Energy Policy 87 (2015) 240249

Contents lists available at ScienceDirect

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

Determinant factors of residential consumption and perception

of energy conservation: Time-series analysis by large-scale
questionnaire in Suita, Japan
Keishiro Hara a,n, Michinori Uwasu a, Yusuke Kishita b, Hiroyuki Takeda c
a
Osaka University, Center for Environmental Innovation Design for Sustainability, 2-1, Yamada-oka, Suita, Osaka 565-0871, Japan
b
National Institute of Advanced Industrial Science and Technology (AIST), 1-2-1 Namiki, Tsukuba, Ibaraki 305-8564, Japan
c
Osaka University, Graduate School of Engineering, 2-1, Yamada-oka, Suita, Osaka 565-0871, Japan

H I G H L I G H T S

Questionnaire was conducted to households in two years for time-series analysis.

We analyzed residential energy consumption and perception of savings in households.

Determinant factors for consumption and perception of savings were identied.

Households being wasteful of energy are also found willing to cut consumption.

Policy intervention could affect consumption pattern and perception of savings.

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

Article history: In this study, we examined determinant factors associated with the residential consumption and per-
Received 27 April 2015 ception of savings of electricity and city gas; this was based on data collected from a large-scale ques-
Received in revised form tionnaire sent to households in Suita, Osaka Prefecture, Japan, in two different years: 2009 and 2013. We
31 July 2015
applied an ordered logit model to determine the overall trend of the determinant factors, and then we
Accepted 12 September 2015
Available online 24 September 2015
performed a more detailed analysis in order to understand the reasons why the determinant factors
changed between the two periods. Results from the ordered logit model reveal that electricity and gas
Keywords: consumption was primarily determined by such factors as household income, number of family mem-
Energy consumption bers, the number of home appliances, and the perceptions of energy savings; there was not much dif-
Perception of energy savings
ference between the two years, although in 2013, household income did not affect the perception of
Questionnaire
energy savings. Detailed analysis demonstrated that households with high energy consumption and
Ordered logit model
Determinant factors those with moderate consumption are becoming polarized and that there was a growing gap between
consumption behavior and the perception of conservation. The implications derived from the analyses
provide an essential insight into the design of a municipal policy to induce lifestyle changes for an en-
ergy-saving society.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction shows why, in Japan, it is essential to pursue energy conservation

Change in lifestyle is a key component to saving energy and
becoming a low-carbon society. As of 2011, households accounted
for about 14.2% of the total energy consumption in Japan; this has
more than doubled since the 1970s (Ministry of Economy, Trade
and Industry, 2013), while the energy consumed by the industrial
sector has almost stabilized due to their conservation efforts. This
n
Corresponding author.
E-mail address: hara@ceids.osaka-u.ac.jp (K. Hara).
at the household level, and we note that the household level is
recognized as one of the indispensable targets of energy policy in
other countries, as well (Saidur et al., 2007; Wang et al., 2011).
Such changes in lifestyle can be enhanced by effective policies,
and citizens' attitudes have traditionally been important factors
when determining energy policy (Viklund, 2004). A thorough
understanding of behavioral patterns and perceptions about en-
ergy consumption and conservation is the basis for effective policy
formations and ultimate changes in lifestyle. Past studies have
considered which factors are essential for affecting the public's

http://dx.doi.org/10.1016/j.enpol.2015.09.016
0301-4215/& 2015 Elsevier Ltd. All rights reserved.
 K. Hara et al. / Energy Policy 87 (2015) 240249
energy efciency and attitudes towards energy conservation (At-
tari et al., 2010; Barr et al., 2005; Becker et al., 1981; Gatersleben
et al., 2002; Gromet et al., 2013; Martinsson et al., 2011; Mizobuchi
and Takeuchi, 2013; Moll et al., 2005; Ouyang and Hokao, 2009;
Sweeney et al., 2013; Wang et al., 2011). For example, Wang et al.
(2011) explored factors related to electricity saving behavior, such
as economic benets, policy and social norms and past experi-
ences, by conducting questionnaire survey in China. Mizobuchi
and Takeuchi (2013) examined the inuences of nancial factors,
such as income and rewards, and non-nancial factors, such as
social norms, on electricity saving behavior by looking into ran-
dom sample of Japanese households. Other studies have explored
policy measures that can induce changes in behavior, such as by
changing pricing and providing information (Albadi and El-Saa-
dany, 2008; Herter, 2007; Ueno et al., 2006; Vine et al., 2013).
However, there have been few large-scale surveys of municipal
households, aimed at systematically discovering attitudes and
perceptions of energy conservation. Furthermore, no studies have
comprehensively performed time-series analyses of changes in
energy consumption and perceptions of conservation in Japan,
although it has been claimed that the Great East Japan Earthquake
that occurred in 2011 could have inuenced behavior, motivation,
and perception of energy consumption and conservation (Kosugi,
2014). We hypothesized that these factors would have been in-
uenced by that earthquake. Following the Great East Japan
Earthquake in 2011, due to the stoppage of nuclear power plants,
there have been shortages of electricity in Japan. This has led to
required electricity conservation, with a specic target of 10% in
the summer of 2012, within the Kansai Electric Power Company's
service area, in which the city of Suita, the case study area of this
study, is located. This could have affected energy consumption and
perceptions of conservation. While some studies have attempted
to delve into the possible impacts of the intervention after the
earthquake (Kosugi 2014; Nishio and Ofuji, 2014), to the best of
our knowledge, no studies have performed substantiated analyses
of any changes in the determinant factors.
In this study, we examine the determinant factors associated
with the residential consumption and perception of conservation
of electricity and city gas (hereinafter gas); this was based on data
collected from a large-scale questionnaire distributed to house-
holds in Suita, Osaka Prefecture, Japan, in two different years:
2009 and 2013. We consider the consumption of electricity and
gas to be an indicator for energy consumption at household level,
and we investigate the factors and attributes of the individual
households that could potentially determine consumption and
perception of savings or willingness to conserve. We then compare
the results of the 2009 and 2013 surveys and explore possible
changes in the determinant factors affecting consumption pattern
or perception of savings over the two-year period. We also address
the policy implications of these results. The information and im-
plications derived from this study provide an essential insight into
the design of a municipal policy that can induce lifestyle changes
that result in an energy-saving society, and can be the basis for
detailed analyses in the future as to how the people's energy
consumption and perception of savings could have been affected
after the policy intervention of the government of Japan following
the earthquake.
2. Methods
2.1. A large-scale questionnaire in Suita
Suita, the location of our study, is the commuter town located
in the northern part of Osaka Prefecture. Its population is around
0.36 million, its area is 36 km2, and there were about 0.16 million
241
households in 2013. Its population has been slightly increasing just
recently while many municipalities in Japan see a declining po-
pulation. It is therefore an urgent task for the city to formulate
policies to pursue energy conservation in households particularly
in the context of combating climate change (City of Suita, 2014).
In close collaboration with the City of Suita, we used the reg-
ister of residents and randomly chose households in all parts of the
city for participation in the survey: 10,000 households in 2009 and
4000 in 2013. The numbers of households selected were different
due to the budget in each scal year, but we ensured that condi-
tions were the same so that the results could be compared. The
questionnaire was conducted between October and November in
2009 and between February and March in 2013. Note that the
questionnaire contained the same questions in both years 2009
and 2013, with the exception that in 2013, a few items were added
to assess the possible impact of the policies imposed following the
Great East Japan Earthquake in 2011 on the behavior and per-
ceptions regarding energy consumption and conservation.
In both years, the questionnaires were mailed to the selected
households. The response rates were 44.5% (4448) in 2009 and
31.1% (1245) in 2013. These rates were sufciently high that we
considered the results sufcient for analysis.
2.2. Items in the questionnaire
The questionnaire was designed to elicit comprehensive re-
sponses about behavior and perceptions regarding energy con-
sumption and conservation, awareness about new energy sources,
including renewable energy, and how these items related to the
various attributes of each household. Table 1 lists the main items
that are contained in the questionnaire. The main items were
questions about the attributes of the chosen households (Q1Q4),
the consumption behavior and perceptions about energy and re-
sources (Q5 and Q6), the use of energy related to transportation
(Q7 and Q8), practices and awareness related to energy con-
servation (Q9 and Q10), and consciousness of new energy sources
and energy-saving equipment (Q11Q15).
Questions Q1Q4 asked about the attributes of the households,
including the location of the residence, the type of residence, the
composition of the family, and the household income. Questions
Q5 and Q6 asked about the types of energy and resources used,
including electricity, gas, kerosene, water, and solid waste. Al-
though the primary target of this survey was energy, we also
collected data on water consumption and solid waste discharge for
comparison in future study.
As discussed above, the questions were the same in 2009 and
2013, with the exception for the addition of questions related to
impacts due to policy changes following the Great East Japan
Earthquake that occurred in 2011; these additional questions were
added to Q9 as sub-items. For example, one of these asks each
household about energy-saving measures and practices that were
adopted following the 2011 earthquake.
The questions were multiple choice in order, as exemplied in
Q4 of Table 1, which asks the annual household income (i.e., 1: less
than two million yen. 2: between two million yen and ve million
yen, 3: between ve million yen and 10 million yen and 4: more
than 10 million yen).
Although the data derived from the questionnaire can be used
for various types of analyses, in this study, we focus on clarifying
the determinant factors for energy consumption and perceptions
about energy conservation; we do this by analyzing their re-
lationships with the attributes of each household and delving into
any possible changes in such factors between the values in 2009
and those in 2013. The data most valuable for this were those
obtained from Q1 to Q6. Table 2 illustrates the details of Q6, which
asks about the current consumption practices and perceptions
242 K. Hara et al. / Energy Policy 87 (2015) 240249

Table 1
Main items in the questionnaire for 2013.

Types/attributes of question Question

Q1 Address and commuting information Please write your place of residence (address). (You may omit the street number.) Also, if you are studying
or working, please write the name of your school or workplace.
Q2 Type of residence Choose the type of your residence. (Select only one answer by circling the appropriate number.)
Q3 Composition of family What is the composition of your family? (Write the number of people in your family, their ages, and
indicate who answered the questionnaire.)
Q4 Household income What is the approximate total annual income of your household?
1. Less than two million yen.
2. Between two million yen and ve million yen.
3. Between ve million yen and 10 million yen.
4. More than 10 million yen.
Q5 Status of electrical equipment (numbers, etc.) Describe the condition of the electrical equipment in your home.
Q6 Status of energy consumption and perception of What is the approximate energy consumption of your home per month (water supply per two months)?
savings And by how much could you reduce consumption if you really tried?
Q7 Usage of personal vehicles Regarding your personal vehicle, what is the engine displacement, fuel type, approximate fuel con-
sumption, and approximate total distance traveled per year?
Q8 Means of transportation When you go out, what means of transport do you use most often, and what types of public transportation
are available near your place of residence?
Q9 Energy-saving practices (1) What energy-saving measures do you take in your home?
(2) If you do not take any, what is the reason?
(3) Circle the energy-saving measures that you became particularly aware of after The Great East Japan
Earthquake. (You may select multiple answers.)
Q10 Awareness of new types of energy and energy-saving Please indicate your familiarity with new energy sources and energy-saving equipment, and indicate
equipment whether you use it. (Circle one answer for each item.)
Q11 Concerns about introducing new energy sources If you have adopted a new energy source (e.g., solar heating and/or photovoltaic power) to your home,
what were your concerns? (Circle all appropriate answers.) If you havent done so, then answer in terms of
future possibilities.
Q12 Assistance by local government to promote adoption In order to promote more widespread adoption of new energy sources (e.g., solar heating or photovoltaic
of new energy sources power), what kind of initiatives should the city take, in your opinion? Circle one answer. Choose the one
that is the most appropriate for each item. If you do not currently use new energy sources, then answer in
terms of future possibilities.
Q13 Willingness to adopt new energy sources when a In Q11, when asked about your plans to purchase any of the choices provided, if you answered (3) I am
subsidy is available interested, but I do not plan to buy it, or (4) I am not interested, then would you consider adopting that/
those new energy source(s) if a government subsidy was offered? (Circle the number of the appropriate
answer.)
Q14 Willingness to pay If you answered 1. I would seriously consider it or 2. I would probably think about it to Q13, what is
the maximum amount you would pay (supposing that government subsidies were offered) for the
adoption of any item of equipment?
Q15 Comments about new energy sources or energy-saving Please comment freely about your expectations regarding new energy sources and energy-saving pro-
equipment ducts, and your impressions if you have tried any of them.

Note: Q9 is further divided into seven subquestions (Q9-1 to Q9-7); Q9-1 is shown in the table.

Table 2 believes it could reduce consumption relative to the current con-

Details of Q6: (a) status of consumption of energy (electricity and gas), and sumption level.
(b) perceptions of savings.

Q6-a: What is the approx- Q6-b: By how much could you 2.3. Statistical analyses
imate energy consumption reduce consumption relative to
of your home per month? present consumption level if you 2.3.1. Analyses using the ordered logit model
really tried?
We used regression analysis to examine which household at-
Electricity 1. Less than 200 kWh 1. 13% tributes were likely to correlate with amount of electricity con-
2. 201300 kWh 2. 46% sumed and the perception of potential energy conservation. As
3. 301400 kWh 3. 710% explained above, the questions were multiple choice. For example,
4. 401500 kWh 4. More than 11% -(Details: )% if the consumption of electricity per month was estimated to be
5. More than 501 kWh 5. Impossible to cut further
-(Details: ) kWh
between 301 kWh and 400 kWh, then the respondent would se-
Gas 1. 15 m 3
1. 13% lect option 3. Likewise, if the respondent believes that household
2. 610 m3 2. 46% electricity consumption could be reduced by 46%, then the re-
3. 1115 m3 3. 710% spondent would select option 2. We carried out a regression
4. 1620 m3 4. More than 11% -(Details: )%
analyses between the option numbers selected for Q6 and those
5. More than 21 m3 -(De- 5. Impossible to cut further
tails: ) m3 selected for the questions regarding household attributes (Q1Q4).
6. No consumption Because the explained variables in our data set are ordered and
discrete, we consider that the ordered logit model was the most
appropriate method for this analysis (Greene, 2008).
regarding conservation of energy (specically, electricity, gas,
water, and waste generation) at the household level. In Table 2, we 2.3.2. Analysis of determinant factors for energy consumption
highlight the details of the questions regarding electricity and gas, We began by hypothesizing that the number of family mem-
as these were the main targets of our study. As indicated in the bers (Q3), the number of individuals over 60 years old (Q3), the
table, Q6 asked about (a) the approximate level of energy con- age of the respondent (Q3), the household income (Q4), the
sumption per month, and (b) to what extent each household number of air conditioners in the house (Q5), and the perceptions
 K. Hara et al. / Energy Policy 87 (2015) 240249 243

of energy savings (Q6-b) could inuence the household con- Table 3

sumption of energy. We thus employed these factors as ex- Determinant factors of electricity consumption (2009, 2013).
planatory variables for the regression analysis using the ordered
Variables 2009 2013
logit model.
We hypothesized that the presence of elderly people in the Number of family members *** ***
household could increase the consumption of energy due to the Number of individuals over 60 *** ***
necessity to maintain a comfortable temperature; thus, we used Age of respondent ***
Household income *** ***
the number of individuals over 60 as the explanatory variable. The Number of air conditioners *** ***
number of air-conditioners was selected as a proxy for the amount Perception of energy savings *** ***
of energy consumed by home appliances. We also assigned the Number of samples 2866 860
variable for the analysis of gas, assuming that it can be a proxy for
energy consumption behavior in general. A variable was assigned Note: ***
and *
represent the 0.01 and 0.05 level of signicance, respectively.
to the perception of energy savings, and it represents the answer
obtained from Q6-b (By how much could you reduce consump- Table 4
tion if you really tried?; see Table 2). This variable is based on the Determinant factors of gas consumption (2009, 2013).
assumption that there may be various possible relationships be-
Variables 2009 2013
tween consumption behavior and the perception of energy saving.
Number of family members *** ***
2.3.3. Analysis of determinant factors for perception of energy Number of individuals over 60  
savings Age of respondent *** ***
Household income ***
As the explanatory variables of the perception of energy saving,
Number of air conditioners *** ***
we selected the number of family members (Q3), the number of Perception of energy savings *** ***
individuals over 60 years old (Q3), the age of the respondent (Q3), Number of samples 2599 733
the household income (Q4), the number of air conditioners in the
house (Q5), and the amount of energy consumed (Q6-a); we as-
sumed that any of these variables could inuence perceptions of analysis; the actual number of samples used in the analysis is
energy saving. shown at the bottom of table.
After identifying key determinant factors and any changes ob- Table 3 indicates that the determinant factors for electricity
served during the two-year period based on the ordered logit consumption included the number in the family, number of in-
model (macrolevel analysis), we then further delve into the data dividuals over 60, household income, number of air conditioners,
from the questionnaires for both years by classifying all the and perception of energy savings. The only apparent difference
households that responded, aiming to investigate possible reasons between 2009 and 2013 in terms of the determinant factors was
for such changes. that the age of respondent was not a determinant factor in 2013.
In fact, no studies have yet examined the interconnection be- Table 4 shows that for both years, the factors inuencing gas
tween energy consumption behavior and perception of savings. consumption included the number in the family, age of re-
Further, possible impacts of the elderly in the family have not been spondent, number of air conditioners, and perception of energy
explored, either, while attributional factors such as family size and savings. The results imply that, unlike the results for electricity, the
household incomes were studies (Mizobuchi and Takeuchi, 2013). number of individuals over 60 did not inuence gas consumption.
Our analysis, thus, will be able to provide a new insight from these Furthermore, although household income was an essential factor
viewpoints.
in 2009, it does not appear to be signicant in 2013.

3. Results 3.2. Determinant factors of perception of energy savings

In this section, we report the estimation results obtained by
regression analysis using the ordered logit model. Because we
wished to identify the attributes of energy consumption behavior
(and we did not intend to estimate the marginal values or mag-
nitude of change of the factors), only the sign and signicance of
the estimated coefcients are reported. For example, if an ex-
planatory variable has a positive and statistically signicant sign, it
means that higher values of the variable will increase the prob-
ability that the explained variable has a higher order in choice
number.
Table 5 shows the results of a regression analysis for the pos-
sible determinant factors for the perception of electricity savings
in 2009 and 2013. Table 6 illustrates the regression results for the
determinant factors for the perception of gas savings in 2009 and
2013. Results from these tables indicate that the determinant
factors for perceptions of energy savings in 2009 are household
income and the actual amount consumed. Notably, this trend ap-
plies for both electricity and gas.
On the other hand, we found different determinant factors for
2009 and 2013. The actual amount consumed was the only factor

3.1. Determinant factors of electricity and gas consumption Table 5

Determinant factors for perception of electricity savings (2009, 2013).
Table 3 shows the results for 2009 and 2013, respectively, of the
Variables 2009 2013
regression analysis using the ordered logit model; these results
indicate the determinant factors of electricity consumption. Ta- Number of family members
ble 4 illustrates the determinant factors of gas consumption in Number of individuals over 60  
2009 and 2013, respectively. Note that while the numbers of re- Age of respondent * 
Household income ***
sponses to the questionnaire were 4448 in 2009 and 1245 in 2013, Number of air conditioner  
due to some incomplete responses, there were some missing data Actual amount consumed *** ***
for the selected explanatory variables necessary for the regression Number of samples 2866 860
analysis. These incomplete responses were omitted from the
244 K. Hara et al. / Energy Policy 87 (2015) 240249
Table 6
Determinant factors for perception of gas savings (2009, 2013).
Variables 2009
Number of family members
Number of individuals over 60
Age of respondent
Household income ***
Number of air conditioners  
Actual amount consumed ***
Number of samples 2559
for both electricity and gas in 2013, indicating that household in-
come was not a determinant factor in 2013. We argue that there
was a change in local perceptions and attitudes between 2009 and
2013.
From the macrolevel analysis using the ordered logit model, we
identied the key determinant factors for energy consumption and
the perceptions of energy savings. We also found that the actual
amount consumed, household income, and perceptions of savings
are closely related, and that in 2013, household income was not a
determinant factor of the perceptions of savings, unlike the case in
2009. The results of regression analysis using the ordered logit
model imply that while the perception of energy saving tended to
increase with household income in 2009, this was not the case in
2013.
4. Discussion
4.1. Possible determinant factors identied by the ordered logit
model
Below, we summarize the results of our analysis of the ques-
tionnaire data. As for the factors that could potentially inuence
the consumption of electricity and gas by the households in Suita,
we found the following points:
1. The primary determinant factors that are common for both
electricity and gas include the number in the family, the
household income, the number of air conditioners, and the
perception of savings.
2. The number of individuals over 60 was only relevant for elec-
tricity consumption.
3. No large difference was observed between the responses in
2009 and those in 2013. However, gas consumption was not
determined by household income in 2013, although it was
statistically signicant as a determinant factor in 2009.
For the rst point above, we note that it makes sense that the
number in the family, the household income, and the number of
air conditioners might determine the amount of energy con-
sumed. The implication is that the greater the number of family
members, the more likely it is that the electricity or gas con-
sumption will increase, and the greater the household income,
the greater the consumption of energy. This is also true for the
number of air conditioners.
It is interesting to note that those who believe they could fur-
ther cut consumption tend to consume more energy. This could
imply a possible gap between peoples behavior and their
awareness of their ability to conserve energy.
For the second point above, it would make sense that a
household with elderly people would tend to rely more on
electricity, since electricity is essential for providing various
services, including protection against the cold or heat. Note that
the questionnaire was carried out during October and
November in 2009, and during February and March in 2013,
when the average temperature in Osaka Prefecture is below
15 C; this probably impacts electricity consumption for pro-
2013
tection against the cold. This nding could have important im-
plications for policy, since it would not be effective to require all
 household to achieve energy conservation by rigid and stan-
dardized policies. The result here implies that elasticity allows
*
each household to respond in accord with its characteristics and
*** attributes. In particular, the potential for saving electricity is
733 lower in households with elderly people.
In terms of perception of electricity and gas savings, we found
the following:
4. In 2009, household income and the actual amount consumed
were found to be the factors that inuence the perception of
energy savings.
5. Actual amount consumed was the only factor affecting the
perception of energy savings in 2013, and the household income
had no effect.
The fourth point above explains the observed tendency that
households with higher incomes were more willing to cut energy
consumption in 2009. However, this tendency vanished in 2013.
From these ndings, we assume that the perception of energy
savings changed in 2013, and is not necessarily based on house-
hold income.
A clear difference is observed between 2009 and 2013 in terms
of the determinant factors, especially for gas consumption and
perception of savings; that is, household income was only relevant
in 2009 as a determinant factor for gas consumption and per-
ception of energy savings.
4.2. Detailed analysis of relationships between three essential
variables
Given that between 2009 and 2013, there was a change in the
relationships among the actual amount consumed, the household
income, and the perception of savings, as indicated by the ordered
logit model, we further analyzed the data from the questionnaires.
For both years, we classied all the households that responded,
based on the actual amount consumed, the household income, and
the perception of savings. In particular, we examined the re-
spondents consumption of both electricity and gas, and their
perceptions of energy savings, classied by household income.
The composition (percentages) of respondents classied by
household income in 2009 and 2013 was 12.1% and 14.5% for level
1 (less than two million yen), 44.6% and 48.3% for level 2 (between
two million yen and ve million yen), 32.5% and 29.2% for level 3
(between ve million yen and 10 million yen), and 10.9% and 8.0%
for level 4 (more than 10 million yen), respectively. Thus the basic
compositions of respondents are equivalent for both 2009 and
2013; the percentage of households of level 2 is the largest in both
years, and this is followed by level 3.
Under each level of annual household income, we then classi-
ed all the responses for both years by electricity and gas con-
sumption. In Fig. 1a, we show the electricity consumption of
households for which the annual household income was less than
two million yen (level 1). The gure shows the percentages of
households for which the monthly electricity consumption falls in
various ranges. The x-axis represents the categories of monthly
electricity consumption, as follows: category 1: less than
200 kWh; category 2: 201300 kWh; category 3: 301400 kWh;
category 4: 401500 kWh; and category 5: more than 501 kWh.
From the gure, we can see a clear difference between 2009 and
2013. For instance, the percent of households that consumed a
large amount of electricity (category 5: more than 501 kWh) was
much higher in 2013 than it was 2009. Similarly, Fig. 1b, c and d
 K. Hara et al. / Energy Policy 87 (2015) 240249 245

Fig. 1. Percent of households in which the monthly electricity consumption is in the range indicated (x-axis), for households with incomes under two million yen (a),
between two million yen and ve million yen (b), between ve million yen and 10 million yen (c) and of more than 10 million yen (d).

shows the percentage of households in each range of electricity We note that Fig. 1ad reveal a stark contrast between the data
consumption, when the household income was between two from 2009 and those from 2013. While it is true that households
million yen and ve million yen, between ve million yen and 10 with higher incomes tend to consume more electricity, in general,
million yen, and more than 10 million yen, respectively. the ratio of households under the category of highest consumption

Fig. 2. Percent of households in which the per month gas consumption is in the range indicated (x-axis), for households with incomes of less than two million yen (a),
between two million yen and ve million yen (b), between ve million yen and 10 million yen (c) and more than 10 million yen (d).
246 K. Hara et al. / Energy Policy 87 (2015) 240249

(i.e., more than 501 kWh per month) was higher in 2013 than in
2009, regardless of the household income. As for the households
in the highest income range (Fig. 1d), the ratio of those whose
consumption level was more than 501 kWh per month was almost
40% in 2013, while in 2009, it was about 25%. Based on these
gures, we argue that the number of households with high levels
of consumption increased between 2009 and 2013, and we note
that this was regardless of income.
In Fig. 2ad, we show the percentage of households with gas
consumption in various ranges, and with income levels of less than
two million yen, between two million yen and ve million yen,
between ve million yen and 10 million yen, and more than
10 million yen, respectively. Note that the x-axis represents the
categories of gas consumption per month, as follows: category 1:
15 m3; category 2: 610 m3; category 3: 1115 m3; category 4:
1620 m3; and category 5: more than 21 m3. In general, house-
holds with consumption levels in the higher ranges, such as ca-
tegories 4 and 5 (1620 m3 and more than 21 m3, respectively),
tended to increase with income level in 2009. However, the ma-
jority of households were at this highest level in 2013, regardless
of income level. This clearly explains that income level was not a
determinant factor for gas consumption in 2013, unlike the case
for 2009, as implied by the result of the ordered logit model.
Again, the detailed analysis here shows that consumption pattern
changed between 2009 and 2013.
Next, we examine the perception of energy savings, both for
electricity and gas, and the relationships of these to income levels,
both in 2009 and 2013. Specically, we estimated the percent of
households with different levels of energy savings (13%, 46%, 7
10%, more than 11%, and unable to cut), under each level of
household income.
In Fig. 3ae, we show the percent of households that claimed to
be able to cut electricity consumption by 13%, 46%, 710%, more
than 11%, unable to cut (0%), respectively. The numbers on the x-
axis show the classes of income level. We found that the percen-
tage of households that claimed to be able to cut electricity con-
sumption by a large amount (more than 11%) was larger in 2013
than in 2009 (Fig. 3d), while the percentage who claimed to be
able to cut by only by 13% was larger in 2009 than in 2013
(Fig. 3a). On the other hand, the percentage of those who reported
that further energy savings were impossible was larger in 2013
than in 2009, especially for low-income households (Fig. 3e).
Overall, more households in 2013 than in 2009 reported that a
large amount of electricity savings were possible, regardless of
income level. In Fig. 4a, we show the change in the percentage of
low-income households (less than 2 million yen) for each level of
electricity savings (i.e., 13%, 46%, 710%, more than 11%, and 0%),
between 2009 and 2013; Fig. 4b shows the same information for
high-income households (more than 10 million yen). From Fig. 4a
and b, we see that the ratio of households willing to use only a

Fig. 3. Percent of households that claimed to be able to cut electricity consumption by 13% (a), 46% (b), 710% (c) and more than 11% (d), and percentage of households
that claimed to be unable to cut electricity consumption (e), by income (x-axis). Note: The y-axis value indicates the percent of total responding households in 2009 and
2013.
 K. Hara et al. / Energy Policy 87 (2015) 240249 247

Fig. 4. Change in the percent of low-income households (less than two million yen) (a) and high-income households (more than 10 million yen) (b) by the level of perception
of electricity savings (i.e., 13%, 46%, 710%, more than 11%, and 0%) and by year (2009 and 2013).

little less electricity (13%) is less in 2013, and that of households
willing to cut more electricity is larger in 2013, regardless of the
income level. Based on this information, we argue that the
awareness of the need to conserve electricity and perception of it
being possible was greater in 2013 than in 2009, regardless of
income level. This also supports the results of the ordered logit
model that, in 2013, income level disappeared as a determinant
factor for the perception of electricity savings.
The perception of gas savings shows a similar pattern to that of
electricity savings, in that in 2013, more households than in 2009
stated a willingness or ability to reduce gas consumption. In
Fig. 5ae, we show the percent of households that claimed to be
able to cut gas consumption by 13%, 46%, 710%, more than 11%,
and unable to cut (0%), respectively. Again, the numbers on the x-
axis indicates the class of income level. Overall, the ratio of
households willing to cut gas consumption by a large extent (more
than 11%) was larger in 2013 than in 2009 (Fig. 5d), while the ratio
of those willing to cut by 13% was larger in 2009 than in 2013
(Fig. 5a), regardless of income level. On the other hand, more
households in 2013 than in 2009 claimed not to be able reduce
consumption, especially for low-income households (Fig. 5e). In
Fig. 6a, we show the change in the ratio of low-income households

Fig. 5. Percent of households that claimed to be able to cut gas consumption by 13% (a), 46% (b), 710% (c) and more than 11% (d), and percent of households that claimed
to be unable to cut gas consumption (e), by income level (x-axis).
248 K. Hara et al. / Energy Policy 87 (2015) 240249
Fig. 6. Change in the percent of low-income households (less than two million yen) (a) and high-income households (more than 10 million yen) (b) by the level of perception
of gas savings (i.e., 13%, 46%, 710%, more than 11%, and 0%) and by year (2009 and 2013).
(less than 2 million yen) for each level of gas savings (13%, 46%,
710%, more than 11%, and 0%), between 2009 and 2013. The g-
ure demonstrates the overall upward trends in terms of the ratio
of households willing to reduce consumption, regardless of the
amount saved, although the number of households that claimed
that no greater savings were possible also increased from 2009 to
2013. In Fig. 6b, we show the same information for high-income
households (more than 10 million yen), and there is an upward
trend in the households claiming a large amount of savings (more
than 11% reduction) and a slight downward trend for the smaller
amount (13% reduction).
From all the above analyses, we conclude that the number of
high-consumption households increased from 2009 to 2013, but the
perception of the possibility of savings also increased. This has two
important implications. First, energy consumption and perception of
energy savings both appeared to have altered between 2009 and
2013. Second, there was a growing gap between behavior and the
perception of savings, in that some households became more was-
teful of energy at the same time that they acknowledged that they
were able and willing to reduce consumption.
5. Conclusion and policy implications
In this study, we identied the determinant factors associated
with electricity and gas consumption as well as those associated
with the perception of savings, for residents of Suita, Japan. Data
was obtained by conducting a large-scale survey in 2009 and 2013.
We used the ordered logit model to identify the essential factors
that could inuence energy consumption and perception of sav-
ings, and we further analyzed the relationships and mechanisms
among the essential components (the amount of energy con-
sumed, the perception of savings, and household income) by
looking into details of the data obtained from the questionnaire.
For the two years surveyed, we identied some changes that were
based on the determinant factors, and we also identied a growing
gap between consumption behavior and perceptions of savings.
These important results, which were obtained from both the
ordered logit model and the detailed analysis, provide some im-
portant implications for energy policy. Specically, the determi-
nant factors we identied in the study can be used as a basis for
designing a policy that effectively induces energy-saving behavior.
For example, we found that policy makers should take into ac-
count the elderly (the number of people over 60) in a given
household when identifying effective policy measures, instead of
simply implementing generalized measures.
According to a white paper for the City of Suita (2014), total
energy consumption decreased between 2009 and 2013. However,
detailed analysis of this study indicated that the number of high-
consumption households actually increased in 2013. This fact,
coupled with the above information from Suita, implies that
households with high energy consumption and those with mod-
erate consumption, are becoming polarized. However, the essen-
tial point is that the high-consumption households believe that
they are capable of reducing energy consumption. Thus, there is a
potential for further energy savings, particularly by those in the
high-consumption category. Indeed, prior to our study, the energy
and environmental policy makers in Suita were not fully aware of
these facts and implications, according to our interview with the
municipality ofcials. Our ndings about the determinant factors
as well as the polarization of consumption patterns in Suita pro-
vide essential insights necessary for designing effective policy
measures to promote energy conservation program.
In Suita, a new action plan to combat climate change was for-
mulated in March 2011; both lifestyle changes and relevant energy
savings were placed at high priority, and the stated goal was to
reduce greenhouse gas emissions in 2020 by more than 25%
compared to 1990 levels (City of Suita, 2011). Under the current
plan, there are no specic energy-saving measures that target in-
dividual households. The action plan is now under discussion for
revision. The revised plan should take into account these results in
order to formulate more-effective measures.
There might be several reasons for the growing gap between
the perceptions about energy savings and the actual consumption
in 2013, and the polarization of households with different levels of
energy consumption. As indicated by some studies (Kosugi, 2014;
Nishio and Ofuji, 2014), after the 2011 earthquake, the required
electricity savings, which had a specic target, could have affected
perceptions and behaviors. There is a possibility that such a strict
policy could have backred. We plan to conduct a detailed analysis
of this in a future study. The ndings, such as the factor of the
elderly in the house and relationships between income level,
consumption behavior and perception of savings, shall provide
new insights into energy policy which the past studies have not
claried. As a future study, we will explore how we could model
consumption behavior and perception of savings based on our
ndings while taking into account other variables and informa-
tion, such as age of electric equipment, which we obtained in the
questionnaire survey. We also intend to explore more in detail
why such changes occurred and the possible impacts of the po-
licies implemented following the Great East Japan Earthquake; in
Q9 of the questionnaire in this study (Table 1), as we also obtained
information relevant to this topic. We believe it is also important
to gather and analyze data, such as those obtained in our survey, in
wider areas of Japan, in order to discuss whether the facts ob-
tained from our study, such as growing gaps between consump-
tion pattern and perception of energy savings, can be seen in other
parts of Japan and to gain more knowledge of the possible impacts
 K. Hara et al. / Energy Policy 87 (2015) 240249
of the earthquake on energy consumption and conservation. The
imposed energy savings following the earthquake varied in dif-
ferent areas, and such differences may have affected attitudes and
perceptions in different way. Comprehensive analyses of such
impacts will be carried out in a future study.
Acknowledgments
We are grateful to the City of Suita, Osaka Prefecture, Japan, for
providing the data and other information relevant to this study.
References
Albadi, M.H., El-Saadany, M.H., 2008. A summary of demand response in electricity
markets. Electr. Power Syst. Res. 78, 19891996.
Attari, S.Z., DeKay, M.L., Davidson, C.I., Bruin, W.B.D., 2010. Public perceptions of
energy consumption and savings. Proc. Natl. Acad. Sci. 107, 1605416059.
Barr, S., Gilg, A., Ford, N., 2005. The household energy gap: examining the divide
between habitual-and purchase-related conservation behaviours. Energy Policy
33, 14251444.
Becker, L.J., Seligman, C., Fazio, R.H., Darley, J.M., 1981. Relating attitudes to re-
sidential energy use. Environ. Behav. 13, 590609.
City of Suita, 2011. New Action Plan to Combat Climate Change in Suita (in Japa-
nese). Available from: http://www.city.suita.osaka.jp/var/rev0/0034/5757/
2011425154319.pdf (viewed on 22.12.15.).
City of Suita, 2014. White Paper on the Environment in Suita (in Japanese). Avail-
able from: http://www.city.suita.osaka.jp/var/rev0/0063/3864/SuitaE-re
portH26_00.pdf (viewed on 05.03.15.).
Gatersleben, B., Steg, L., Vlek, C., 2002. Measurement and determinants of en-
vironmentally signicant consumer behavior. Environ. Behav. 34, 335362.
Greene, H.W., 2008. Econometric Analysis, sixth ed. Prentice Hall, Upper Saddle
River, New Jersey.
Gromet, D.M., Kunreuther, H., Larrick, R.P., 2013. Political ideology affects energy-
249
efciency attitudes and choices. Proc. Natl. Acad. Sci. 110, 93149319.
Herter, K., 2007. Residential implementation of critical-peak pricing of electricity.
Energy Policy 35, 21212130.
Kosugi, M., 2014. Public Opinion on Environmental and Energy Issues: Results of
the Census After 3 Years of the Great East Japan Earthquake. CRIEPI Research
Report, Y14004.
Martinsson, J., Lundqvist, L.J., Sundstro m, A., 2011. Energy saving in Swedish
households. The (relative) importance of environmental attitudes. Energy
Policy 39, 51825191.
Ministry of Economy, Trade, and Industry, 2013. The White Paper on Energy Usage,
2013 (in Japanese). Available from: http://www.enecho.meti.go.jp/about/
whitepaper/2013html/2-1-2.html.
Mizobuchi, K., Takeuchi, K., 2013. The inuences of nancial andnon-nancial
factors on energy-saving behaviour: a eld experiment in Japan. Energy Policy
63, 775787.
Moll, H.C., Noorman, K.J., Kok, R., Engstrom, R., Throne-Holst, H., Clark, C., 2005.
Pursuing more sustainable consumption by analyzing household metabolism in
European countries and cities. J. Ind. Ecol. 9 (1), 259276.
Nishio, K., Ofuji, K., 2014. Ex-post Analysis of Electricity Saving Measures in the
Residential Sector in the Summer of 2013. CRIEPI Research Report, Y13010.
Ouyang, J., Hokao, K., 2009. Energy-saving potential by improving occupants' be-
havior in urban residential sector in Hangzhou City, China. Energy Build. 41,
711720.
Saidur, R., Masjuki, H.H., Jamaluddin, M.Y., 2007. An application of energy and
exergy analysis in residential sector of Malaysia. Energy Policy 35, 10501063.
Sweeney, J.C., Kresling, J., Webb, D., Soutar, G.N., Mazzarol, T., 2013. Energy saving
behaviours: development of a practice-based model. Energy Policy 61,
371381.
Ueno, T., Sano, F., Saeki, O., Tsuji, K., 2006. Effectiveness of an energy-consumption
information system on energy savings in residential houses based on mon-
itored data. Appl. Energy 83, 166183.
Viklund, M., 2004. Energy policy options: from the perspective of public attitudes
and risk perceptions. Energy Policy 32, 11591171.
Vine, D., Buys, L., Morris, P., 2013. The effectiveness of energy feedback for con-
servation and peak demand: a literature review. Open J. Energy Efc. 2, 715.
Wang, Z., Zhang, B., Ying, J., Zhang, Y., 2011. Determinants and policy implications
for household electricity-saving behaviour: evidence from Beijing, China. En-
ergy Policy 39, 35503557.