consumers to care about the relia-
bility of smart meter devices. Furthermore, the consumers charged
with the greatly increased rate led a lawsuit against the utility
company for the inaccuracy of smart meters (Venturebeat,
2009:11, 17).
3. Research model and hypotheses
3.1. Design of Risk Integrated TAM (RITAM)
In this study, in addition to Perceived Usefulness and Perceived
Ease of Use of the smart grid presented in TAM, Perceived Risk is
considered as a major factor affecting the Intention to Accept the
smart grid. As previously shown, TAM has developed into its own
TAM II and TAM III, at the same time adopting the perceived risk
theory. It has evolved to take the perceived risk into account for a
variable inuencing the intention to use besides perceived useful-
ness and perceived ease of use.
Wu and Wang (2005), in their research on consumer accep-
tance of mobile commerce, conrmed that perceived risk had a
negative impact on technology acceptance. Lee (2009), in a study
on the factors of Internet banking acceptance, showed that
perceived risk negatively inuenced attitude towards using and,
eventually, intention to use. Sanayei and Bahmani (2012) also, in
the study on customer acceptance of internet banking, analyzed
that perceived risk had a negative inuence on intention to use.
Yoon et al. (2011) also in the research on the acceptance of smart-
phone applications established a model to verify that perceived
risk affected intention to use.
While some studies focused on the proposition that perceived
risk had an effect on the intention to use a new technology, others
analyzed that perceived risk had an inuence on perceived
Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
3
Fig. 2. RITAM.
usefulness as well as intention to use. Siegrist et al. (2007), in
the study on the acceptance of nano-technology, maintained that
the factors determining the intention to buy nano-technology
foods were benets and risks, and that the benet variable had a
negative effect on the risk variable. Shim (2009) in a research on
the acceptance of nuclear-related technology argued that per-
ceived risk negatively affected acceptance, and that the awareness
of the benets of nuclear power plants had an effect to reduce the
perception of risks of nuclear power plants.
Meanwhile, Featherman and Pavlou (2002), in a study on the
acceptance of electronic service (e-service), made a conclusion
that perceived risk had a negative inuence not only on the
acceptance of e-service directly, but also on perceived usefulness.
Li and Huang (2009), in a research on the factors of online
shopping channel use, maintained that perceived risk along with
perceived usefulness and perceived ease of use, affected intention
to use, and that perceived risk even inuenced perceived ease
of use.
While all precedent studies reported that perceived risk had a
negative impact on intention to use, the causal relationship
between perceived risk and perceived benet sometimes indicated
the opposite direction. In other words, some presented perceived
risk as affecting perceived benet, and others provided that
perceived benet affected perceived risk. As for the relationship
between perceived risk and perceived benet, Finucane et al.
(2000), as a result of the analysis of the reference documents,
argued that the two variables had a negative effect on each other,
especially when people relied on emotions in judging the risks and
benets of a specic product and technology.
In this study to examine the acceptance of the smart grid, a
non-recursive model1 is designed, taking the relationship of the
two variables of perceived risk and perceived benet as mutually
negative. As shown in Fig. 2, the variable of the perceived risk is
added to the existing TAM. Perceived risk inuences Intention to
Use and has a mutually affecting relationship with perceived
benet.
3.2. Research model and hypotheses
In this study, the research model has been established by
adding exogenous variables to the earlier mentioned RITAM as
shown in Fig. 3.
1
When the causal relationship among the endogenous variables does not
direct to one way, but mutually as cause and effect to each other at the same time,
the model is called a non-recursive model.
4 C.-K. Park et al. / Energy Policy ()
Fig. 3. Research model.
The RITAM as a modied model reects the proposition that
users' Perceived Risk about a new technology has a negative
impact on the Intention to Use. Further, it shows that the more
the usefulness of the technology is perceived, the less the concerns
about the risk of the technology are perceived. Also, the more the risk
of the technology is perceived, the less the usefulness of the
technology is perceived. The RITAM has set a total of six hypotheses.
H1: Perceived Ease of Use has a positive () effect on Perceived
Usefulness.
H2: Perceived Usefulness has a negative (  ) effect on
Perceived Risk.
H3: Perceived Risk has a negative ( ) effect on Perceived
Usefulness.
H4: Perceived Risk has a negative (  ) effect on Intention
to Use.
H5: Perceived Usefulness has a positive () effect on Intention
to Use.
H6: Perceived Ease of Use has a positive () effect on Intention
to Use.
Perceived Compatibility and Perceived Understanding have
been selected for the exogenous variables that affect the Perceived
Ease of Use. Perceived Compatibility refers to the degree to which
the way of using and the value of a new technology is consistent
with those of the existing technology, enabling users to easily
accept the use of a new technology (Wu and Wang, 2005).
The importance of compatibility is especially emphasized in the
Innovation Diffusion Theory, in which the higher compatibility a
new innovative technology has with the existing technology, the
sooner it is accepted. The reason Perceived Understanding has
been selected as an exogenous variable for Perceived Ease of Use is
that the more understanding of a new technology users think to
have, the less perception of the complexity in the use of it they
have. The precedent studies showed that the higher the under-
standing of the smart grid the users have, the higher the
acceptance of smart grid technologies is (Pike Research, 2010;
GE, 2010). In this study, it is not maintained that Understanding of
the Smart Grid directly affects Intention to Use, but rather it is a
factor increasing the awareness of Ease of Use to raise Intention to
Use. The reason for the limited understanding of the smart grid
currently is that terminology such as demand response, dynamic
pricing, and so on is difcult to consumers (SGCC, 2010). Therefore,
many researches emphasize the importance of education to
promote consumers' understanding (Zpryme, 2011; IBM, 2011).
Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
H7: The Perceived Compatibility has a positive ( ) effect on the
Perceived Ease of Use.
H8: The Perceived Understanding has a positive ( ) effect on
the Perceived Ease of Use.
For the exogenous variables affecting Perceived Usefulness,
have been selected Perceived Understanding, Perceived Power
Supply Reliability, Perceived Electricity Rate Saving, and Perceived
Eco-Environment.
Perceived Understanding has an effect not only on Perceived
Ease of Use, but also on Perceived Usefulness because the higher
the understanding of a technology people have, the higher the
awareness of its usefulness they will have. As mentioned above,
the precedent studies showed that the understanding of a tech-
nology inuenced the acceptance of the technology. It is assumed
that the understanding of a technology has a positive impact on
the perception of usefulness, and it, in turn, has a positive effect on
intention to use.
Perceived Power Supply Reliability is also presumed to have a
positive effect on the awareness of the Usefulness of smart grid
technologies. According to the survey conducted by SGCC (Smart
Grid Consumer Collaborative) (2010) on the local residents of
Atlanta, Lansing, and Los Angeles as samples, the detection and
reduction of power outage and the expansion of renewable energy
ranked highly among the expected benets of smart meters in the
smart grid. As for the most important potential benets of the
smart grid in the survey by Zpryme (2011), the reduction of power
loss was 58.3%, and the smooth recovery in case of power failure
was 52% (Kim et al., 2012:102).
Another variable, the Perceived Electricity Rate Saving, is
supposed to have the most direct impact on Perceived Usefulness.
As a result of the survey of 10,000 power consumers in 15
countries done by IBM (2011), even though the reduction of
energy costs was not an absolute factor for the usefulness of the
smart grid, it was revealed to have the highest inuence of 62% on
consumers' behavioral patterns. This pecuniary benet received
more priority than the eco-environment by 12% at most in smart
grid usefulness. According to the research of some 1000 adults in
the United States conducted by EcoAlign (2010), eight out of 10
showed the most interest in the nancial factor such as cost saving
in the smart grid. Accenture (2011) conducted a survey targeting
10,199 electrical power consumers from 18 countries (more than
500 from each country) and presented the result that 91% of
respondents, an increase of 3% compared to the previous year, said
that the saving on the electricity bill was the most important
factor in using a power consumption management program.
The Perceived Eco-Environment can be thought as a factor
positively affecting the Perceived Usefulness of smart grid tech-
nologies. As the worldwide interest in climate change increases,
the consumers' environmental protection consciousness also
grows. According to the results of the following studies, the
Eco-Environment ranks second following the monetary benet
of electricity rate saving in inuencing the usefulness of green
energy technologies (Accenture, 2011; IBM, 2011; Oracle, 2009).
H9: Perceived Understanding has a positive ( ) effect on
Perceived Usefulness.
H10: Perceived Power Supply Reliability has a positive ( )
effect on Perceived Usefulness.
H11: Perceived Electricity Rate Saving has a positive ( ) effect
on Perceived Usefulness.
H12: Perceived Eco-Environment has a positive () effect on
Perceived Usefulness.
As for the exogenous variables having inuence on Perceived
Risk, Perceived Cyber Security Threats (Perceived Cyber Insecurity),
 C.-K. Park et al. / Energy Policy ()
Perceived Fear of Electromagnetic Radiation (Perceived EMR Fear),
Perceived Concerns about Equipment Malfunction, and Perfor-
mance Deterioration (Perceived Performance Concerns) have been
established. As we emphasized in the previous section of the
necessity of considering perceived risk in TAM, there are impor-
tant issues that hinder the acceptance of smart grid technologies
such as cyber security threats in a smart grid and concerns about
EMR and performance reliability of smart meters. The analysis of
the smart grid acceptance factors should include the consideration
of those risk factors.
H13: Perceived Cyber Insecurity has a positive ( ) effect on
Perceived Risk.
H14: Perceived EMR Fear has a positive () effect on
Perceived Risk.
H15: Perceived Performance Concerns has a positive () effect
on Perceived Risk.
4. Research methodology
4.1. Respondent selection and data collection
The purpose of this study is to examine the factors for potential
smart grid consumers to accept the smart grid, so electricity
consumers in the major metropolitans of South Korea were
selected as respondents of the survey. The metropolitans included
in the actual survey were Busan, Incheon, and Daegu, and a total of
300 adults aged over 19 by 100 people from each city were
surveyed. The survey was conducted for 18 days from 4th to 21st
of June, 2012 by the Korea Research Company.
As for the method of data collection for empirical analysis, a
personal interview method was used. As mentioned earlier, the
population was male and female adults aged over 19 living in
Busan, Incheon, and Daegu, with the sample being 300 people. The
sampling was done according to the constituent ratio by sex, age,
and region in the statistics of residents as of April 30, 2012.
Assuming random sampling, the maximum sampling error is
75.7% with a condence level of 95%. This indicates the chances
are 95 in 100 that the survey results are within 75.7% points of
the true percentage for the whole population.2
We discussed the survey questions with ve scholars in the
management of technology and reworded the questions based on
their comments to enhance content validity. Next, a pilot test was
performed prior to the main survey to reword the questions based
on the feedback and increase the reliability and validity of the
research. The pilot test was done online by 60 people of the Korea
Research panel.
In the main survey, 300 questionnaires were collected through
personal interviews, and 45 questionnaires among them were
excluded because of many missing values or a lack of consistency.
Finally a total of 255 questionnaires were used in the empirical
analysis.
4.2. Data analysis method
The 7-point Likert scale was used in the questionnaire. In the
questions about the use of the smart grid, the answer of not
2
The total population aged over 19 of the three sampled city is about 7 million.
A general method to determine the required sample size in large population
(greater than 50,000) is Cochran (1977)'s way; n0 z2 p1  p=e2 , where n0 is the
sample size, Z is the Z value (e.g., 1.96 for 95% condence level), p is the estimated
proportion of an attribute that is present in the population, expressed as decimal,
and e is the acceptable sampling error, expressed as decimal. When we do not have
prior information about p, we use p 0.5 to obtain a conservative estimate of the
required sample size.
Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
5
agree at all was equivalent to 1, and strongly agree was
equivalent to 7.
The exploratory factor analysis (EFA) was carried out to
examine whether the measured variables of each factor affecting
the smart grid acceptance had been appropriately designed and
answered, while Bartlett's unit matrix was checked to see whether
the survey data was t for factor analysis. To this end, Bartlett's
test and KaiserMeyerOlkin (KMO) goodness-of-t test were
conducted. The principal component analysis was utilized for the
factor model, and the Varimax method of orthogonal rotation was
used for the factor rotation.
After the EFA, the Structural Equation Modeling (SEM) analysis
was performed by using AMOS 17 to test the hypotheses. The SEM
analysis consisted of two steps: conrmatory factor analysis (CFA)
and path analysis (PA).
5. Empirical analysis
5.1. Exploratory factor analysis
In the EFA, in order to see whether the survey data was t for
factor analysis, Bartlett's sphericity test and KMO's goodness-of-t
test were conducted. In Bartlett's sphericity test, the value was
7419.2 (p o0.001, df 990), showing that the correlation of the
variables was statistically signicant. Further, the KMO value of
0.882 was a high goodness-of-t for factor analysis, meaning that
the raw data was suitable for the factor analysis. Note that the
closer the KMO value is to 1, the higher the signicance the factor
analysis has. Also, the minimum value should not be less than
0.5 and a value greater than 0.8 is good (Huh, 2001:528).
The exploratory factor analysis in this study made use of the
principal component analysis whose purpose was to minimize the
loss of information and enable variables to have a minimum
number of factors. As for the factor rotation, the Varimax method
was used of orthogonal rotation that rotated while maintaining
the independence between factors. In determining the number of
factors, a total of 12 factors were derived from the test of the eigen
value and factor loading. Generally speaking, when factor loading
was more than 0.5, the concentration of the measured variables in
a factor could be seen as good. The factor loadings of each factor
were generally higher than from 0.6 to 0.7, showing high
concentration.
Furthermore, in order to test the internal consistency of the
measured variables, Cronbach's values were obtained to mea-
sure reliability. The range of the values is between 0 and 1, and
generally, when the value is less than 0.3, the concentration
between variables is weak, and when it is more than 0.7, the
concentration between variables can be said as strong. As a result
of the analysis, all the factors except Concerns about Equipment
Malfunction and Performance Deterioration had a value of more
than 0.7, showing high reliabilities. However, the value of the
factor of Concerns about Equipment Malfunction and Perfor-
mance Deterioration was 0.69, close to 0.7, which could be said
to have high reliability as well.
5.2. Conrmatory factor analysis
The CFA has difference with the EFA in that it can model a form
of hypothesis with theoretical knowledge or results and constrain
some elements of the matrix. The reason to conduct CFA after EFA
is that it can identify the outline of factor structure of the latent
variables used in a research model and remove the measured
variables contrary to the validity more thoroughly (Kim, 2008).
CFA is a process to verify whether the concepts in EFA have
a similarity with the research model in the structural equation.
6 C.-K. Park et al. / Energy Policy ()

To check the similarity, the statistical test can be done to the

goodness-of-t of the CFA model.
2, the initial t indexes of the model for CFA were attained as
1841.944 (P 0.00),3 2/df 1.756, RMR0.056, RMSEA 0.055,
GFI 0.781, AGFI 0.745, NFI 0.793, CFI 0.897, IFI 0.899,
PNFI0.707, and PGFI 0.800.4 Since the overall t indexes of
the initial model were deemed unt for the criteria, the measured
variables with the high adjusted index were removed one by one
by the adjusted index analysis. The adjusted index was divided
into the covariance adjusted index and the regression weights
(regression coefcient) where adjusted index was based on. The
goodness-of-t of the model can be improvised by removing the
measured variables and latent variables with the high adjusted
index. However, in order to increase the goodness-of-t, removing
too many variables can spoil the validity of the research model,
even though it attained a high goodness-of-t.
The overall t indexes of the adjusted model in the range not
spoiling the validity of the research model were 2 867.595
(P 0.00), 2/df 1.541, RMR 0.051, RMSEA 0.046, GFI0.850,
AGFI0.813, NFI 0.850, CFI 0.940, IFI 0.942, PNFI0.718, and
Fig. 4. Result of path analysis.
PGFI 0.795. These were tter for the criteria than the t indexes
of the initial model.
When the goodness-of-t indexes out of CFA satised the to have discriminant validity between them (D.U. Kim, 2008:46;
criteria, reliability and validity were veried. First, the signicance K.S. Kim, 2008:372373). As a result of the analysis, all the AVEs of
of the coefcient showing the degree to which latent variables two variables were greater than the square of their correlation
affected measured variables was examined with the critical ratio coefcient, so the validity was veried.
(CR) of unstandardized coefcient. The analysis showed that all CR
values were signicant in condence level of 99%. 5.3. Testing of hypotheses through path analysis
After the signicance of the unstandardized coefcient was
veried, the reliability of each measured variable could be checked By PA,9 the t indexes of the model were gained as 2/
by the standardized coefcient. Here the criterion was that the df 1.633, RMR0.077, RMSEA 0.05, GFI 0.838, AGFI0.806,
standardized coefcient was more than 0.5.5 According to the NFI 0.834, CFI 0.927, IFI 0.928, PNFI 0.734, and PGFI 0.816.
analysis, all the standardized coefcients were more than 0.5, Almost all t indexes except GFI and AGFI met the criteria. Even
therefore it can be said that the measured variables had reliability. the rest two indexes were close to their criteria, so they can be said
Next, the reliabilities of the latent variables should be tested to be acceptable. In addition, the number of measured variables
with the standardized coefcients. To test the reliability, the maintained more than three for the sake of the validity of the
construct reliability,6 also known as the composite reliability, research model.
and the average variance extracted (AVE)7 were veried.8 When According to the analysis (Fig. 4), all the endogenous variables
the construct reliability was more than 0.7 and the AVE was more of the Perceived Usefulness, the Perceived Ease of Use, and the
than 0.5, the reliability of a latent variable was accepted (Fornell et Perceived Risk of the smart grid have a positive or a negative effect
al., 1982). The analysis showed that all the latent variables met the on the Intention to Use the smart grid, and the Perceived Ease of
criteria of the reliability. Use has a positive impact on the Perceived Usefulness. However,
After the test of the reliability of latent variables, the discrimi- the negative relation between the Perceived Usefulness and the
nant validity should be checked. If AVE1 and AVE2 of two latent Perceived Risk (H2, H3) does not have statistical signicance.
factors were greater than the coefcient of determination; i.e., the As for the total effect10 of the three variables, the Perceived Ease
square of their correlation coefcient, the two factors can be said of Use is 0.329, the Perceived Usefulness is 0.614, and the
Perceived Risk is 0.198. Therefore, the Perceived Usefulness has
the greatest inuence on the Intention to Use.
3
2 test is a statistical test method to determine whether or not the samples All the exogenous variables were veried as inuential factors for
which researchers obtained are the same with the model designed by the researchers.
When the p-value of test is greater than the signicance level of 5% at least, the null the endogenous variables. The exogenous variables affecting the
hypothesis that the two covariance matrices are equal can be accepted. However, not Perceived Ease of Use are the Perceived Compatibility and the
a few social science researches using structural equation modeling technique have
p-value close to 0, rejecting the null hypothesis. This is because it is sensitive to the
9
sample size. Thus, the judgment of the goodness-of-t of a model should not rely only Path analysis (PA) is an extension of multiple regression analysis and used
on test, but on other t indexes (Kim, 2008:36, 401). mainly to examine the comparative strength of direct and indirect relationships
4
For a detailed explanation of the goodness-of-t of a model, refer to Hooper among variables. A series of parameters are estimated by solving one or more
et al. (2008). structural equations to test the t of the correlation matrix between two or more
5
Because the explanatory power of error is greater than that of the measured causal models which are hypothesized. It is useful in that it forces researchers to
value of the measured variables when the standardized coefcient is lower than specify the relationships between variables and thus encourages the development
0.7, Fornell et al. (1982) recommended the standardized coefcient should be of logical theories of the processes inuencing a particular outcome (Lleras, 2005).
10
higher than 0.7. However, Bagozzi and Yi (1988) suggested that the standardized As the Perceived Ease of Use has an indirect effect on the Intention to Use via
coefcient should be higher than 0.5 and less than 0.95 (Kim, 2008:51). the Perceived Usefulness, this indirect inuence should be considered as well. Since
6
Construct Reliability Standardized coefficients2 = the value of the effect of the Perceived Ease of Use on the Perceived Usefulness is
Standardized coefficents2 masurement error variance 0.186, and that of the Perceived Usefulness on the Intention to Use 0.614, the
7 2 2
AVE standardized coefficients =standardized coefficents  indirect inuence can be calculated by the multiplication of the two values. In order
masurement error variance to look into the indirect effect in the analysis of structural equation modeling, an
8
In this study, the measurement error variance used the value calculated by additional test is needed. As the standard error of the indirect effect of the
the algorithm of the AMOS program. The measurement error variance can be Perceived Ease of Use on the Intention to Use is 0.054, and the test value is
calculated as (1  [Standardized coefcient]). 2.222, they veried the statistic signicance at condence level of 95%.

Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
 C.-K. Park et al. / Energy Policy ()
Perceived Understanding, with the Perceived Compatibility with the
path coefcient of 0.379 having relatively greater inuence than the
Perceived Understanding with the path coefcient of 0.144.
The Perceived Understanding, the Perceived Power Supply
Reliability, the Perceived Electricity Rate Saving, and the Perceived
Eco-Environment have a statistically signicant effect on the
Perceived Usefulness. The path coefcient values of the Perceived
Electricity Rate Saving, the Perceived Power Supply Reliability, and
the Perceived Eco-Environment are mostly similar, and inter alia,
the value of the Perceived Electricity Rate Saving is the highest
at 0.282.
As for the Perceived Risk, all of the Perceived Cyber Insecurity,
the Perceived EMR Fear, and the Perceived Performance Concerns
have a statistically signicant impact, and the Perceived EMR Fear
has the greatest inuence on the Perceived Risk, having a
path coefcient of 0.411. Perceived Performance Concerns and
Perceived Cyber Insecurity are next in order.
6. Discussion and Conclusions
The existing TAM based on the theory of reasoned action
emphasizes the factors of the perceived usefulness and the
perceived ease of use. However, the acceptance of a new technol-
ogy is bound to be affected by subjective and irrational factors like
emotions and images as well as objective and rational factors. In
this study, consumers' smart grid acceptance factors including an
irrational factor, so-called the perceived risk, were examined with
the existing TAM based on the theory of reasoned action.
In precedent studies, there were some cases considering the
perception of risk in the use of a technology, generally as an external
variable, yet rare are those considering it as an endogenous variable
like the perceived usefulness, the perceived ease of use, and the
intention to use. In this study, in the name of Risk Integrated TAM
(RITAM), a separate endogenous variable called the perceived risk is
added, and as external variables having inuence on this variable,
social/psychological risk, functional/economic risk, and physical risk
factors are reected on the model through examining the perception
of risk in the use of the smart grid more thoroughly.
According to the analysis of structural equation modeling, all
the endogenous variables of Perceived Usefulness, Perceived Ease
of Use, and Perceived Risk of the smart grid have a positive or a
negative effect on the Intention to Use the smart grid. Meanwhile,
the Perceived Ease of Use has a positive impact on the Perceived
Usefulness but the negative relation between the Perceived
Usefulness and the Perceived Risk does not have statistical
signicance.
All the external variables set in the research model were
veried as the factors affecting each endogenous variable.
The perceptions of the understanding of the smart grid and the
compatibility with the existing technology have an inuence on
the recognition of the ease of use of the smart grid and the
improvement of power supply reliability, electricity rate saving,
and environment-friendliness, which affect the perception of the
usefulness of the smart grid. The awareness of cyber security
threats, the fear of electromagnetic radiation, and the concerns
about devices' malfunction and performance deterioration have
effects on the perception of the risk of the smart grid.
As a result of this study, the importance of consumer education
and public relations of the smart grid can be conrmed as well. It
is important to improve consumers' understanding of the smart
grid and to offer consumer-oriented education and public rela-
tions. The analysis shows that the understanding of the smart grid
improves the perceptions of ease of use and usefulness. It is also
important to provide lucid terminology and unexaggerated infor-
mation. Consumers need to be able to easily understand reasons,
Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
7
benets, and impacts of the smart grid deployment. Sources of
information on smart meters or the smart grid can be various as to
be government agencies, research institutes, related industries,
and consumer groups. The exchange of smart grid-related infor-
mation needs to be coordinated not to cause consumers' confusion
and misunderstanding.
Another shortcut to ensure the acceptance of the smart grid
is to mitigate the anxiety about the risk in the use of the smart
grid. Particularly, it is important to reduce awareness on the
worry about electromagnetic radiation from smart meters,
which is shown as the strongest perception of the risk in the
use of the smart grid. Even though domestic media or research
institutions have not intensively dealt the hazard of a smart
meter's electromagnetic radiation yet, consumers feel appre-
hension that the electromagnetic radiation from smart meters
might be serious. In order to dispel this worry, smart meter
distributors or electricity providers need to treat the issue
transparently and proactively. The regulatory agency should
check and conrm that all the already-installed and to-be-
installed smart meters meet the standard for the electromag-
netic radiation, and thereafter provide the standards or guide-
lines on the minimum effort for utility providers to relieve
consumers' uneasiness, which is related to the electromagnetic
radiation of smart meters.
The concerns about equipment malfunction and cyber inse-
curity have also been recognized as a major factor to affect
the perception of smart grid risks. Therefore, it is necessary to
transparently inform the public that smart grid devices und
ergo stringent tests before leaving the manufacturing plant and
are veried by the regulatory body. Moreover, the remedy
procedures for any problems should be informed. It is also
important to strengthen the standard for smart grid privacy
and cyber security. The rules regarding the ownership of con-
sumer data, access limitations, and data use permissions should
be provided.
In addition, in order to improve the compatibility of smart grid
products and services, it is important to enhance the intuitiveness
of a smart grid user interface. As Facebook or Google shows data as
if they are not data, it is necessary that energy data should be
made more enjoyable and easier for smart grid users to interface.
This study has an implication in that it has provided theoretical
and empirical ground, based on which the policies to promote
consumer participation in the deployment of the smart grid can be
developed. Since there are few studies on the policies from the
perspective of the smart grid users, this study will contribute
directly to the development of the strategy to ensure the accep-
tance of the smart grid.
However, this study has interview errors in terms of the
technical nature of the personal interviewer survey. Some
sampling errors and non-sampling errors result from the sample
test. In addition, it has a limitation in content that the survey
was done to only the potential users of the smart grid. Provided
that the perceptions of the current users of smart grid products
and services are compared with those of the potential users, it
will be possible to gain more practical policy suggestions on the
smart grid acceptance factors. Moreover, we will be able to
include other exogenous and endogenous variables for further
studies.
Acknowledgments
This study was supported by the Korea Energy Economics
Institute (RP-12-06). We thank Kwon, Young Sun and Song, Chan
Hoo, professors in KAIST for great comments. Also, comments by
8 C.-K. Park et al. / Energy Policy ()
two anonymous reviewers greatly helped to improve an earlier
version of this manuscript.
Appendix A. Supplementary material
Supplementary data associated with this article can be found in
the online version at http://dx.doi.org/10.1016/j.enpol.2014.03.017.
References
Accenture, 2010. Understanding Consumer Preferences in Energy Efciency.
Accenture, Dublin.
Bauer, R.A., 1960. Consumer behavior as risk taking. In: Hancock, R.W. (Ed.),
American Marketing Association.
Bagozzi, R.P., Yi, Y.J., 1988. On the evaluation of structural equation models. Journal
of the Academy of Marketing Science 16, 7494.
BCG (Boston Consulting Group), 2010. The Smart-Meter Opportunity. BCG, Boston
Brooker, G., 1984. An assessment of an expanded measure of perceived risk. Adv.
Consum. Res. 11, 439441.
Chicagotribune, 2011. Naperville Smart Grid Opponents Sue. Chicagotribune,
Chicago.
Cochran, W.G., 1977. Sampling Techniques. Wiley, New York.
Cox, D.F., 1967. Risk handling in consumer behavior: an intensive of study two case.
In: Cox, D.F. (Ed.), Risk Taking and Informational Handling in Consumer
Behavior. Harvard University Press, London, pp. 3637.
Davis, F.D., 1989. Perceived usefulness, easy of use, and the user acceptance of
information technology. MIS Q. 13 (3), 319340.
Earth2tech, Smart Meter Backlash, Again: This Time in Texas, 2010, Earth2tech; San
Francisco, CA.
EcoAlign, 2010. Separating Smart Grid from Smart Meters? Consumer Perceptions
and Expectations of Smart Grid. EcoAlign, Boston.
EPRI, 2012. Understanding Electric Utility CustomersSummary Report. Technical
Report. EPRI, Palo Alto, CA.
Featherman, M.S., Pavlou, P.A. 2002. Predicting E-services adoption: a perceived
risk facets perspective. in: Proceedings of the Eighth Americas Conference on
Information Systems. pp. 10341046.
Finucane, M.L, Alhakami, A., Slovic, P., Johnson, S.M., 2000. The affect heuristic in
judgements of risks and benets. J. Behav. Decis. Mak. 13, 117.
Fornell, C., Tellis, G., Zinkhan, G. 1982. Validity assessment: a structural equations
approach using partial least squares. In: Proceedings of AMR education's
conference. pp. 405-409.
GE, 2010. Smart Grid Survey: Majority Down Under still wonder. GE, New York.
Gefen, D., Karahanna, E., Straub, D.W., 2003. Trust and TAM in online shopping: an
interacted model. MIS Q. 27 (1), 5190.
GlobalData, 2011. Smart Grid Stakeholder AnalysisEconomic Impact of Smart Grid
Technology Deployments. GlobalData, Los Angeles.
Greentech Media, 2011. Embracing Lifestyle Plans and Social Media. Greentech
Media, Boston.
GridWise Alliance, 2013. 2013 Grid Modernization Index. GridWise Alliance,
Washington, DC.
Harris Interactive, 2010. Consumers Have Little Awareness of Smart Grid and Smart
Meters. Harris Interactive, New York.
Hooper, D., Coughlan, J., Mullen, M., 2008. Structural equation modelling: guide-
lines for determining model t. Electronic Journal of Business Research
Methods 6 (1), 5360.
Huh, M.H., 2001. Statistics Analysis. Bobmunsa, Seoul.
Please cite this article as: Park, C.-K., et al., A study of factors enhancing smart grid consumer engagement. Energy Policy (2014), http:
//dx.doi.org/10.1016/j.enpol.2014.03.017i
IBM, 2011. Knowledge is power. White Paper. IBM, New York.
IEA, 2011. Smart Grid Smart Customer Policy Needs. IEA, Paris.
Jacoby, J., Kaplan, L.B., 1972. The component of perceived risk. In: Venkatesan, M.
(Ed.), Proceedings from Third Annual Conference of the Association for
Consumer Research. University of Chicago. pp. 382393.
Jun, S.Y., Huh, J.H., Kang, S.J., 2003. The effects of risk perception on the relative role
of brand and price in internet shopping mall. J. Consum. Stud. 14 (2), 1943.
Kassarjian, H.H., Robertson, T.S., 1991. Perspectives in Consumer Behavior, 4th ed.
Pretice Hall, New Jersey.
Koufaris M., Hampton-Sosa W., 2002. Customer trust onlne: examining the role of
the experience with the web site, CIS Working Paper Series. Baruch College;
New York.
Kim, D.U., 2008. AMOS A to Z: Structural Equation Modeling for Writing Papers.
Hakhyunsa, Paju.
Kim, H.J., Park, J.H., Park, C.H., 2012. A study on key inuencing factors in smart grid
consumer engagement. Informatiz. Policy 19 (1), 91106.
Kim, K.S., 2008. Structural Equation Modeling Analysis. Hannarae, Seoul.
Kim, S.Y., 2009. Studies of Acceptance Model of New ScienceTechnology. STEPI
Report. Science and Technology Policy Institute, Seoul.
Lee, M.C., 2009. Factors inuencing the adoption of internet banking: an integration
of TAM and TPB with perceived risk and perceived benet. Electron. Commer.
Res. Appl. 8(3), 130141.
Li, Y.H., Huang, J.W., 2009. Applying theory of perceived risk and technology
acceptance model in the online shopping channel. World Acad. Sci. Eng.
Technol. 53.
Lleras, C., 2005. Path analysis. Encycl. Soc. Meas. 3, 2530.
Oracle, 2009. Turning Information into Power Report: Moving Toward the Smart
Grid. Oracle, California.
Pike Research, 2010. Smart Grid Consumer Survey. Pike Research, Boulder.
Roselius, T., 1971. Consumer rankings of risk reduction methods. J. Market. 35,
5661.
Sanayei, A., Bahmani, E., 2012. Integrating TAM and TPB with perceived risk to
measure customers' acceptance of internet banking. Int. J. Inf. Sci. Manag.,
2537
SGCC (Smart Grid Consumer Collaborative), 2010. Consumer Voice: Results of
Baseline Focus Groups. SGCC, Atlanta.
Shim, J.S., 2009. Trust in nuclear power plant, perceived risk and benet, and
acceptance. Korean Policy Stud. Rev. 18 (4), 213238.
Siegrist, M., Keller, C., Kastenholz, H., Frey, S., Wiek, A., 2007. Laypeople's and
experts' perception of nanotechnology hazards. Risk Anal. 27 (1), 5969.
Stone, R.N., Gronhaug, K., 1993. Perceived risk: further considerations for the
marketing discipline. Eur. J. Market. 27, 3950.
Taylor, S., Todd, P.A., 1995. Understanding information technology usage: a test of
competing models. Inf. Syst. Res. 6 (2), 145176.
Venkatesh, V.A., Bala, H., 2008. Technology acceptance model3 and a research
agenda on interventions. Decis. Sci. 39 (2), 273315.
Venkatesh, V.A., Davis, F.D., 2000. A theoretical extension of the technology
acceptance model: four longitudinal eld studies. Manag. Sci. 46 (2), 186204.
Venturebeat, 2009. PG&E Lawsuit Spreads Down Smart Grid Supply Chain. Ven-
turebeat, San Francisco, CA.
Vincent, M., Zikmund, W.G., 1976. An experimental investigation of situational
effects on risk perception. Adv. Consum. Res. 3, 125129.
Wu, J.H., Wang, S.C., 2005. What drives mobile commerce? An empirical evaluation
of the revised technology acceptance model. Inf. Manag. 42, 719729.
Yoon, Y.B., Park, K.C., Ahn, S.M., Lee, J.Y., Lee, B.G. 2011. Study on the effects of
perceived riskness on smartphone application acceptance. In: Proceedings of
the Korean Society for Internet Information Summer Conference. pp. 291292.
Yann, Y., Jung, M.K., 1999. Perceived risks and consumer characteristics in CATV
homeshopping. Korean J. Advert. 10 (2), 115139.
Zpryme, 2011. The New Energy Consumer. Zpryme, Mountain City.