Академический Документы
Профессиональный Документы
Культура Документы
ABSTRACT
Since sustainable development and energy efficiency are being pursued in many countries, use of smart meters is
considered a first step in allowing residential consumers to control their energy consumption remotely. Despite the
growing interest in smart meters, many consumers are still sceptical due to their limited awareness, knowledge, and
understanding of these devices. Social acceptability is a crucial determinant of the failure or success of plans by policy
makers to implement smart meter systems. Therefore, this study investigated the factors that affect consumer acceptance
of smart meters in residential buildings. Data obtained in a survey of energy consumers in Indonesia were analysed by
Structural Equation Modeling (SEM) to determine the interacting factors in consumer acceptance of smart meters. The
findings of this research reveal several factors that affect consumer decisions related to their acceptance of smart meters
in residential buildings.
INTRODUCTION
Energy efficiency and sustainable development are important global issues and are major concerns in most countries. In
terms of energy efficiency, the resource use must be rationalized and overall energy system must be monitored. It causes
simply monitoring consumption is no longer sufficient. In response of this demand and respect to energy efficiency, this
moment is perfectly aligned with the changes to propose smart meter as an automatic remote metering and sub metering
of energy consumption. On the other hand, several studies show that the public is still confused about what can be
achieved through smart meters application. This phenomenon reinforces the notion that the technology availability does
not guarantee its widespread adoption.
Related to smart meters issue, many researches were conducted and most of them used a detailed technical
perspective or general conceptual perspective as the initial approach. Whereas, social acceptability is an important
determinant of the success or failure of policy maker decisions. The first step in smart meter development is to ensure
consumer acceptance of smart meters in their residences (McKenna et al., 2012). Smart meter deployment could be easily
refused if consumers underestimate their benefits or overestimate their problems. To design an appropriate policy, a clear
understanding of consumer concerns and preferences about smart meter deployment is crucial. For utility companies,
measuring consumer tendencies to adopt smart meters is crucial for accurate forecasting and effective targeting of high-
tech product and services.
This research investigates factors that affect consumer acceptance of smart meters installed in residential buildings in
Indonesia. A research model is developed based on technology acceptance theory. The findings of this study improve
understanding of consumer perceptions and behaviors related to smart meters. The data can be used by energy utility
companies in Indonesia and by policy makers when developing strategies for implementing a “one-size-fits-all” program
related to smart meter use in residential buildings.
1
J.-S. Chou, N. Yutami
However, a potential risk of smart meter installation is that consumers may object to increased electricity bills
resulting from improved billing accuracy or from the use of flat billing compared to bills based on manual meter reading.
Manual meter reading is subject to human error in recording, and since smart meters are digital devices, they would
avoid this problem. However, the increased billing accuracy could result in increased bills for consumers who have been
previously undercharged. Another potential risk of implementing smart meters is violation of consumer privacy. Detailed
billing data could reveal personal data of consumers such as when they are home, their habits and routines, and how they
use their appliances. For instance, such data may allow hackers to disconnect power to large areas remotely and to obtain
decrypted personal information.
Consumer Acceptance of New Technology
In terms of technology acceptance concerns, the most frequently used model of user acceptance of information
technology (IT) is the technology acceptance model (Krishnamurti et al., 2012). The technology acceptance model (TAM)
explains the acceptance of information technology in performing a task and identified two beliefs that influence the
information system (IS) usage: Perceived usefulness (PU) and ease of use (PEU). Prior studies of TAM regarded actual
IT usage as an indicator of acceptance. Here, however actual usage was not used as an indicator of acceptance because
smart meters have not been widely implemented by energy consumers in the research setting (i.e., Indonesia).
Previous studies in related management disciplines (e.g., organizational behavior, marketing, and information
systems) have examined the relationship among factors such as user involvement, satisfaction, usage, and behavioral
intentions. Hence, this study focused on behavioral intention to use the system rather than actual system usage and
posited it as an indicator of technology acceptance. Figure 1 shows that, compared to TAM, the model developed in this
study is more comprehensive and provides a better understanding of the factors that affect behavioral intention to use
smart meters. The model also incorporated variables related to social issues (social influence/subjective norms),
individual issues (energy tariff/cost, privacy and safety), and technical issues (program content/feedback features and
technological complexity). From the initial proposed model, we derived 20 hyphoteses.
ETC 1
ETC 2
H16
ETC 3 Energy Tariff/Cost PERCEIVED RISK
ETC 4
ETC 5
H17 PR 1 PR 2 PR 3 PR 4 PR 5 PR 6 H11
PSC 1
ATB 1 ATB 2 ATB 3 ATB 4 ATB 5
PSC2
PSC3 Privacy/Savety Concern H9 H12
PSC4
PSC5
H8
ATTITUDE TOWARDS
UES 1 UES 2 UES 3 H4 BEHAVIOR
H10
H13
SIN 1
SIN2 H18
Social Influence/ PERCEIVED EXPECTED USER EXPECTED
SIN3 H15
SIN4 Subjective Norms USEFULNESS H1 SATISFACTION
SIN5
H6 H14
H7
TCC 1
TCC2 H20
TCC 3
Technological Complexity PERCEIVED EXPECTED
TCC4
EASE OF USE
TCC5
2
J.-S. Chou, N. Yutami
Perceived Risk (PR) or uncertainty affects the confidence of decision makers. Risky situations include those in
which the probabilities of outcomes are unknown, regardless of whether the outcome is unknown. Prior research argued
that perceived risk negatively impacts consumer satisfaction (Udo et al., 2010).
H8: Perceived risk significantly affects user expected satisfaction
H9: Perceived risk significantly affects perceived expected usefulness
H10: Perceived risk significantly affects perceived expected ease of use
H11: Perceived risk significantly affects attitude
H12: Perceived risk significantly affects behavioral intention to use
User satisfaction is a mediating factor in the services or system performance improvement perceived by users and a
successful information system. Other researcher defined user satisfaction as “the degree of meeting the information
needed by users”. This study added the term “expected” for a precise and clear differentiation between potential adopters
and current users. Prior research found that satisfaction affects behavioral intention to use technology (Son et al., 2012).
H13: User expected satisfaction significantly affects attitude towards behavior
H14: User expected satisfaction significantly affects behavioral intention to use a technology
The capability to measure attitudes toward a new system is a central feature of TAM. Prior researcher defined
attitude as “an individual’s positive or negative feelings (evaluative effect) about performing the target behavior” and
suggested that attitude can be affected by both psychological factors and the situations encountered. Other researchers
have found that attitude positively affects behavioral intention to use a technology (Liu et al., 2013).
H15: Attitude towards behavior significantly affects intention to use a technology
Behavioral intention to use is defined as a measure of the strength of the intention to perform a specific behavior; the
behavior considered here is the use of information technology. This research model uses behavioral intention as an
individual-level variable to measure the acceptance of smart meters. The literature suggested that the determining factors
in the behavioral intention to use a system must be analysed to identify their role in the successful implementation of
information systems.
Determinants of Success in Smart Meter Device Implementation
Energy tariffs/costs
Mah et al. (2012) found that energy consumers are sensitive to tariff increases related to the policy of smart meter
deployment in their country (Mah et al., 2012). This study argues that energy tariffs/costs have a significant positive
effect on perceived risk and a significant negative effect on technology adoption.
H16: Energy tariffs/costs significantly affect perceived risk
Program contents/features
Studies show that user behavior toward technology is affected by feature, which is defined as “vendor-created software
tools designed to complete tasks requested by user”. Stromback et al. (2011) argued that program content/ features can
affect consumer behavior related to smart meters (Stromback et al., 2011).
H19: Program contents/features significantly affect perceived expected usefulness
Technological complexity
Technology complexity is defined as “the individual's perception of the degree of difficulty involved in understanding
and using new technology”. Since most users are unfamiliar with smart meters, technological complexity is usually the
main obstacle to smart meter adoption. Empirical data show that complexity directly affects perceived ease of use (Son et
al., 2012).
H20: Technological complexity significantly affects perceived expected ease of use
3
J.-S. Chou, N. Yutami
4
J.-S. Chou, N. Yutami
The questionnaire was initially written in English and translated into Indonesian by native speakers. To avoid
misconceptions, the orientation of the 7-point Likert scale was applied uniformly; low scores represented negative
settings while high scores represented favorable situations. The appropriatess of the measurements were also evaluated
in a pilot study performed in 100 Indonesia students attending National Taiwan Universities of Science and Technology.
The pre-test result showed that the constructs and their indicators were easily understood and answered. Out of 400
questionnaires, 309 questionnaires were returned; of these, eight contained incomplete or incomprehensible responses
and were excluded. Therefore, the response rate was 77.25%, and 75.25% of the questionnaires were valid. Among valid
questionnaires retrieved from each surveyed area, 22.9% from respondent who living in Sumatra, 51.5% were from Java,
5% were from Kalimantan, 9.6% were from Sulawesi, 2% were from Papua, and 9% were from Bali. Most (58.5%)
respondents were male, and 41.5% were female. Most (43.9%) were aged 21-30 years.
The respondents were engaged in various occupations: 8.3% were students, 8.6% from commercial service field,
37.9% work in industrial and manufacturing, 5.3% work in military or government, 5.3% were homemakers, 1.3% work
in forestry, fishing, animal husbandry fields, 6.6% were entrepreneurs, 1.7% were retired, and the rest (approximately
24.9%) worked in other fields, including a doctor and employees of a public company. Most respondents had a senior
high school (about 32.2%) or university or college (about 55.8%) education. Most (about 57.1%) had no complaint about
current electricity service, and the most common (approximately 32.6%) complaints were frequent outages and unstable
voltage. Additionally, 67.8% respondent had never seen/heard about smart meters, 20.3% had seen/heard about smart
meters but did not understand their function, 5% were general , 5.3% were familiar with smart meters, and 1.7% were
very familiar with smart meters. Most (approximately 97%) had no experience with smart meters. All respondents who
indicated that they were familiar or very familiar with smart meters or indicated that they had experience using smart
meters were employees of the national electricity company of Indonesia. Compared to the studies about sampling data
reported in the literature, this study had a comparable sample size (4 times of the indicators/variables).
Confirmatory Factor Analysis and Model Modification
The data analyses in this study were performed using structural equation modeling (SEM), a family of statistical
methodologies for using a confirmatory approach to analysing a structural theory bearing on some phenomenon. The
SEM is also an emerging statistical tool in the social sciences (Kline, 2011). The adequancy of the measurement model
was assessed by reliability test, confirmatory factor analysis (i.e., convergent and discriminant validities), and model fit
test.
Reliability of results
Before model assessment, the internal consistency of the results was tested by calculating Cronbach α (Hu and Bentler,
1999). The Cronbach α reliability coefficient obtained for each instrument ranged between 0.867 and 0.967 (Table 2),
which is well above the minimum acceptable threshold of 0.70. Thus, the measurement instruments were deemed
reliable gauges of the constructs.
Convergent validity
The relative convergent validity among item measures can be estimated in several ways. Fornell and Larcker (1981)
suggested using item reliability for each measure, composite reliability for each construct, and the average variance
extracted (AVE) to assess the convergent validity of the items measured (Fornell and F.Larcker, 1981). For a given
measure, item reliability is confirmed by a factor loading of 0.70 or greater, which indicates a well-defined structure. To
ensure adequate composite reliability, a value of 0.70 or higher is recommended. The composite reliability values
obtained here ranged from 0.872 to 0.965, which far exceeded the recommended value of 0.7. The AVE is a measure of
the overall variance attributed to the construct relative to the variance attributed to measurement error. Each construct
should have an AVE of at least 0.50, the point above which the variance captured by the construct exceeds the variance
due to measurement error. Table 3 shows that AVE values ranged from 0.763 to 0.874, considerably above the threshold
of 0.50. Factor loading, Cronbach , AVE, and CR coefficients were used to assess the reliability and validity of the
proposed model. In the originally proposed model, some coefficients were lower than the recommended limit due to the
high complexity of the model, and some paths and indicators were insignificant. After modification, All AVE, CR and
Cronbach coefficients obtained for the constructs were within recommended ranges (Table 2).
Discriminant validity
A given construct has discriminant validity when the variance between that construct and another construct in the model
is smaller than the variance among the indicators for that construct (Hair et al., 2010). Discriminant validity was assessed
by comparing the square root of the AVE for a given construct with the correlations between that and all other constructs
(Son et al., 2012). If the square roots of the AVEs are greater than the absolute values of the off-diagonal elements in the
corresponding rows and columns of the correlation matrix, the construct has a stronger correlation with its own indicators
than with the other constructs in the model.
5
J.-S. Chou, N. Yutami
ETC 2 0.820
Energy tariff/ cost 0.773 0.872 0.867
ETC 5 0.935
PSC 3 0.989
Privacy & safety concerns 0.805 0.891 0.880
PSC 4 0.795
SIN 2 0.885
SIN 3 0.931
Social influence/subjective norms 0.871 0.964 0.963
SIN 4 0.962
SIN 5 0.954
PCF 3 0.892
Program contents/features 0.817 0.899 0.900
PCF 4 0.916
TCC 2 0.934
TCC 3 0.949
Technological complexity 0.874 0.965 0.967
TCC 4 0.925
TCC 5 0.932
PR 2 0.930
PR 4 0.959
Perceived risk 0.838 0.954 0.961
PR 5 0.933
PR 6 0.834
PEU 4 0.858
Perceived expected usefulness + PEEOU 4 0.933
Perceived expected ease of use PEEOU 5 0.864 0.767 0.929 0.928
PEEOU 6 0.845
UES 1 0.894
User Expected Satisfaction UES 2 0.945 0.831 0.937 0.946
UES 3 0.895
ATB 2 0.926
Attitude towards behavior + ATB 3 0.947 0.763 0.941 0.948
Behavioral intention to use ATB 4 0.920
BIU 3 0.779
BIU 5 0.780
The original model did not satisfy the discriminant validity test. The correlations between some constructs were
larger than the square roots of their AVE values. Hence, the constructs were merged. After confirming its discriminant
validity, the modified model showed increased goodness of fit. In Table 3, the diagonal elements of the matrix are the
modified AVE values, and the elements below the diagonal are the correlation coefficients. The square root value of each
diagonal element in the matrix is greater than the absolute values of the off-diagonal elements in its row and column,
which confirms its discriminant validity. Figure 2 shows the indicators, constructs, and paths remaining after
modification of the model. Because of the merging process, the proposed hypotheses differed between the original model
and the final model. The final model retained 11 of the original 20 hypotheses.
6
J.-S. Chou, N. Yutami
Model fit
In this study, eight different indices were used to measure model fit: the chi-square test (χ2/degree of freedom (df)), the
goodness of fit index (GFI), the adjusted goodness of fit index (AGFI), parsimony-goodness of fit index (PGFI),
incremental fit index (IFI), the root mean square error approximation (RMSEA), the comparative fit index (CFI), and the
Tucker Lewis index (TLI). These indices were chosen because of their relative stability and insensitivity to sample size.
In these eight indices, the criteria for a good fit are as follows: χ2/df<5 (Mokhtarian and Ory, 2009), GFI<0.8, AGFI<0.7,
PGFI<0.5, CFI>0.80, TLI>0.80, and RMSEA<0.1 (Hair et al., 2010). The AGFI is set to < 0.7 because it is adjusted to
GFI. The AGFI values are typically lower than GFI values in proportion to model complexity.
ETC 2
Energy Tariff/Cost
H7'
ETC 5
PERCEIVED RISK
H 2'
Program Contents/ H10'
PCF 3
Features Available PEU + PEEOU
PCF 4
H11'
TCC 3
Technological Complexity
TCC4
TCC5
Perceived risk did not significantly affect attitude towards behavior and behavioral intention to use. Therefore, H5'
was rejected. User expected satisfaction significantly (p<0.001) affected attitude towards behavior and behavioral
intention to use, which supported H6'. Energy tariff/cost (p<0.001) and privacy/safety (p<0.001) significantly affected
perceived risk, which supported H7' and H8'. Social influence/subjective norms significantly (p<0.01) affected perceived
expected usefulness and perceived expected ease of use, which supported H9'. Program content/features significantly
(p<0.001) affected perceived expected usefulness and perceived expected ease of use, which supported H10'.
7
J.-S. Chou, N. Yutami
Technological complexity significantly (p<0.05) affected perceived expected usefulness and perceived expected ease of
use, which supported H11'.
REFERENCES
CHEUNG, R. & VOGEL, D. 2013. Predicting user acceptance of collaborative technologies: An extension of the
technology acceptance model for e-learning. Computers & Education, 63, 160-175.
FORNELL, C. & F.LARCKER, D. 1981. Evaluating Structural Equation Models with Unobservable Variables and
Measurement Error. Journal of Marketing Research, 18, 39-50.
HAIR, J. F., BLACK, W. C., BABIN, B. J. & ANDERSON, R. E. 2010. Multivariate Data Analysis: A global
perspective, New Jersey, Pearson Education International.
HU, L.-T. & BENTLER, P. M. 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria
versus new alternatives. Structural Equation Modeling, 6, 1-55.
KLINE, R. B. 2011. Priciples and Practice of Structural Equation Modeling The Guilford Press.
KRISHNAMURTI, T., SCHWARTZ, D., DAVIS, A., FISCHHOFF, B., DE BRUIN, W. B., LAVE, L. & WANG, J.
2012. Preparing for smart grid technologies: A behavioral decision research approach to understanding
consumer expectations about smart meters. Energy Policy, 41, 790-797.
LIU, W., WANG, C. & MOL, A. P. J. 2013. Rural public acceptance of renewable energy deployment: The case of
Shandong in China. Applied Energy, 102, 1187-1196.
MAH, N., VAN DER VLEUTEN, D., JOHANNES MARINUS, J., HILLS, P. & TAO, J. 2012. Consumer perceptions of
smart grid development: Results of a Hong Kong survey and policy implications. Energy Policy, 49, 204-216.
MCKENNA, E., RICHARDSON, I. & THOMSON, M. 2012. Smart meter data: Balancing consumer privacy concerns
with legitimate applications. Energy Policy, 41, 807-814.
MOKHTARIAN, P. L. & ORY, D. T. 2009. Structural Equations Models. In: EDITORS-IN-CHIEF: ROB, K. &
NIGEL, T. (eds.) International Encyclopedia of Human Geography. Oxford: Elsevier.
ROSSITER, J. R. & BRAITHWAITE, B. 2013. C-OAR-SE-based single-item measures for the two-stage Technology
Acceptance Model. Australasian Marketing Journal (AMJ), 21, 30-35.
SON, H., PARK, Y., KIM, C. & CHOU, J.-S. 2012. Toward an understanding of construction professionals' acceptance
of mobile computing devices in South Korea: An extension of the technology acceptance model. Automation in
Construction, 28, 82-90.
STROMBACK, J., DROMACQUE, C. & YASSIN, M. H. 2011. The potential of smart meter enabled programs to
increase energy and systems efficiency: a mass pilot comparison. VaasaETT, Global Energy Think Tank.
European Smart Metering Industry Group
UDO, G. J., BAGCHI, K. K. & KIRS, P. J. 2010. An assessment of customers’ e-service quality perception, satisfaction
and intention. International Journal of Information Management, 30, 481-492.