Content uploaded by Attila Bai
Author content
All content in this area was uploaded by Attila Bai on Jan 28, 2024
Content may be subject to copyright.
Energy Strategy Reviews 50 (2023) 101192
Available online 25 September 2023
2211-467X/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
The monetary value of convenience and environmental features in
residential heat energy consumption, in particular its social determinants
Attila Bai
a
, Ibolya Czibere
b
, Imre Kov´
ach
b
,
c
, Boldizs´
ar Megyesi
c
,
*
, P´
eter Balogh
a
a
Faculty of Economics and Business University of Debrecen, Debrecen, Hungary
b
Department of Sociology and Social Policy, University of Debrecen, Hungary
c
Institute for Sociology, Centre for Social Sciences, Hungarian Academy of Sciences, Hungary
ARTICLE INFO
Handling Editor: Xi Lu
Keywords:
Energy price
Convenience
Hybrid choice
Consumer group
Social characteristics
ABSTRACT
Environmental and comfort value of renewable energy sources is a less studied area. The article introduce the
economic value of these characteristics, related to social determinants. The data basis is representative of the
Hungarian population above 18 years of age, by gender, age and level of education. Our model estimation was
developed by hybrid choice context in latent class modelling. In addition to the usual WTP (willingness to pay)
calculation, we also used WTI (willingness to invest) calculations. The results show that the value of
environmentally-friendly nature in Hungary exceeds the convenience factor. The latter cannot be signicantly
detected for the majority of the Hungarian population. This statement is supported by results for both fuels and
boilers. Our ndings show the strong attachment of the Hungarian average person to the use of rewood and
natural gas, which are very typical in Hungarian heat consumption. Regarding socially selected consumer
groups, strong correlation can be observed between social capital supply, income level, access to information and
environmental awareness in Hungary. Regarding policy implications, our results in development of clusters may
be useful for the establishment of a selective support policy and two clearly identiable groups should be
highlighted in the energy policy.
1. Introduction
The aim of this study is to analyse the social determining factors of
the energy decisions on residential heating, one of the most important
areas of energy consumption. Residential energy decisions are inu-
enced by differences in social and economic position, which are
responsible for unequal access to energy resources [1]. The value of
comfortable and environmentally friendly energy consumption is
becoming more and more important.
A previous research [2] assessed the attitudes and knowledge of the
Hungarian population on energy consumption using a national, repre-
sentative quantitative survey. An international comparative analysis of
the specic characteristics of Hungarian households’ energy use [3] and
the preferences of the Hungarian population regarding heating systems
and their socio-demographic background were examined on the basis of
a) similar international researches, b) two types of data collection, c)
expert opinions and d) the experience gained during previous pilot tests
including similar questions. This study assumed that cost factors, the
operation of the heating system, and the environmental impact are
important considerations in the selection. A further assumption was that
there are identiable segments of the population where comfort and
environmentally friendliness may play a key role in making such a de-
cision, as opposed to those who, are unable to make such value-based
decisions. In 2019, 69.8% of persons in the EU lived in an
owner-occupied dwelling, but the Hungarian value is signicantly
higher (91.7%) [4]. In addition, mainly young people lived in rented
apartments in Hungary, but most of them want to own a home in the
near future.
The main research questions are the identication of the social
components that most inuence the decisions of the households
included in the survey (e.g. income, age, education, occupation), and the
examination of the willingness to pay and invest for the purpose of
comparison. The research aim of this study is to seek groups of house-
holds with markedly different preferences for heating, but with a good
t, mapping the underlying factors of decision making using a hybrid
latent class model. Our research use a new econometric model, evalu-
ated a statistically reliable national survey, which considers the main
elements of previous (Hungarian and international) surveys and in this
* Corresponding author.
E-mail addresses: bai.attila@econ.unideb.hu (A. Bai), Megyesi.Boldizsar@tk.hu (B. Megyesi), Balogh.peter@econ.unideb.hu (P. Balogh).
Contents lists available at ScienceDirect
Energy Strategy Reviews
journal homepage: www.elsevier.com/locate/esr
https://doi.org/10.1016/j.esr.2023.101192
Received 25 April 2022; Received in revised form 31 August 2023; Accepted 4 September 2023
Energy Strategy Reviews 50 (2023) 101192
2
way it shows the similarities and differences in energy decisions be-
tween the Hungarian and international energy consumers.
2. Literature review
2.1. Characteristics of household energy consumption in Hungary
In 2019, the global energy consumption was 13 975 Mtoe (585 EJ)
(7), an increase of 43% compared to the turn of the millennium. This
value represents an average annual increase of more than 2% since
2000. Households are responsible for 26.1% of nal energy consumption
in the EU, with heating accounting for the highest share at 63.6% (9).
Residential buildings presents considerable energy savings potential [5]
and strongly contribute to global CO2 emissions due to the high energy
demand for electricity and heating [6]. Then the COVID-19 caused a
uctuation, but the worldwide energy consumption already reached
595 EJ in 2021 [7]. Several earlier studies showed the role of renewable
energy consumption in economic growth [8] in the OECD countries.
In Hungary, the share of the household sector (34%) [9] and, more
specically, the utilization for heating purposes (75% [10]) in total nal
energy consumption are signicantly higher. Regarding energy sources
used for heating, it can be stated that the share of renewable energy and
waste is the most signicant in the EU and also in Hungary (27.0% and
32%, respectively), in addition to natural gas (38.0% and 56.3%,
respectively) [11]. Although the share of renewables in Hungary is
higher than the EU average due to biomass combustion, more than 56%
share of natural gas use (in the absence of any signicant Hungar-
ian/national gas deposits) results in serious dependence on gas imports.
In terms of costs, Hungarian households spent 20% of their total ex-
penditures per month (EU average: 24%) on housing and household
energy. More specically, the share of household energy accounted for
more than 60%, including gas, electricity and solid fuels, in this order
[10,12]. However, Franceschinis et al. [13] stated, that the diffusion of
RES (renewable energy sources) technologies still limited, while heating
systems based on fossil fuels are still predominant [8,14]. The results of
Ruokamo [5] reveal that Finish households view supplementary heating
systems (especially solar-based) favourably. Besides, Brodny-Tutak [15]
proved that higher ratio in RES results a positive impact on the economic
growth, so more intensive use of RES could be a perspective way for the
less developed countries [16].
In general, Hungarians link signicant environmental pollution and
high-cost requirements to the use of rewood in a stereotypical way [2].
On the contrary, heating with natural gas is associated with the ste-
reotype of convenient and inexpensive heating, probably not indepen-
dently of the subsidised residential pricing that has been in force for ten
years. The perceptions of comfort and environmental considerations are
nearly the same, although different social groups show signicant dif-
ferences. The public perception of rewood (i.e. it is uncomfortable and
polluting) was also supported by a U.S. study [17]. It is important to note
that regarding external costs of biomass is higher than natural gas and
mineral oil, similar to the lignite, just the coal has higher value [18].
The housing stock in Hungary is predominantly outdated, i.e. most
residential buildings are poor energy efcient, which leads to high
overheads, carbon dioxide and air pollution. The most common type of
housing - almost one in ve - is a detached house built in the 1960s or
1980s. Heating such a building without insulation requires up to twice
as much energy per square metre as a typical panelled dwelling, and four
times as much as a (insulated) 2000s apartment block. Gas is the most
common energy carrier in 2019, with solid fuel (e.g. wood, coal, bri-
quettes) in second place and district heating in urban areas in third
place. Almost 40% of households in the bottom quintile of income
groups heat exclusively with solid fuels, compared to only 9% in the top
quintile. In other words, wood heating is mainly the fuel of choice for the
poorer classes, as opposed to gas heating. Residential solid fuel com-
bustion is one of the main causes of air pollution in Hungary, due to
outdated stoves and poor insulation of buildings [19,20]. Household
solid fuel combustion is responsible for more than 80% of the seriously
harmful particulate matter emissions in Hungary (the EU average is
41%, average particulate matter emissions per capita in the EU are less
than a third of those in Hungary) [21].
2.2. Assessment of environmental awareness in international literature
Environmentally conscious behaviour is evaluated in the literature
with the willingness to pay and willingness to accept [22]. To eliminate
this problem, Kocsis-Marjain´
e [23] and Whittington [24] recommended
examining the use of overtime instead of monetary value, especially in
developing countries. Studies of various environmental problems in
various developing countries showed that residents are less willing to
spend money on environmental protection, but more willing to devote
time to this purpose [25–28].
Even in developed countries, people prefer to offer their time as
opposed to their money [29], but the value of their free time depends
signicantly on how they spend this time, their income, and their social
position [30]. The authors of this paper tried to eliminate this problem
by means of a nationally representative questionnaire survey and the
formation of clusters that can be considered as characteristic in terms of
thermal energy consumption, as well as by conducting a discrete choice
experiment. The analysis of latent variables, such as attitudes, can now
be included in the long-used discrete choice experiment. For this reason,
the effect of environmental attitudes was incorporated into the model to
answer the research questions [31]. Michelsen and Madlener [32]
examined the preferences of German homeowners regarding the intro-
duction of an innovative heating system, providing their estimates for
the representative sample using a multinomial logit model. Based on the
obtained results, there are different drivers for adopting an innovative
heating solution for newly built and existing single-family homes. For
the former, decisions are largely related to the characteristics of the
given system, while for the latter, choices are greatly inuenced by
various socio-demographic factors. Rouvinen-Matero [33] examined the
preferences of Finnish housing properties with regard to heating sys-
tems. The aim of their research was to reveal which properties most
inuence residents’ decisions concerning the choice of heating system. A
stated choice (SC) experiment involved six heating alternatives,
including wood pellet boiler, solid wood-red boiler, district heat,
electricity, ground heat (pump) and oil boiler in a labelled form. Attri-
butes included in their experiment included investment cost, annual
operating cost, CO2 emissions, ne particle emissions, and personal
presence/work. In addition, estimates were made using multinomial
logit and random parameter logit specications. The investment cost is
the most important consideration when making a decision, and other
(non-nancial) characteristics also have a signicant impact. The het-
erogeneity in preferences was caused on the one hand by the availability
of district heating, the existing heating system, and forests owned by
certain residents. In addition, the authors pointed out the signicant role
of non-observed preference heterogeneity. In their 2020 representative
survey conducted in three countries (Poland, Sweden, UK), Mills and
Schleich [34] aimed at assessing households’ preferences for a new
heating system and making estimates of willingness to pay. The attri-
butes involved in their experiment included the heating bill (25%, 50%,
75% reduction), the duration of installation (half a day, three days, one
week), the duration of warranty (2, 5, 10 years), the cost of investment
(3–20 thousand EUR), the extent of subsidies in relation to the invest-
ment cost (0, 5, 15, 25%) and the form of subsidy (public agency, energy
provider, no subsidy (if subsidy =0%)). Based on Mixed logit model
estimation, respondents generally have a positive judgement of dis-
counts for a new heating system, although it depends on the given
country, Poland being the most signicant in this respect. In the case of
Sweden, it was found that discounted prices were more effective if they
were offered by a public rather than a private source of funding. In
addition, it was shown that respondents do not prefer longer time of
installation and show a higher willingness to pay for a longer warranty
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
3
period.
The methodological guide of Mariel et al. [35] is a signicant
contribution to the literature. The hybrid choice modelling was used in
the context of the traditional latent class specication, where classes
were allocated based on latent attitudes. Young, low-income men who
live closer to wind farms have a stronger pro-wind power generation
attitude. Based on the performed latent class modelling, two groups with
opposite attitudes were distinguished.
In the experiment of Achnicht (2011), the 400 sampled German
house owners could choose either a modern heating system or an
improved thermal insulation for their home and the results showed that
environmental benets have a signicant impact on choices of heating
systems. Based on Altnicht et al. (2014) results, those homeowners are
more likely to undertake energy modernisation activities who are able to
pay and who can see a return on this investment [36].
Research by Yu et al. [37] examined the combined effects of energy
poverty, GDP, renewable energy consumption, natural gas consumption
and free trade on carbon emissions in 25 developing countries between
2001 and 2019. This was interpreted in the context of access to elec-
tricity. The countries studied currently have the highest demand for oil
and the lowest demand for renewables. The lack of access to electricity is
signicant. The results show a direct link between natural gas con-
sumption and CO
2
emissions. Energy poverty, i.e. low access to elec-
tricity, causes an increase in CO
2
. Increased economic activity in turn
leads to a decrease in CO
2
emissions. The authors therefore recommend
that governments concerned increase access to electricity through
infrastructure investments to ensure that the energy grid is developed
throughout all countries. This will help to displace the most polluting
fuels (oil, natural gas, coal) in the developing countries studied. It would
be necessary to extend electricity supply to rural areas and to provide
low-cost or free household electrical appliances.
Khan et al. [38] studied the relationship between per capita carbon
emissions from fossil energy consumption and the factors of renewable
energy consumption, technological innovation, carbon dioxide taxes,
GDP, industrialisation, foreign direct investment and government
integrity in 19 high-income countries of the European Union. While the
increase in technological innovation contributes to the reduction of
carbon emissions, the regression analysis results show that the rela-
tionship is not signicant at the highest quantiles. Carbon taxes reduce
carbon emissions at initial pollution levels, but at higher levels, carbon
taxes become a legal mechanism for continued pollution. Their empir-
ical results conrm that renewable energy consumption contributes to
carbon emission reductions.
Hu and Wang [39] investigated the relationship between environ-
mental regulation and carbon production capacity in China. The results
of empirical research based on the spatial spillover effect show that there
is a positive spatial correlation between the carbon productivity of
different regions and there is a threshold for the impact of environ-
mental regulation on carbon productivity. As the intensity of environ-
mental regulation changes from weak to strong, the impact on carbon
productivity of local regions changes from negative to positive. Empir-
ical research based on the spatial spillover effect shows that there is a
positive spatial correlation between the carbon productivity of different
regions in China and a threshold for the impact of environmental
regulation on carbon productivity. Improved carbon productivity in
local regions also improves carbon productivity in neighbouring regions.
Based on the regression results of the control variables, it is found that an
increase in GDP per capita contributes to an improvement in carbon
productivity and that technological innovation is also a positive factor in
improving carbon productivity. However, the industrial structure and
energy structure are barriers to improving carbon productivity.
Wang et al. [40] also investigated spatial effects in China: the impact
of green technological innovation on green total factor productivity
(GTFP). Their results show that green technological innovation has a
signicant positive effect on the change in green total factor produc-
tivity. However, they obtained different results by region. GTFP is
highest in the Eastern region and lowest in the Western region. The
analysis results of the spatial Durbin model suggest that green techno-
logical innovation has a signicant positive effect on its own change in
GTFP, but a negative effect on that of its neighbours. Green technolog-
ical innovations in the Eastern and Central regions have a negative effect
on their own GTFP, while those in the Western regions have a positive
effect.
Yin et al. [41] analysed the dynamic changes and inuencing factors
of forest carbon sequestration efciency at the provincial level in China
for 30 years. Their results showed that per capita gross domestic product
(GDP), urbanization and length of highway network had signicant
positive effects on carbon sequestration efciency, while total imports
and exports had signicant negative effects. Their results also show that
urbanization, ecological forest cover, temperature, GDP per capita,
population, and total imports and exports have spillover effects. Large
cities have a spillover effect, which also promotes the development of
surrounding cities. Urban communities are also affected by ecological
afforestation, i.e. forest policy development, in neighbouring cities.
Xu et al. [37] studied the impact of highway infrastructure im-
provements on carbon emissions. In their study, they used the length of
highways and CO
2
emissions and agglomeration as proxy variables
based on panel data of 278 cities in China from 2003 to 2016. There is an
inverted U-shaped relationship between highway infrastructure and CO
2
emissions, and the effect is driven by agglomeration as a route. Empir-
ical results show that the development of highway infrastructure has an
impact on CO
2
emissions. The development of motorway infrastructure
has a threshold on its impact on CO
2
emissions, i.e. motorway infra-
structure has an emission-reducing effect through its own positive ex-
ternality if it exceeds the threshold.
Ponce et al. [42] analysed the long-run relationship between eco-
nomic growth and nancial development, non-renewable energy,
renewable energy and human capital in 16 Latin American countries.
The analysis was based on statistical data from the World Bank and Penn
Word Table databases for the period 1988–2018. Their results show that
there is a long-run equilibrium relationship between nancial develop-
ment, non-renewable energy consumption, renewable energy con-
sumption, human capital and long-term economic growth. They show a
positive relationship between the variables of nancial development,
non-renewable energy consumption, renewable energy consumption
and human capital and economic growth, indicating that an increase in
these variables leads to an improvement in long-term economic growth.
A two-way causal relationship was found between nancial develop-
ment, human capital variables and economic growth. One-way causality
was observed between non-renewable energy consumption and eco-
nomic growth, and between economic growth and renewable energy
consumption.
Many studies are available regarding correlations between utiliza-
tion of RES and other macroeconomical indicators at country, or
regional level. Peng et al. [43] found, that the use of green energy cor-
relates with changes in GDP and inversely proportional to population
density. According to an up-to-date study by Wang et al. [44], carbon
emissions and income levels at the global level can be characterised by
an inverted U-shaped curve. Before its peak (USD 19 203 per capita),
renewable energy sources tend to play a dominant role in emissions
reductions, and afterwards the role of human capital considered more
important. A previous article of Wang et al. [14] based on data of 104
countries the ndings show that the role of RES in economic develop-
ment shows strong correlation in developed countries, while a U-shaped
relationship emerges in developing countries. The energy needs of rising
living standards in developing countries are usually based on
non-renewable energy sources, so the effectiveness of energy saving
measures there may outweigh the use of renewables [8,45,46]. Wang
et al. [14] analysed the importance of composite, political, economic
and nancial risk in the OECD countries on economic growth. They
found that RES has positive impact after a threshold (in case of the rst
two risk types) and between the two threshold (in case of the last two
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
4
risk types). The importance of strong connection between political sta-
bility and energy and food security was also emphasized by Popp et al.
[47]. Regarding OECD countries, Wang et al. [8] proved that research
and development considered the most important driving force in
spreading of RES, however, contribution of policy and environmental
pressure also have high importance in the highest-income countries.
Wang and Wang [8] stated (based on data of OECD countries) that RES
shows a clearly positive impact on economic growth, especially in case
of high-level, or growing energy consumption. In our opinion it proves
the competition between RES and energy saving methods. Important
note, that due to food-energy-feedstock debate, the energy use of
biomass is also highly dependent on factors inuencing food consump-
tion (population growth, urbanization, religious habits, changing di-
etary patterns) [48].
We present and overview the main points of the literature on envi-
ronmental awareness in Table 1, which summarizes the main details of
the reviewed studies by comparing the variables used, the method and
main results achieved.
2.3. Social determinants of energy use
The literature on social determinants of energy use examines factors
inuencing energy efciency in households [49,50]. The unequal social
distribution of access to information and knowledge is a key research
issue in this area [51–53]. In addition to reference groups, the network
effect [54], various forms of knowledge [34] also play a key role in
spreading sustainable attitude patterns [55]. The role of consumer
habits [56] and energy poverty [57] are the main topics of another
research focusing more on social inequalities. Another generally
accepted pattern is that well-educated residents tend to be more inter-
ested in energy efcient technologies, retrots and behaviours [58].
Cattaneo [59] examines the social determinants and limitations of
energy use in the context of a possible form of favourable policy inter-
vention. In relation to energy efciency, two domains of consumer
behaviour are distinguished. One of them is behaviour according to
everyday routines, while the other is investor behaviour resulting in
long-term energy efciency. Schleich et al. identify external and internal
boundaries as a guiding principle for consumer behaviour [60]. Con-
sumers invest less in energy-efcient technologies [61], and Schleich
et al. [62] include the factors in the theoretical framework of internal
and external constraints that can explain low adoption behaviour.
External constraints include primarily institutional factors that limit
the introduction of energy-efcient technologies, which are also
commonly referred to as market failure explanations [63]. In the system
described by Schleich et al. [62], external constraints include capital
market failures, lack of information, asymmetric and ambiguous infor-
mation, as well as nancial and technological risks. According to
Schleich et al. internal constraints include reference-dependence and
Table 1
On the main results of the selected literature.
Author, year Region Variables Methods Results
Marjain´
e et al., 2011 Hungary Survey Choice experiment “The local population has a zero willingness to pay for reduction of ood frequency,
so this outcome is of no value to the local population. In relation to water quality
changes, WTP is positive.”
Kocsis T, Marjain´
e
Szer´
enyi Z. 2018.
Hungary Survey WTP “Both WTP for saving the environment and WTA (willingness to accept) costing
externalization to others increases when the parameters of the decision are described
in temporal terms instead of in money.” (p. 1.)
Whittington D. 2010 Less developed
countries
Meta-analysis stated preferences The WTP is low in the analysed cases, apart from the services and goods provided.
Rai RK, Scarborough
H. 2015
Less developed
countries
Survey Choice experiment The WTP increases if participants can contribute also by labour.
Tilahun et al., 2015 Ethiopia Survey Contingent Valuation
Analysis
Properly designed payments with complementary policy interventions support
sustainable resource use and poverty reduction.
Lankia et al., 2014 Finland Survey WTP & willingness to
contribute (WTC)
Both WTP & WTC are different among social groups.
Eom –Larsson, 2006 South-Korea Survey WTP The authors compared the value of house-work time and market wage.
Kastner I, Matthies E.
2016
Germany Survey Choice experiment The strategies to foster investment into renewable energies has to be adopted to the
different social groups.
Michelsen CC,
Madlener R. 2012
Germany Survey Choice experiment The decision to invest in the modernisation of the heating system depends on the
preference of being more independent from fossil fuels and on the age of the home
(existing vs. new-built).
Rouvinen S, Matero J.
2013
Finland Survey Choice experiment Financial and non-nancial factors inuence residents choices, also the availability of
district heating and ownership od forest have an effect.
Mills B, Schleich J.
2012
EU & Norway Survey Statistical analysis Family age composition, and educational level has an impact on energy-investment.
East-West differences can be found also.
Yu et al., 2022 Developing
countries
Statistical data Quantile-on-Quantile
(QQ)’,
Low access to electricity, causes an increase in CO2 emissions, while increased access
to electricity help to displace the most polluting fuels (oil, natural gas, coal) in the
developing countries.
Hu and Wang (2020) China Statistical data Statistical analysis There is a spatial spillover effect and a positive spatial correlation between the carbon
productivity of different regions in China, and there is a threshold for the impact of
environmental regulation.
Ponce et al. (2021) Latin-America Statistical data Statistical analysis The results show that there is a long-run equilibrium relationship between nancial
development, non-renewable energy consumption, renewable energy consumption,
human capital and long-term economic growth.
Wang et al. (2023) 208 countries Long-term
statistical data
Differential GMM
estimation
Inverted U-curve between income level and CO
2
emission, before peak: impact of
RES, after peak: human capital is determining.
Wang et al. (2022a) 104 countries Long-term
statistical data
Regression estimation Relation between RES and economic development: strong in developed countries, U-
shaped curve in developing countries
Wang et al. (2022b) OECD Long-term
statistical data
Panel threshold model Impact of RES on economic growth is differential in case of differential country risk
types.
Wang et al. (2020) G20 Long-term
statistical data
Multiple co-integration
estimation
Driving forces in spreading of RES: research and development (generally), policy,
environmental pressure (high-income countries).
Wang and Wang
(2020)
OECD Long-term
statistical data
Non-linear panel data
analysis
RES has a positive effect on economic growth, especially in case of high energy
consumption.
Vida et al. (2020) World Statistical data Statistical analysis Energy use of RES is limited by the factors inuencing food consumption
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
5
non-linear probability weighting, rational inattention, bounded ratio-
nality, present bias, and status quo bias [62]. Of external constraints,
lack of adequate information is the primary reason for lack of investment
in energy efciency [64,65].
The interactions between social inequalities and household energy
use were analysed in a study by Czibere et al. [3], according to which the
attitude of individuals concerning energy use shows great differences
based on social and demographic variables. There are also
country-specic differences.
Heat energy demand also depends on age, since it is common for
older people to spend large amounts of time in their home [66]. The
effect of aging as a determinant of the use of renewable energy sources is
that a growing proportion of households are spending more and more
time in their own homes. As older people tend to lead more sedentary
lifestyles, they are more sensitive to the temperature of their environ-
ment. As a result, older people tend to set higher temperatures for
themselves at home. In other words, ageing is associated with an
increase in household energy demand. As incomes fall after retirement,
they no longer modernise heating systems, which were outdated by
then. Owners of buildings with poor or no insulation, outdated heating
systems, outdated glazing, draughty doors and windows will not accept
expensive renewable energy sources that would require signicant
capital investment. Older people are also more afraid of change and of
learning and using new technologies than younger people [67]. This is
also true for the adoption of more energy efcient heating and lighting
systems [68]. Willis et al. [69] used conditional logit and mixed logit
models to examine the effect of age and showed that it does indeed affect
the adoption of technologies that affect energy efciency. Older
households are less likely to install microgeneration technologies (solar
thermal, photovoltaic, wind). In other words, an ageing population re-
duces the likelihood of adopting microgeneration technologies in each
country [69].
3. Research questions, materials and methods
3.1. The sample
The data collection was carried out by Z´
avecz Research
1
in
November 2020, the sample is representative of the Hungarian popu-
lation above 18 years of age, by gender, age, level of education and type
Table 2
Sociodemographic characteristics of the sample.
Sociodemographic indicators Sample (N =1000) Sociodemographic indicators HCSO
Gender
Female 53.43 Female
2
52.09
Male 46.57 Male 47.91
Age Age (2016)
18–29 18.14 20–29 15.47
30–39 19.47 30–39 17.48
40–49 16.26 40–49 19.26
50–59 17.75 50–59 15.44
60- 28.38 60- 32.36
Educational level
Elementary or below 27.21 Elementary or below 20.85
Vocational training 22.60 Vocational training 24.52
Secondary school 32.09 Secondary school 32.74
Higher education 18.10 Higher education 21.89
Frequency of using the Internet (%)
Daily 71.14 Almost everyday
3
81.4
At least once a week 4.92 At least once a week 5.5
At least once a month 0.66 At least once a month 0.875
Less frequently than monthly 0.30 Less frequently 0
Never 22.98 Never 12.5
Frequency of shopping online (%)
At least once a week 2.48 At least once in a quarter of a year
4
49.3
At least once a month 11.57
At least once a quarter 16.34
At least once every six months 11.89 At least once 3–12 monthly 10.7
Less frequently 15.67 Less frequently the once a year 5.7
Never 42.05 Never 34.3
Type of residence
Village 29.72 Village
5
31.46
Rural town 35.79 Rural town 31.99
County centre 17.41 County centre 18.89
Budapest 17.08 Budapest 17.67
Employment status
Employee 33.59 Employee
6
58.13
Leader, entrepreneurial with employee 11.39 Leader, entrepreneurial with employee 4.55
Entrepreneurial without employee 24.95 Entrepreneurial without employee 2.38
Retired 25.50 Retired
7
29.68
Other (unemployed, student) 4.57 Other (unemployed, student) 5.26
Financial status
8
0-2 asset 43.60
3-5 assets 39.84
More than 6 assets 16.56
2
https://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_wdsd003c.html.
3
87.85 of the respondents have Internet access; https://www.ksh.hu/docs
/hun/xstadat/xstadat_eves/i_oni017.html.
4
https://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_oni019.html.
5
https://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_zhc060b.html.
6
http://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_qlf001.html.
7
https://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_fsp001.html.
8
Its distribution within society as a whole has not been examined.
1
http://www.zaveczresearch.hu/.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
6
of residence. The questionnaire was part of an omnibus survey, and the
present study focuses on the correlations between attitudes related to
energy use and environmental protection, in addition to the re-
spondents’ set of values and their opinions on climate change. In the
questionnaire, questions related to energy use are followed by questions
on environmental protection. The questionnaire shows well the trends in
Hungary, its questions are suitable for comparison with the results of
previous questionnaires in Hungary [2] and abroad [70], but also reect
the effects of the specic Hungarian residential energy price regulation
(“reduction of residential energy bill”). Table 2 Presents the detailed
description of the sampe.
The following limitations should be taken into account at interpret-
ing our results. We did not ask how condent respondents were that they
would actually buy the house with the options they had chosen, as this
would have allowed us to lter out any unrealistic choices. It would
therefore be worth measuring the proportion of uncertain choices in
future surveys.
In the scientic literature, several authors have found that people
often answer lower amounts [71–73]. Ladenburg and Olsen [74] suggest
that it is possible that the use of so-called “cheap talk” (this involves
describing the hypothetical bias and its causes in the questionnaire)
before WTP questions could reduce this hypothetical bias, but the results
of the literature on this are not clear. We therefore did not apply this
technique.
Our method is a discrete choice experiment, which is a preference
estimation method based on the theory of random utility, it assumes
utility-maximising behaviour known in the eld of microeconomics
[75]. In addition, according to the theory of characteristics, utility levels
of alternatives in a decision set are considered to be derived from their
attributes [76]. Finally, it breaks down utility into a systematic and a
random component [77]. In addition to the application being a
frequently used procedure in many elds: transportation studies [78,
79]; health economy [80,81]; marketing [82,83]; energetics [84,85] in
recent decades, several novel methodological innovations were per-
formed on it. In order to resolve one of the most signicant limitations of
the multinomial logit model associated with McFadden [86], a number
of new specications attempted to capture the heterogeneity inherent in
taste in order to abolish the assumption of homogeneous preferences.
Some of these specications approach by including deterministic com-
ponents, and others by including stochastic components [87]. In their
work, Bujosa et al. [88] and Greene-Hensher [78], present a
bi-directional approach. However, the latest trend is the use of so-called
hybrid choice models, which supplement the standard choice model
with a latent construct as a different member. The basic assumption of
the model is that individuals’ choices are greatly inuenced by different
attitudes and perceptions, which, although not directly observable, can
be incorporated into the context of a hybrid choice model through
related statements and a latent variable [87,89]. In the literature of
recent years, researchers have generally argued that individuals’ pref-
erences are not only inuenced by the characteristics and observable
attributes they examine, but are also related to individuals’ attitudes and
perceptions [90–94]. An appropriate and widely used way to collect
data on attitudes or perceptions is to ask respondents to indicate how
much they agree with it [95–97]. The present study contributes to the
literature with an approach that is becoming increasingly widely used
today by specifying a latent class (LC) model that captures taste het-
erogeneity and simultaneously allocates individuals to classes according
to underlying attitudes that also inuence the answers to a number of
attitudinal questions [98–100]. Hence, this article not only aims at
determining the monetary value of convenience and environmental
features in residential heat energy consumption but also aims at addi-
tionally incorporating individuals’ attitudes toward energy consump-
tion in a hybrid choice model.
Hybrid choice modelling became an increasingly researched topic in
the early 2000s. One of its key methodological issues is described by
Bolduc et al. [101]. The authors point out that the use of the specica-
tion has so far been restricted to small-scale models due to methodo-
logical limitations (solving complex, multidimensional integrals). At the
same time, they point out that the development of computer technology
already allows us to apply more complex models through the use of
different simulation-driven methods. Bolduc et al. [102] analysed the
choices of Canadian residents regarding passenger cars (when faced
with technological innovations) using a hybrid choice model. Their
analysis identied two latent variables, the rst of which was “envi-
ronmental concern” and the second was “appreciation of new car fea-
tures”. Based on their results, several socio-demographic factors
(gender, age, and education of respondents) had a signicant effect on
the explanation of latent variables (among others, for example, women
with a university degree over 56 years of age had a more positive
environmental concern attitude). Daziano and Bolduc [103] used
Bayesian estimation procedure to process their data from the stated
preference approach, which analysed vehicle purchase decisions and
environmental considerations. Their results highlighted that environ-
mentally conscious consumers are willing to pay more for low-emission
vehicles. From a methodological point of view, it was emphasized that
the application of the Bayesian estimation technique is
Table 3
Attributes, their levels and coding.
Attributes Description of attributes Attribute level Coding
Monthly energy cost (thousand
HUF/month)
Includes the cost of energy use, the efciency of use and, in the case of gas,
the base charge (xed cost) *
10 (coal)
18 (rewood, natural gas)
26 (biobriquette)
30 (rewood pellets)
Continuous
variable
One-time investment cost (thousand
HUF)
It includes the heater, its installation and space requirements. 200 (coal, normal rewood boiler)
300 (natural gas)
700 (rewood, biobriquette gasier
boiler)
1000 (pellet boiler)
Environmentally- friendly nature The impact of the chosen heating option on the environment during the
entire life cycle
Very polluting (coal)
Slightly polluting (natural gas,
rewood)
Slightly environmentally- friendly
(biobriquette)
Very environmentally- friendly (re
pellets)
1
2
3
4
Type of operation (convenience) Manual or automatic feeding of fuel Manual: coal, rewood, biobriquette
Automatic: natural gas, re pellets
1
2
Notes: * Excluding personnel costs and maintenance/repair cost, due to their subjective nature.
**The subjective assessment of rewood in Hungary is less favourable than its actual environmental characteristics.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
7
methodologically easier to implement in the context of hybrid models,
and the denition of condence intervals for willingness to pay calcu-
lations is also clearer. Daly et al. [104], on the one hand, provided an
example of the applicability of hybrid modelling in transport study and
highlighted the role of latent attitudes in the context of train travel, on
the other hand, they provided methodological innovation. In contrast to
previous studies (where responses to attitude-type statements were
analysed as a continuous model), ordered logit structure was used to
model indicators related to attitudes. Sarman et al. [105], analysed
decisions on leisure trips through hybrid choice modelling, with a
particular focus on the risks from life-threatening events associated with
travel. The latent variable of their model was dened in the context of
risk-taking. Their results highlighted the impact of different threats,
their potential magnitude, and respondents ’risk-tolerant attitudes on
decisions.
This study uses hybrid latent class modelling to examine the envi-
ronmental and economic sensitivities of the population and to show how
preferences differ across the population of the country under study. The
aim of this approach is to adequately capture individual heterogeneity in
tastes through attitude indicators. Some of the heterogeneity may be
related to sociodemographic characteristics of the respondents, but
unobserved attitudes may also be the main cause of heterogeneity.
Therefore, using this case study from Hungary, we jointly estimate
attitude and choice models, analysing the role of latent attitudes in an
environmental context.
3.2. Experiment design
The research started with a detailed literature review and expert
interviews, with the aim to be able to identify which properties in-
dividuals consider most important in choosing their heating method in
Hungary, and what levels should be considered when comparing them.
Experts on both biomass energy processes and environmental pro-
tection with signicant international publication activity were invited for
the purpose of this task. In addition to the ndings of the eight expert
responses received, the methodology used so far in the international
literature was also taken into account the ndings of the authors’ previ-
ously published paper regarding the Hungarian population’s preferences
[2]. Subsequently, based on a 50-person pilot study, the questionnaire,
decision situation questions, and different levels of attributes were nal-
ised. Four attributes were identied, the levels of which were determined
at 4-4-4-2. The number of attributes was inuenced by the characteristics
of the energy sources taken into account and the technological solutions of
the boilers to be considered during use.
In the case of economic factors, in addition to the above aspects,
market and statistical data were taken into consideration, as follows:
Basic data for determining the monthly energy cost:
•The prices and price ratios of ve energy sources per unit of caloric
value, which cover almost 100% of individual heat consumption in
Hungary, but differ from the point of view of comfort and environ-
mental protection, were taken into account, as follows:
o Prices (thousand HUF/GJ, Internet 1–3)
Coal: 2.55 Natural gas: 2.8 Firewood: 3.0 Biobriquette: 4.1 Fire-
wood pellets: 4.9
o Price ratios (rounded):
Coal: 1 Natural gas: 1.1 Firewood: 1.15 Biobriquette: 1.6 Fire-
wood pellets: 2
•These values already include the typical energy efciency values for
combustion (and the base charge for natural gas to be paid regardless
of consumption), which were as follows (Internet 4–5)
o Coal: 80% Natural gas: 95% Firewood: 80% Biobriquette: 85%
Firewood pellets: 95%
•Finally, the two basic statistical data on individual heating in the
population were also used to determine the values to be developed:
o Energy demand of a 100 m
2
family house with average insulation:
73 GJ/year (Internet-6)
o The amount of the annual household energy cost per capita ac-
cording to income deciles (129–396 thousand HUF/year) and the
average family size (3.01), (Internet 7–8)
Basic data for determining the one-off investment cost:
•Based on the price offers of the most important boiler manufacturers,
the following boiler prices were taken into consideration (thousand
HUF, Internet 5–6):
o Coal: 200 Natural gas: 300 Firewood: 200/700 Biobriquette: 200/
700 Firewood pellets: 1000
o If there are two values, the rst refers to a normal wood-red
boiler and the second to a more advanced wood-gasier boiler.
Table 3 provides a summary of the attributes used and their details.
To compile decision alternatives and situations, the D-efcient
experimental design was chosen and the decision situations (16 pieces)
were arranged into two blocks. Ngene 1.2 [106] was used for imple-
mentation. In the nal questionnaire, respondents faced eight decision
situations, each containing three choices (none of which included “no
answer” as an option, i.e., respondents were faced a so-called “forced
choice” situation). An example of a decision situation is shown in Table 4.
3.3. Model specication
In this paper, in addition to traditional latent class modelling, a
hybrid choice approach is used which, similarly to Mariel et al. [109],
describes a latent class (LC) specication complemented with a latent
Table 4
An example of a decision situation.
1st heating method 2nd heating method 3rd heating method
Monthly energy cost (thousand HUF/month) 26 10 30
One-time investment amount (thousand HUF) 700 1000 200
Environmentally- friendly nature Very polluting Very polluting Very environmentally- friendly
Type of operation Automated Manual feeding Manual feeding
Data were processed and models were estimated using R: Apollo [107,108].
Internet-1:
https://matrabrikett.hu/tuezeloanyag-valasztas.
Internet-2:
https://tuzelocentrum.hu/puspokladany?gclid=EAIaIQobChMIybrk
-9zq6gIVWxV7Ch2XvQbjEAAYASAAEgLoBfD_BwE.
Internet-3:
http://pelletexpert.hu/.
Internet-4:
https://netkazan.hu/.
Internet-5:
http://www.kazanwebaruhaz.hu.
Internet-6:
https://365.reblog.hu/majdnem-a-felere-csokkentette-a-hoszigeteles-
egy-atlagos-csaladi.
Internet-7:
http://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_zhc021b.html.
Internet-8:
https://www.ksh.hu/docs/hun/xstadat/xstadat_eves/i_zrk007b.html.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
8
variable (hereinafter referred to as a hybrid latent class (HLC) model).
As a rst step, the traditional multinomial logit (MNL) model is
estimated in order to obtain basic information about the sign of the
characteristics involved in the study, as well as their relative weight as a
function of utility. As a next step, a latent class (LC) model was used, i.e.,
distinct heterogeneous classes were formed within which members have
homogeneous preferences [110]. The class allocation equation of the LC
model is supplemented with a latent variable and 12 attitude-related
statements are modelled (in the context of environmental awareness)
with measurement equations.
3.3.1. Multinomial logit model (MNL)
The biggest advantage of the MNL model attributed to McFadden
(1973) is that its estimation process and interpretation of its results can
be performed relatively easily. However, it also has several disadvan-
tages, most notably the assumption of independence of homogeneous
preferences and irrelevant alternatives. In the case of the model based on
the RUT (Random Utility Theory) approach, the systematic part of the
total utility (Equation (1)) can be written down according to Equation
(2).
Un,i,t=Vn,i,t+
ε
n,i,t,(1)
where Un,i,t is the total utility of decision-maker n related to alternative i
in the decision situation t, Vn,i,t is the systematic part of utility (arising
from the observed characteristics), while
ε
n,i,t is the random (non-
observable) part of utility.
Vn,i,t=βXn,i,t,(2)
where β is the coefcient related to the observed variable, and X is the
observed variable.
In the model, the probability of choosing alternative i of the elements
of the decision set J related to decision maker n in the decision situation t
can be expressed according to Equation (3):
Pn,i,t=exp βXn,i,t
J
j=1
exp βXn,j,t
(3)
3.4. Latent class model (LC)
The LC specication is able to address one of the major disadvan-
tages of the MNL model, the assumption of homogeneous preferences,
through forming a discrete number of classes. Separate classes, within
which member preferences are already homogeneous, have separate β
parameters for the studied properties. In the case of the model, the
systematic part of the utility can be expressed according to Equation (4)
[111].
Vn,i,t=βqXn,i,t,(4)
where βq expresses the estimated coefcient for the observed variable
and class q (q =1, …,Q).
In the model, the probability of choosing alternative i of the elements
of the decision set J related to decision maker n in class q in the decision
situation t can be expressed according to Equation (5):
Pn,i,t|q=expβqXn,i,t
J
j=1
expβqXn,j,t
q=1,…,Q,(5)
It is clear from Equation (5) that it is structured according to a similar
structure as in the case of the MNL model (Equation (3)), however, from
the aspect of the LC, it is modied according to Equation (6) in order to
determine the probability of individuals belonging to each class.
Pn,i,t=
Q
q=1
Pn,i|qHn,q,(6)
where Hn,q is the probability of individual n belonging to class q [112].
In the case of the model, it is difcult to identify the ideal number of
classes. This is usually decided on the basis of information criteria,
mostly AIC (Akaike information criterion), CAIC (Consistent AIC) and
BIC (Bayesian information criterion) [113].
3.4.1. Hybrid latent class model (HLC)
The purpose of estimating so-called “hybrid” or “latent variable”
models is to incorporate directly non-measurable effects (e.g., attitudes,
perceptions) into the model described in this paper, as they also form an
essential part of individuals’ decision-making processes [114]. On the
one hand, these effects appear through the standard choice model
(Equation (1)), supplementing it in a way that corresponds to Equation
(7). On the other hand, they appear through measurement equations
related to different attitudes (Equation (9)).
Un,i,t=Vn,i,t+λLVn+
ε
n,i,t,(7)
where LVn is the latent variable for individual n, while λ shows its effect.
Among other characteristics, hybrid models are built from structural
(describing the structure of latent variables as a function of observable,
explanatory variables, in a typical utility function formula) and mea-
surement equations (linking the latent variable(s) to questions related to
different attitudes). This research includes a very important latent var-
iable that represents respondents ’attitudes toward environmental
awareness, and its structure can be described according to Equation (8)
[102,104].
LVn=γFn+
η
n,(8)
where γ is the coefcient estimated for the observed socio-demographic
characteristic; Fn is the variable related to the observed socio-
demographic characteristic; while
η
n is the random member that is
assumed to have normal distribution.
The measurement equations related to the decision-maker n (k =1,
…K), where the answers given to statements represent the dependent
variable, can be expressed in the structure of Equation (9).
MEk,n=ζkLVn+
σ
k,n,(9)
where ζk is the coefcient estimated for the latent variable in question k,
LVn is the latent variable, while
σ
k,n is the random part of the mea-
surement model in relation to decision-maker n and question k.
3.5. Utility function, class allocation, and structural equations
In the models described in this paper, the systematic part of the
utility is constructed according to Equation (10).
Vn,i,t=ASCi+βPricePricen,i,t+βCost Costn,i,t+βEnvMedium EnvMediumn,i,t
+βEnvHigh EnvHighn,i,t+βEnvVery highEnvVery highn,i,t+βComfortManual ComfortManualn,i,t,
(10)
〖ASC〗 _i is the specic constant value of the alternative for the i-th
alternative (set to 0 for Alternative 1 in each case); Price, Cost, Env,
Comfort denote the properties included in the study, of which Table 2
provides a detailed overview. The baseline levels (for discrete variables)
always included Low (for Environmental friendliness) and Automatic
(for Comfort).
Where ASCi is the specic constant value of the alternative for the i-
th alternative (set to 0 for Alternative 1 in each case); Price, Cost, Env,
Comfort denote the properties included in the study, of which Table 2
provides a detailed overview. The baseline levels (for discrete variables)
always included Low (for Environmental friendliness) and Automatic (for
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
9
Comfort).
In the case of the latent class (LC) model described here, only one
constant term was included in the class allocation equation according to
Equation (11), while in the case of the hybrid latent class (HLC) it was
extended it with the latent variable (Equation (12)).
An,q=δq,(11)
where δq denotes the constant of the q-th class (one of them is set at 0,
only q-1 constants are to be estimated).
An,q=δq+φqLVn,(12)
where φq denotes the coefcient estimated as a result of the latent
variable for the q-th class, while LVn denotes the latent variable for the n-
th individual.
The structural equation for the latent variable can be expressed ac-
cording to Equation (13).
LVn=γEducation2Education2n+γEducation3Education3n+γEducation4Education4n
+γResidence2Residence2n+γResidence3Residence3n+γResidence4Residence4n
+γWealth2Wealth2n+γWealth3Wealth3n+
η
n,
(13)
where Education, Residence and Wealth are socio-demographic vari-
ables, as described in detail in Table 1.
The measurement equations in this paper were dened in the
structure of Equation (9) and were estimated for the following 12 Likert-
scale type statements [3,115]. The scales represent four different atti-
tudes: the rst three of them measure environmental identity, the
fourth-sixth measure the role attributed to governments, the
seventh-ninth help to understand the role of personal norms in energy
use, while the last three focuses on the role of social norms in shaping
attitudes inuencing energy use. Individuals with a strong environ-
mental self-identity consider active participation in environmental ac-
tivities to be a key characteristic of their identity [116]. Personal
standards represent how people feel about their moral commitment to
energy-saving behaviour [117]. Corporate environmental responsibility
means increasing the environmental performance of organizations and
reducing their environmental impact [118]. Social norms, on the one
hand, include how people considered as references reduce their energy
use and how they think about what an individual can do [119].
•Acting pro-environmentally is an important part of who I am.
(Statement 1)
•I am the type of person who acts pro-environmentally. (Statement 2)
•I see myself to be a pro-environmental person. (Statement 3)
•I think the government has a goal to minimise its impact on the
environment. (Statement 4)
•I think the government has implemented policies and procedures to
minimise its impact on the environment. (Statement 5)
•I think the government has stated its mission to implement a sus-
tainable (pro-environmental) policy. (Statement 6)
•I feel morally determined to save energy. (Statement 7)
•It would t my standards if I used sustainable energy. (Statement 8)
•I feel personally responsible to try to save energy. (Statement 9)
•Most of the people who are important to me think I should try to use
as little energy as possible. (Statement 10)
•Most of the people who are important to me will approve if I try to
use as little energy as possible. (Statement 11)
•Most people who are important to me try to use as little energy as
possible. (Statement 12)
Descriptive statistics on responses to the statements are presented in
Table 5.
Responses ranged from 1 (strongly disagree) to 5 (strongly agree) on
a scale. The table shows that the average exceeded 3 in all cases, which
means that respondents tended to be in agreement with the above
statements. It can also be concluded that respondents are the least
convinced of the government’s commitment to the environment, as the
proportion of those who agree is the lowest. However, there is also the
greatest disagreement among respondents, as the standard deviation is
the highest in the case of these questions. Personal standards and soci-
etal expectations regarding environmental protection are particularly
high; there is less disagreement among respondents than in case of
previous responses. A recently published paper arrives at similar results
[3]
,
.
9
4. Results and discussions
This chapter presents the results of the model estimates according to
the structure presented in the model specication section. As a rst step,
the estimation of the multinomial logit (MNL) specication is
Table 5
Descriptive statistics of the examined statements.
Statement 1
(%)
2 (%) 3 (%) 4 (%) 5 (%) Mean Standard
deviation
Statement
1
1.86 6.97 28.22 37.86 25.09 3.77 0.96
Statement
2
2.32 7.78 30.89 36.70 22.30 3.69 0.98
Statement
3
2.79 8.36 27.29 36.93 24.62 3.72 1.01
Statement
4
4.76 11.61 31.36 34.03 18.23 3.49 1.06
Statement
5
6.85 16.84 34.84 26.83 14.63 3.26 1.11
Statement
6
6.85 13.24 30.66 32.06 17.19 3.39 1.12
Statement
7
1.63 6.50 25.09 40.07 26.71 3.84 0.95
Statement
8
1.39 4.99 24.16 40.42 29.04 3.91 0.92
Statement
9
1.63 6.62 25.90 40.19 25.67 3.82 0.95
Statement
10
2.09 5.11 27.99 40.53 24.27 3.80 0.94
Statement
11
1.63 4.88 26.48 39.61 27.41 3.86 0.93
Statement
12
2.09 5.92 27.87 40.19 23.93 3.78 0.95
Table 6
Results of the MNL model estimates.
Properties and data describing the model Estimates Robust t-values
ASC alternative 2 −0.0091 −0.32
ASC alternative 3 −0.2401*** −7.68
Price/1000 −0.0351*** −17.22
Investment costs/10 000 −0.0110*** −18.31
Slightly polluting 0.4452*** 7.52
Slightly environmentally- friendly 0.7723*** 12.61
Very environmentally- friendly 1.2202*** 24.27
Manual −0.2543*** −7.37
Individuals 861
Observations 6888
Parameters 8
Log-likelihood (nal) −6310.684
Pseudo R2 0.1661
AIC 12637.37
BIC 12692.07
Note: ***, **, * indicate that the coefcients are signicant at the 1%, 5% and
10% levels, respectively.
9
Factors inuencing environmental attitudes are examined in detail in a
different study.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
10
performed, followed by the latent class (LC) and hybrid latent class
(HLC) models. It is important to mention that, due to ltering out
incomplete respondents, a sample of 861 people is used during the
model estimations.
4.1. Results of the multinomial logit (MNL) model estimation
The results of the estimates of the MNL model constructed according
to the utility function specication of Equation (10) are shown in
Table 6.
Based on the results of the MNL model, it can be concluded that there
were some heuristics in the decisions, shown by the fact that Alternative
3 was chosen signicantly less often than Alternative 1 at the base level
(which is shown by the signicant coefcient of the “ASC Alternative
3”). As previously expected, price and investment cost have a negative
effect on individuals’ sense of utility (with price having a more signi-
cant effect), i.e., an increase in these factors is accompanied by a
decrease in utility levels. As the level of pollution decreases, consumers’
sense of utility increases, and the manual heating alternative is consid-
ered less preferred to the automatic solution. All estimated parameters
for the examined characteristics were signicant (at the 1% level),
suggesting that the characteristics that most inuenced individuals
’decisions were included in the experiment. Our results show great
similarity to the ndings of Khan et al. [38] and Hu and Wang [39],
highlighted the importance of automatized technologies in the reduction
of carbon dioxyde emissions, the tendency in Hungary ts to the West-
ern European characteristics. Considering Yu et al. [37] statements,
spreading these technological innovations is also a great reserve in
improving carbon productivity and GDP.
Results of the latent class (LC) model estimation.
To eliminate the homogeneous preferences assumed by the MNL
model, LC model estimates were also performed in order to distinguish
classes with different preferences. For this purpose, several cases with
different numbers of classes were tested and the version considered to be
the most ideal from the statistics aspect was further analysed. The values
of the indicators forming the basis for the authors’ choice are shown in
Table 6. The structure of our model was based on the research method of
Mills and Schleich [34] and Mariel et al. [35].
Based on the results of, It can be clearly seen that the values of the
information criteria (log-likelihood (nal), AIC, BIC) showed a decrease
even in the 6-class case compared to the previous 5-class case (Table 7).
Although this conclusion suggests a better t of the 6-class model, the
authors nevertheless decided to analyse the 5-class specication. This
choice was justied by the fact that in the 6-class case, a group with a
rather low (<10%) class probability value was already visible, the
number of signicant parameters decreased signicantly in the case of
the model. Based on these conclusions, the performed research focused
on 5-class models (for both the LC and HLC models). The results ob-
tained for the LC model are shown in Table 8.
The applied typology is necessary to understand how the different
consumer groups will react on (1) energy policy legislation, (2) what
kind of tools are necessary to reach the desired goals of energy strategy –
considering the proportion of the cluster members (3) what are the ef-
fects of the energy policy decisions at country level.
Table 7
Values of information criteria for LC models with different numbers of classes.
Information criteria 2 Classes 3 Classes 4 Classes 5 Classes 6 Classes
Parameters 15 22 29 36 43
Log-likelihood (nal) −5857.476 −5655.795 −5556.196 −5452.076 −5380.438
AIC 11744.95 11355.59 11170.39 10976.15 10846.88
BIC 11847.52 11506.02 11368.68 11222.3 11140.89
Class probability values 0.72 0.28 0.22 0.29 0.23
0.19 0.17
0.27 0.28 0.12 0.07
0.28 0.28 0.12
0.45 0.21 0.19 0.20
0.22 0.20
Table 8
Results of the LC model estimates.
Characteristics and data
describing the model
Neutral Cost sensitive Comfort and
environmental concern
Sensitive to energy price Environmental concern
Class probability 0.19 0.19 0.29 0.12 0.22
Estimates Robust t-values
ASC alternative 2 0.0437 1.05
ASC alternative 3 −0.4061*** −8.02
Estimates Robust t-
values
Estimates Robust t-
values
Estimates Robust t-
values
Estimates Robust t-
values
Estimates Robust t-
values
δ Base 0.0090 0.04 0.4276** 2.04 −0.4236* −1.67 0.1394 0.59
Price/1000 −0.0155*** −2.72 −0.0847*** −3.72 −0.0424*** −5.08 ¡0.1680*** −9.19 0.0137 0.76
Investment costs/10 000 −0.0048*** −2.86 −0.0580*** −6.01 −0.0163*** −5.37 −0.0066** −2.07 −0.0020 −0.64
Slightly polluting 0.1913 1.05 0.7128** 2.03 1.7284*** 9.98 −0.0369 −0.12 0.3238 0.55
Slightly environmentally-
friendly
0.5535*** 3.54 1.3092*** 3.08 1.8057*** 7.84 0.2767 0.97 2.4384*** 9.81
Very environmentally-
friendly
0.1814 1.17 0.4181 1.28 1.9949*** 8.60 0.4881** 2.04 3.8230*** 13.03
Manual 0.0437 0.33 0.0719 0.60 ¡1.3750*** −4.96 0.1417 0.80 −0.4743 −1.58
Individuals 861
Observations 6888
Parameters 36
Log-likelihood (nal) −5452.076
Pseudo R2 0.2795
AIC 10976.15
BIC 11222.3
Note: ***, **, * indicate that the coefcients are signicant at the 1%, 5% and 10% levels, respectively.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
11
4.2. Social characteristics of LC
The latent classes are briey described below based on socio-
demographic aspects (Table 9).
The Neutral class is one of the youngest groups (class 1), with the
lowest proportion of those over 60 and the highest of those under 40
years of age. The educational level of this group is average. Members of
the group are above-average daily Internet users, as it is one of the most
Internet-using groups (following class Environmental concern). There is
a predominance of those who do not shop regularly online, but the
proportion of those who never shop online is also below average (42% of
the sample never shop online). These respondents mostly belong to the
metropolitan population, they live in county seats and in the capital and
they have the best nancial situation in the sample. The average level of
those in weak and average nancial situation and the above-average
level of those in very good nancial situation is a good indicator, as
the proportion of those who own the most assets is the highest in this
class. Most of the members have no children, even though this class has
the most households with one and two children. The employment
structure of the group is mainly characterised by the high proportion of
employees and individual entrepreneurs and the below-average pro-
portion of pensioners.
The cost sensitive class is one of the oldest groups (class 2), with an
above-average proportion of people over 60 (one-third of the group), but
the proportion of young people under 30 is also one of the highest. This
is the least educated class, with most respondents on an elementary
school or vocational school level and there are no respondents with
higher education degree in this class. Many (one in four people) never
use the Internet, daily users are slightly below average, and people are
more likely to use the Internet once a week or once a month than the
average. Almost half of them never buy online, and those who do shop
only very rarely. These respondents live mainly in larger cities (county
seats and in the capital), with rural and small-town populations being
Table 9
Sociodemographic characteristics of the respondents in the sample and latent classes.
Sociodemographic Factors Sample (N =
1000)
Neutral
(19%)
Cost sensitive
(19%)
Comfort and environmental
concern (29%)
Sensitivity to energy
price (12%)
Environmental concern
(22%)
Gender (%)
Female 53.43 50.08 52.86 56.12 56.13 51.17
Male 46.57 49.92 47.14 43.88 43.87 48.83
Age category (%)
18–29 18.14 16.67 19.39 15.96 13.37 22.21
30–39 19.47 26.24 18.95 16.93 15.38 18.88
40–49 16.26 18.84 11.52 19.16 14.68 15.75
50–59 17.75 15.50 16.40 18.96 14.48 20.69
60- 28.38 22.75 33.74 28.99 42.09 22.47
Highest level of education (%)
Elementary or below 27.21 10.10 18.17 7.95 14.76 3.81
Vocational training 22.60 35.79 43.30 32.39 42.28 21.31
Secondary school 32.09 40.60 36.15 43.12 36.30 48.68
Higher education 18.10 13.51 2.38 16.54 6.66 26.20
Frequency of using the Internet (%)
Daily 71.14 82.91 64.77 77.88 63.28 97.62
At least once a week 4.92 5.47 8.15 5.70 7.11 2.37
At least once a month 0.66 0.07 1.48 0.48 1.22 0.01
Less frequently than monthly 0.30 0.00 0.00 0.00 0.00 0.00
Never 22.98 11.55 25.60 15.94 28.39 10.00
Frequency of shopping online (%)
At least once a week 2.48 3.81 2.34 2.52 3.82 2.07
At least once a month 11.57 15.34 7.13 14.30 8.12 15.57
At least once a quarter 16.34 18.78 14.36 18.95 14.46 23.43
At least once every six months 11.89 11.32 18.93 11.99 13.26 6.71
Less frequently 15.67 17.21 12.59 19.40 13.79 19.12
Never 42.05 33.54 44.65 32.84 46.55 33.10
Type of residence (%)
Small town, village 29.72 17.11 13.29 13.32 14.52 33.84
City 35.79 17.74 15.16 22.94 18.12 14.07
County seat 17.41 35.59 36.28 35.53 38.74 32.50
Budapest 17.08 29.56 35.27 28.21 28.62 19.59
Financial situation (%)
10
0-2 assets, weak nancial situation 43.60 38.28 53.17 31.3 53.01 32.28
3-5 assets, average nancial
situation
39.84 38.57 38.26 47.80 30.90 46.52
More than 6 assets, very good
nancial situation
16.56 23.15 8.57 20.90 16.09 21.20
Persons under 18 years of age in the household (%) (households with children)
0 72.05 68.92 74.86 72.79 81.77 78.55
1 15.83 19.50 14.18 17.70 13.68 13.20
2 12.12 11.58 10.96 9.51 4.55 8.25
Occupation of the respondent (%)
Employees 33.59 35.04 35.98 39.14 33.88 37.12
Leaders, or entrepreneurs with
employees
11.39 11.96 4.02 13.60 7.67 14.88
Entrepreneurs without employees 24.95 32.12 27.69 22.88 18.18 24.65
Pensioners 25.50 17.77 30.47 21.97 35.76 21.04
Other (unemployed, student) 4.57 3.11 1.84 2.41 4.51 2.31
10
The nancial situation of the respondents was measured by the following
items: owning a car (less then 10 years old), a second home, a motocycle, a LED
TV, a play console, a video-camera, a printer, a dishwasher, a laptop, or an
automatic coffee machine.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
12
well below average. This class has the worst nancial situation (similar
to class Comfort and environmental concert). Only a few households
have children, mostly one child. The proportion of employees and
pensioners is almost identical (about 30%), but the presence of indi-
vidual entrepreneurs is also the most typical in this class, in addition to
the neutral class.
The Comfort and environmental concern class is one of the “most
feminine” groups (class 3), with the highest proportion of women in
addition to class Comfort and environmental concert. Among the women
in the class, those above 60 years of age represent the highest proportion
(29%), but all other age groups have a relatively proportionate presence
of women. The comfort and environmental concern class is a relatively
highly educated group, with the predominance of higher education and
secondary education degrees, while the proportion of those with
elementary degree is the lowest. The vast majority (77%) use the
Internet on a daily basis. They do not shop online frequently, as one third
of them never shop online, and others do so only very rarely. Most re-
spondents in this class live in cities, county seats and in the capital, while
small towns and villages are rare. They have relatively good nancial
situation, with the majority having a medium level of possessions, but
the proportion of people in adverse nancial conditions is also the
lowest here (still about one third of the group). The majority have no
children, and households with children typically raise one child. The
proportion of employees is the highest in this class, while the proportion
of pensioners and entrepreneurs is not outstanding, but rather around or
below average.
The Sensitivity to energy price class (class 4) is also characterised by a
slight predominance of women, with an exceptionally high proportion of
those over 60 (this is the oldest group, 42%), but that of people under 30
years of age is also the lowest. Most of them have secondary education
degrees, while the number of people with higher education degrees is
minimal (6.7%) and the proportion of elementary education degrees is
relatively high (15%). The share of daily Internet users is lowest (63%)
and the proportion of those who have never used the Internet is also the
highest (28.4%). Typically they are short of knowledge and - according
to the ndings of [27,32–34] – they are less interested in the investment
into RES technologies. Furthermore, a signicant number of people
never shop online (47%). Class sensitivity to energy price consists of
urban population, with people living mainly in county seats and in the
capital, small towns and villages are not dominant. In addition to class
Sensitivity to boiler price, it is the group with the weakest nancial
situation, more than half of them have no more than two assets. The
majority are childless (82%), and those who have children usually raise
one child. The proportion of pensioners is the highest (35.8%) in this
class and that of employees and individual entrepreneurs is the lowest.
In the Environmental concern class (class 5), the proportion of those
under 30 years of age and people in their 50’s is above average, but that
of the elderly is not signicant. Elementary education degree is almost
non-existent in the class, with the proportion of secondary education
and higher education degrees being extremely high. This is the most
educated group. Almost all respondents in the group use the Internet
every day, although one-third of them never shop online, and those who
do so usually shop online quarterly or less frequently. This class is
characterised by the highest proportion of those living in villages (34%)
and in county seats (32.5%). A signicant part of the group is in a me-
dium nancial position (46%), but the proportion of those in weak
nancial conditions is also notable (32%). At the same time, people in
good nancial situation have the largest proportion in this group,
although it is only 20%, but no other class have such high value, apart
from class Comfort and environmental concern. The number of children
is also low, with 79% of households not raising any children. The pro-
portion of employees is high, but the proportion of individual entre-
preneurs and joint ventures is also signicant, while the proportion of
pensioners remains below average.
4.3. Analysis of willingness to pay (WTP)
Based on the LC model estimates shown in Table 8, negative and
signicant coefcients were obtained for price and investment costs
(consumer price sensitivity decreases with increasing price and invest-
ment cost) in the case of four classes (as expected) and no signicant
effect was shown on these factors for one class (class 5). In terms of
environmental awareness, two classes (class 3 and class 5) have envi-
ronmentally conscious behaviour. Within these classes, the sense of
utility increases as the level of environmental protection increases.
Regarding the amount of work required during heating, a signicant
effect appears only for one class (class 3). Members of this class prefer
automatic operation to manual operation.
In the next step, the research focused on the willingness to pay (WTP)
which characterises the different classes for the examined attributes.
The results are shown in Table 10.
The results in Table 10 clearly show that signicant willingness to
pay values for all factors were obtained only in one class (Comfort and
environmental concern – class 3). Members of this class would pay
approximately 47 000 HUF extra per month for the very
environmentally-friendly heating alternative, as opposed to the very
polluting one; in addition, they would pay approximately 32 thousand
HUF less in a month if they had to heat manually, as opposed to auto-
matic feeding. These factors have a great signicance in the modern-
isation of existing family houses [32].
The provided minimum and maximum values (10 and 30 thousand
HUF/month, respectively) suggest that the members of Class 3 (Comfort
and environmental concern), which makes up 29% of the population,
would not burn coal in any way, but they would choose other, more
environmentally-friendly energy sources, even in the case of a much
larger price difference than the current one. At the same time, the dif-
ference between the other energy sources is small, but it follows an
environmentally-friendly nature. It is noteworthy that there would be a
potential demand for rewood pellets (as a convenient and also
environmentally-friendly fuel) even in the case of a signicant price
increase of up to 100%. In the questionnaire, a higher-than-average
proportion of wealthier and younger respondents occurred in this clus-
ter, which may also explain higher willingness to pay, the importance of
convenience, and environmental awareness. Under the current price
Table 10
WTP calculation for the LC model.
Levels of Attributes Willingness to pay
Neutral – class
1
Cost sensitive –
class 2
Comfort and environmental concern –
class 3
Sensitivity to energy price –
class 4
Environmental concern –
class 5
Slightly polluting n.s. 8419.0*** 40 736.4*** n.s. n.s.
Slightly environmentally-
friendly
35 775.9** 15 463.7*** 42 556.5*** n.s. n.s.
Very environmentally-
friendly
n.s. n.s. 47 016.0*** 2904.4* n.s.
Manual n.s. n.s. −32 405.4*** n.s. n.s.
Note: ***, **, * indicate that the coefcients are signicant at the 1%, 5% and 10% levels, respectively.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
13
conditions, the use of natural gas (comfortable but slightly polluting)
and bio-briquettes (slightly environmentally-friendly but inconvenient)
and the transition to these fuels are also likely to occur in clusters that
make up two-thirds of the population. This tendency is in line with
Western European trends, where solvency is less of a constraint on actual
purchases. It should be noted that, according to the authors’ previous
research [2], a signicant part of the Hungarian population classies
rewood as “very polluting”, therefore, the dissemination of knowledge
and the distribution of high-efciency equipment could be a step for-
ward for this fuel type, especially when granting permits for new houses,
as recommended also by Michelsen-Madlener [32].
After determining willingness to pay, calculations were performed to
determine willingness to invest (WTI), based on the fact that, according
to Rouvinen-Matero [33], investment costs are the primary inuence on
household decisions. The results of this research, covering several
countries with signicantly different economic status (Table 11).
Table 11 presents signicant values for all characteristics were ob-
tained for the same class (Comfort and environmental concern – class 3)
as in the case of willingness to pay. Members of this class would invest
more than 1 million HUF in order to ensure that their heating is not very
harmful to the environment. In addition, they would invest less than
approximately 840 thousand HUF if their heating required manual
feeding, as opposed to the automatic solution.
These results essentially support the authors’ ndings regarding
Tables 10 and i.e. members of Class 3 (Comfort and environmental
concern) would prefer not only to burn the comfortable and very
environmentally-friendly rewood pellets, but also to purchase a much
more expensive boiler that allows to burn rewood pellets. In addition,
the amount of money to be paid for boilers running on increasingly
environmentally-friendly fuels shows little difference compared to coal-
red boilers.
As in the previous case, two-thirds of the population is willing to pay
for slightly more expensive boilers capable of burning natural gas and
biobriquettes. However, convenient use is not relevant for about 70% of
the population, which can be explained by the existing natural gas boiler
(that is not suitable for burning the other fuels studied here) and less
favourable income situation (i.e. favouring the cheapest possible solu-
tions). The latter explanation is supported by the maximum value of
rewood boilers in the slightly environmentally-friendly category
among the members of the “Neutral” class (Class 1) with the largest
share. In addition, this nding is closely related to the fact that the
importance of natural gas and rewood combustion and the proportion
Table 11
WTI calculation for the LC model.
Levels of Attributes Willingness to invest
Neutral – class
1
Cost sensitive –
class 2
Comfort and environmental concern –
class 3
Sensitivity to energy price –
class 4
Environmental concern –
class 5
Slightly polluting n.s. 122 924.0** 1 059 665.0*** n.s. n.s.
Slightly environmentally-
friendly
1 154
470.0***
225 783.0*** 1 107 012.0*** n.s. n.s.
Very environmentally-
friendly
n.s. n.s. 1 223 014.0*** n.s. n.s.
Manual n.s. n.s. −842 953.0*** n.s. n.s.
Note: ***, **, * indicate that the coefcients are signicant at the 1%, 5% and 10% levels, respectively.
Table 12
Results of the HLC model estimates.
Characteristics and data
describing the model
Neutral (class 1) Cost sensitive (class 2) Comfort and
environmental concern
(class 3)
Sensitivity to energy price
(class 4)
Environmental concern
(class 5)
Class probability 0.20 0.18 0.30 0.12 0.21
Estimates Robust t-values
ASC alternative 2 0.0454 1.12
ASC alternative 3 −0.4061*** −8.28
Estimates Robust t-
values
Estimates Robust t-
values
Estimates Robust t-
values
Estimates Robust t-
values
Estimates Robust t-
values
δ 0.1994 0.82 Base 0.5242** 2.37 −0.3881 −1.44 0.1939 1.06
φ 0.3396** 2.08 Base 0.9411*** 5.49 −0.0090 −0.05 0.8573*** 5.42
Price/1000 −0.0174*** −3.07 −0.0902*** −4.56 −0.0399*** −6.21 ¡0.1697*** −9.85 0.0178 1.47
Investment costs/10 000 −0.0048*** −3.11 ¡0.0605*** −7.39 −0.0174*** −6.04 −0.0073** −2.20 −0.0011 −0.51
Slightly polluting 0.1581 0.99 0.8630** 2.22 1.7499*** 9.32 −0.0570 −0.20 0.2511 0.61
Slightly environmentally-
friendly
0.5703*** 3.90 1.3683*** 3.45 1.8657*** 7.92 0.2450 0.87 2.4676*** 9.09
Very environmentally-
friendly
0.1897 1.40 0.3145 1.11 2.0627*** 10.06 0.5077** 2.25 3.9111*** 13.98
Manual 0.0059 0.04 0.0796 0.66 ¡1.3052*** −7.40 0.1640 0.95 −0.5432*** −2.62
Individuals 861
Observations 6888
Parameters 72
Log-likelihood (nal) for
the total model
−17038.21
Log-likelihood (nal) for
the standard choice
model
−5440.419
Pseudo R2 (for choice
model)
0.281
AIC 34220.42
BIC 34712.73
Note: ***, **, * indicate that the coefcients are signicant at the 1%, 5% and 10% levels, respectively.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
14
of heating costs in residential combustion are much higher in Hungary
than the EU average [11].
In order to learn more about the underlying factors of decision
making, we estimated a hybrid latent class (HLC) model in the next step
(Table 12).
Based on the estimates of the HLC model (Table 12), similar con-
clusions can be drawn for the attributes as in the case of the LC model
(Table 8). However, the comparison of models shows that a model with a
better t was obtained (Pseudo R2: 0.281) by including the latent var-
iable. The additional information obtained from the HLC model appear
in three parts (class allocation, structural, and measurement equations).
These results are shown in Table 13.
The results in Table 13 clearly show that three socio-demographic
variables (education level, residence type, wealth type) were included
in the structural equation. Based on the obtained results, it can be
concluded that the value of the latent variable (environmental aware-
ness) will be more positive for people living in the city, who have more
than three assets, higher education degree (college, university degree) or
only elementary school level (although the conclusion on higher edu-
cation was not signicant in the model). The ζ values in the measure-
ment equations (which represent the effect of the latent variable in the
equations) have a positive value for each statement, indicating that
these statements will be rated higher as the latent variable increases. As
an example, the higher the level of environmental awareness among
respondents, the more they agree with the statement “Environmental
awareness is an important part of my self-image”. According to pa-
rameters of the class allocation model, the φ values (which represent the
effect of the latent variable in the class allocation model) show a sig-
nicant effect in several cases. We can conclude that people with a
higher latent variable value (higher environmental awareness) are more
likely to belong to classes 3 and 5. This conclusion is clearly reected in
respondents’ choices and in the coefcients estimated for the attributes
(in the case of class 3 and 5, a clear increasing trend was observed in the
level of utility, at the same time as the level of environmental pollution
decreased).
The different classes were named according to their characteristics
based on the LC and HLC model estimates, as well as the calculations of
willingness to pay and invest. The estimated cost coefcients of the
models were the highest in absolute terms in the fourth class, while the
coefcients estimated for investment cost were the highest in absolute
terms in the second class. For this reason, the fourth class was named
“Price sensitive” and the second “Cost sensitive”. The presence of envi-
ronmental awareness can be clearly assumed for the third and fth
classes (this conclusion is conrmed on the one hand by the coefcients
estimated for the attributes and on the other hand, in the case of the HLC
model, by the parameters of the class allocation model). In the former
case, there is also a clear need for automated operation, which is also
conrmed by calculations of willingness to pay and invest. Accordingly,
the third class was named “Comfort and environmental concern” and the
fth was named “Environmental concern”. Since no clear trend could be
observed in the rst class, it was named “Neutral”.
5. Conclusions
One of the vital question of future energy-use is people’s willingness
to pay for convenience and environmental friendliness, and the factors
inuencing their choices. The article identies the most important social
determinants of energy consumption, using a nationally representative
data basis, determines groups of consumers with markedly different
preferences and quantify their expectable nancial decisions when
purchasing energy and heating systems.
The value of environmental-friendliness nature in Hungary exceeds
the convenience factor of the population in the assessment of energy
sources. This nding is supported by results for both fuels and boilers
cases with signicant differences. The very environmentally-friendly
alternative exceeds the extra cost of comfort by about 50% for fuels
and by almost 20% for boilers.
Signicant results indicate different values among members of the
“Neutral” and “Comfort and environmentally concerned” clusters in the
case of fuel and boiler purchases. The former would pay about 20% less
for slightly more environmentally-friendly fuels than the latter, and
about 5% more for the boiler that burns such fuel. The values of comfort
cannot be signicantly detected for the majority of the Hungarian
population. These ndings show the strong attachment of the Hungarian
average consumer to the use of rewood and natural gas that are still
signicant in the heat consumption of the population.
We also found a segment that accounts for 29% of the total popu-
lation where the need for environmentally-friendly and also convenient
solutions is clearly identiable. These people would in principle be
willing to pay more for fuel and equipment that can be considered
Table 13
The results of the HLC model – structural equation, class allocation model and measurement equation parameters.
Structural equation parameters Estimates Robust t-values Class allocation model parameters Estimates Robust t-values.
γEducation2 −0.376*** −4.09 δ (Class 1) 0.199 0.82
γEducation3 −0.173 −1.57 φ (Class 1) 0.340** 2.08
γEducation4 0.176 1.32 δ (Class 3) 0.524** 2.37
γResidence2 0.173* 1.92 φ (Class 3) 0.941*** 5.49
γResidence3 0.124 1.17 δ (Class 4) −0.388 −1.44
γResidence4 0.057 0.46 φ (Class 4) −0.009 −0.05
γWealth1 0.180** 2.10 δ (Class 5) 0.194 1.06
γWealth2 0.232* 1.82 φ (Class 5) 0.857*** 5.42
Measurement equation parameters Estimates Robust t-values Measurement equation parameters Estimates Robust
T-values
ζq1 0.763*** 24.84 ζq7 0.769*** 26.07
σ
q1 0.586*** 28.61
σ
q7 0.551*** 25.94
ζq2 0.784*** 25.17 ζq8 0.731*** 24.24
σ
q2 0.581*** 30.26
σ
q8 0.559*** 24.40
ζq3 0.826*** 26.98 ζq9 0.791*** 25.15
σ
q3 0.585*** 30.72
σ
q9 0.514*** 28.24
ζq4 0.483*** 10.82 ζq10 0.709*** 20.58
σ
q4 0.948*** 31.71
σ
q10 0.607*** 21.69
ζq5 0.443*** 9.61 ζq11 0.727*** 21.97
σ
q5 1.017*** 38.54
σ
q11 0.576*** 24.81
ζq6 0.426*** 9.04 ζq12 0.668*** 18.86
σ
q6 1.038*** 37.45
σ
q12 0.667*** 20.94
Note: ***, **, * indicate that the coefcients are signicant at the 1%, 5% and 10% levels, respectively.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
15
modern in both respects. For them, coal and rewood burning is not an
alternative in practice.
Higher educated and nancially upper classes are open to environ-
mental values in terms of the use of heating energy. Young adults not yet
raising a child and people over the age of fty with independent children
before retirement, belong to this social class. It indicates that strong
correlation can be observed between social capital supply, income level
and environmental awareness in Hungary. The “Neutral” class also has
social indicators similar to the environment class, and it shows that
strengthening their environmental awareness in the eld of heating
could be achieved through more active involvement and cooperation of
the public and civil spheres. The “Comfort and environmental concern”
group typically includes lower-middle-class city dwellers, where there is
no lack of openness to environmental values in the assessment of heat
energy consumption, but they are strongly inuenced by limited access
to livelihoods and social capital. In their case, mass social and material
rise can bring a signicant strengthening of environmental values in
energy use. For the classes of “Sensitive for stove price” and of “Sensitive
for energy price”, scarce material resources also strongly limit the pur-
suit of environmental values. Both classes includes higher age of classes,
the higher proportion of retirees with uncertain sources of income and
the higher proportion of villagers disadvantaged in terms of access to
information are unlikely to change their environmental values.
People living in poorer households typically live in lower quality,
more energy-intensive properties and have no nancial means to change
this. External support is therefore needed to alleviate energy poverty.
Energy poverty has a number of factors and impacts, including energy
efciency, public subsidy schemes or health risks. The main inuencing
factors are the price of energy, the type, condition, equipment and en-
ergy efciency of housing and household income [20]. In Hungary,
75–80% of energy-poor households live in a family house. Energy
poverty mainly affects low-income elderly people, unemployed people,
families with many children and single-parent families. In rural,
economically disadvantaged areas, family houses are on average larger
than condominiums, and energy poverty is particularly high in these
areas due to higher overheads. The housing stock in Hungary is pre-
dominantly outdated, i.e. residential buildings are typically of poor
energy efciency, which leads to high overheads, carbon dioxide and air
pollution [120]. A series of legislative measures could reduce the con-
sequences of energy poverty. The modernisation of heating systems in
dwellings should be strongly promoted, in particular the replacement of
solid fuel stoves, but this intervention mainly affects the poorest, needs
to be designed in a way that does not cause serious negative conse-
quences and therefore requires a large-scale programme of stove and
chimney replacement [20]. Low-income households typically have
lower energy consumption than better-off households, but at the same
time have less access to modern, energy-efcient and environmentally
friendly solutions. This further exacerbates differences in energy use
levels between different social groups and long-term, predictable,
interdependent and differentiated residential energy renovation pro-
grammes are needed.
The results of our study show that the role of environmental values in
heating energy use is highly dependent on the level of material and
social capital and inequalities in their access. This may be particularly
true in countries and regions of the European Union where the middle
class and the supply of material and social capital do not reach the level
of development centres. Reducing national and regional disparities has
shown a long-term process. In strengthening environmental values,
therefore, a radical change can be achieved in the distribution of EU
nancial aids and of state subsidies, which provides much more re-
sources and harmonizes the cooperation of the state, local government
and non-governmental organizations, especially for the younger
generations.
Our results in development of clusters may be useful for the estab-
lishment of a selective support policy. Meeting with the EU expectations
regarding GHG emission and ratio of RES is impossible without relevant
and selective support policy instead of the present system which is
available for near any person. Based on our results, two clearly identi-
able groups should be highlighted in the energy policy. Our recom-
mendations for the most rapid progress:
•Signicant investment subsidy for quality wood stoves and rewood
(to avoid heating with non-wood wastes, which are highly dangerous
for the environment), as well as energy saving investments (e.g.
insulation) for the 60 year +persons, since they typically live in
under-insulated houses, in order to save energy for the long run. It is
important to inform them via television, radio, or local newspapers,
since – according to Bai et al. [121] these are the primary source of
information for them. They can not afford to buy effective stoves
without any nancial help, but able and ready to use the cheapest
and less comfortable types of biomass, as well as eligible for social
rewood, so price support would not be so effective. The more
effective stoves in their homes can save a high external cost (by
substitution of coal and reducing the rewood consumption). Pri-
ority support for old people is in line with current national social
policy, too. It would be very important not just social, but also in
health aspect, since according to Hughes et al. [122] there are in-
dications related to the correlation of low temperatures and respi-
ratory health of the elderly people.
•Spreading of pellet burners is not a question of nancial matters,
since 29% of the population able and ready to pay for both comfort
and environmental protection. This group is well-educated, younger
and typically at higher level of energy-efciency retrots, so the
subsidy would not be efcient for them. It would be important to
establish inland background of pellet stove production and pellet
production with their spill-over effects. In this case the support
should aim to the establishment of productive assets for enterprises.
The effect of subsidization for these segments should improve eco-
nomic activity and technological development as well as – in
accordance with Yu et al. (2022) [37], Hu and Wang (2020) [39] and
Khan et al. (2021) [38] results – can decrease the GHG emissions.
High natural gas and electricity prices in these days result higher
substitution value of renewable energies, which trigger price hiking
also in the market of renewable equipments, in spite of the technical
development. However, not only the fossil energy sources, but the
energy saving investments can be taken competitors. According to
practical experiences, energy saving investments should be imple-
mented rst, followed by renewable investments.
It should be emphasized that protability of the RES technologies
strongly depends on the price uctuation of fossil energy sources
(especially of oil and gas prices), which are affected many times by
political decisions [123], general infrastructural progress [124], spatial
[41] and nancial [42] development and it makes uncertain the
viability, effectiveness and planning of renewable systems, too. The best
example is the years of 2021 and 2022, when extremely high price
hiking was observed in the market of both above-mentioned fossil en-
ergy sources. Another important issue is considering country-specic
factors in national/regional RES policy, since income level, energy
consumption, differential risks, human capital, research and develop-
ment has very divergent impact on the efcient use of RES, as it was
proved by Wang et al. in several previous articles [14,14,44,94].
Model estimation in hybrid choice context is a relatively rarely used
practice in latent class modelling. By incorporating environmental
awareness as a latent attitude into the class allocation equation, we had
opportunity to distinguish classes with different preferences for heating
systems so that the grouping is based on the level of environmental
awareness of the respondents. As a result, we were able to know what
characteristics respondents actually have environmental awareness and
whether they make consistent decisions in their choices. In addition to
the usual WTP (willingness to pay) calculation, we also used WTI
(willingness to invest) calculations, which also have signicant
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
16
information content and can provide further guidance.
Our paper has two important limitations. On the one hand radical
changes of energy prices in recent years (due to epidemics, Russian-
Ukrainian war) may also orientate the originally environmentally and
comfort-oriented consumers towards economic preferences. On the
other hand, the balanced operation of the world economy, the increasing
information and education level of the everyday people, the energy
effective innovations and the increased integration of environmental
and energy storage considerations into energy prices are likely to bring
about a positive change in environmental awareness in the long term.
We therefore believe that future research should pay particular attention
to comparing the impact of extreme and normal conditions on energy
market as well as to continuing the estimation of short- and long-term
energy trends and their driving forces in differential types of economies.
Funding sources
The research was nanced by the project no. NKFI-OTKA K 128 965.
This research was also supported by the 2019–2.1.13-T´
ET_IN-2020-
00061 “Waste algae to biogas for clean energy and environment: techno-
environ-economic prospects” projectTKP2020-IKA-04 project, which
has been implemented with the support provided from the National
Research, Development and Innovation Fund of Hungary, nanced
under the 2020–4.1.1-TKP2020 funding scheme. The research was
nanced by the Higher Education Institutional Excellence Programme of
the Ministry of Innovation and Technology in Hungary, within the
framework of the Energy thematic programme of the University of
Debrecen. Boldizs´
ar Megyesi received Bolyai J´
anos Postdoctoral Schol-
arship of the Hungarian Academy of Sciences and a New National
Excellence Program ÚNKP 2021-5 Stipendium during the works on the
article.
Credit author statement
Attila Bai; Conceptualization, Methodology, Formal analysis,
Writing - Original Draft, Writing - Review & Editing. Imre Kov´
ach;
Conceptualization, Methodology, Writing - Original Draft, Writing -
Review & Editing, Funding acquisition. P´
eter Balogh; Conceptualiza-
tion, Methodology, Formal analysis, Writing - Original Draft, Writing -
Review & Editing. Ibolya Czibere; Conceptualization, Methodology,
Writing - Original Draft, Writing - Review & Editing. Boldizs´
ar Megyesi;
Conceptualization, Methodology, Writing - Original Draft, Writing -
Review & Editing.
Declaration of competing interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Data availability
Data will be made available on request.
References
[1] B.K. Sovacool, J. Axsen, S. Sorrell, Promoting novelty, rigor, and style in energy
social science: towards codes of practice for appropriate methods and research
design, Energy Res. Social Sci. 45 (2018) 12–42, https://doi.org/10.1016/j.
erss.2018.07.007.
[2] Z. Szak´
aly, P. Balogh, E. Kontor, Z. Gabnai, A. Bai, Attitude toward and awareness
of renewable energy sources: Hungarian experience and special features, Energies
14 (2021) 22, https://doi.org/10.3390/en14010022.
[3] I. Czibere, I. Kov´
ach, G.B. Megyesi, Environmental citizenship and energy
efciency in four European countries (Italy, The Netherlands, Switzerland and
Hungary), Sustainability 12 (2020) 1154, https://doi.org/10.3390/su12031154.
[4] Living conditions in Europe - housing - Statistics Explained, (n.d.). https://ec.
europa.eu/eurostat/statistics-explained/index.php?title=Living_conditions_in_
Europe_-_housing (accessed January 26, 2022).
[5] E. Ruokamo, Household preferences of hybrid home heating systems – a choice
experiment application, Energy Pol. 95 (2016) 224–237.
[6] M. Achtnicht, Do environmental benets matter? Evidence from a choice
experiment among house owners in Germany, Ecol. Econ. 70 (2011) 2191–2200.
[7] Global Energy Review 2021, 2021.
[8] Q. Wang, L. Wang, Renewable energy consumption and economic growth in
OECD countries: a nonlinear panel data analysis, Energy 207 (2020), 118200,
https://doi.org/10.1016/j.energy.2020.118200.
[9] KSH (HCSO), V´
egs˝
o energiafelhaszn´
al´
as szektoronk´
ent (1990–2016) (Final
energy consumption (1990-2016), (n.d.). https://www.ksh.hu/docs/hun/eurost
at_tablak/tabl/tsdpc320.html (accessed April 16, 2021)..
[10] Eurostat 2., household expenditure in 2018. https://ec.europa.eu/eurostat/we
b/products-eurostat-news/-/DDN-20191127-1, 2018. (Accessed 16 April 2021).
[11] Eurostat 1, Energy Consumption in Households - Statistics Explained, 2021. htt
ps://ec.europa.eu/eurostat/statistics-explained/index.php?title=Energy_consum
ption_in_households. (Accessed 16 April 2021).
[12] KSH (HCSO), A. h´
aztart´
asok fogyaszt´
asa, I. F´
el´
ev (El˝
ozetes Adatok Alapj´
an)
HCSO. Statistical Mirror. Consumption of Households, Hungarian Central
Statistics Ofce: Budapest, Hungary, 2018, 2018, p. 3, 2018.
[13] C. Franceschinis, M. Thiene, R. Scarpa, J. Rose, M. Moretto, R. Cavalli, Adoption
of renewable heating systems : an empirical test of the diffusion of innovation
theory, Energy 125 (2017) 313–326, https://doi.org/10.1016/j.
energy.2017.02.060.
[14] Q. Wang, Z. Dong, R. Li, L. Wang, Renewable energy and economic growth: new
insight from country risks, Energy 238 (2022), 122018, https://doi.org/10.1016/
j.energy.2021.122018.
[15] J. Brodny, M. Tutak, Analysis of the efciency and structure of energy
consumption in the industrial sector in the European Union countries between
1995 and 2019, Sci. Total Environ. 808 (2022), 152052, https://doi.org/
10.1016/j.scitotenv.2021.152052.
[16] Y. Sun, Z. Han, Climate change risk perception in taiwan: correlation with
individual and societal factors, Int. J. Environ. Res. Publ. Health 15 (2018) 91,
https://doi.org/10.3390/ijerph15010091.
[17] R.R. Plate, M.C. Monroe, A. Oxarart, Public perceptions of using woody biomass
as a renewable energy source, J. Ext. 48 (2010). https://www.cabdirect.org/ca
bdirect/abstract/20103218687. April 15, 2021.
[18] S. Wehrle, K. Gruber, J. Schmidt, The cost of undisturbed landscapes, Energy Pol.
159 (2021), 112617, https://doi.org/10.1016/j.enpol.2021.112617.
[19] V. Bene, I. Bihari, I. Czibere, I. Kov´
ach, B. Megyesi, V. Pacz´
ari, B. Pataki, Factors
inuencing households’ energy consumption in Hungary, in: Case-study
Conducted in the City of Debrecen, Debrecen University Press, Debrecen, 2019.
[20] N. Feldm´
ar, Energiaszeg´
enys´
eg, in: Gosztonyi ´
Akos Szerk ´
Eves Jel. Lakhat´
asi
Szeg´
enys´
egr˝
ol 2020, Habitat for Humanity Magyarorsz´
ag, Budapest, 2020,
pp. 42–56. https://habitat.hu/sites/lakhatasi-jelentes-2020/wp-content/uploa
ds/sites/9/2020/10/hfhh_lakhatasi_jelentes_2020.pdf.
[21] Air emissions accounts by NACE Rev. 2 activity (env_ac_ainah_r2), (n.d.). htt
ps://ec.europa.eu/eurostat/cache/metadata/en/env_ac_ainah_r2_esms.htm
(accessed December 12, 2022).
[22] Z. Marjain´
e Szer´
enyi, S. Kerekes, Z. Flachner, S. Milton, The possibility of the
economic evaluation of ecosystem services described through a domestic case
study, in: G.G. Nagy, V. Kiss (Eds.), Borrow. Serv. Nat. Methodol. Eval. Ecosyst.
Serv. Focus. Hung. Case Stud., CEEweb for Biodiversity, Budapest, 2011,
pp. 64–75. http://unipub.lib.uni-corvinus.hu/1188/. (Accessed 15 April 2021).
[23] T. Kocsis, Z. Marjain´
e Szer´
enyi, P´
enz´
aldozat vagy id˝
o´
aldozat? K¨
olts´
eg´
ath´
arít´
as ´
es
az ´
ert´
ekel´
es dilemm´
ai a term´
eszeti k¨
ornyezetet ´
erint˝
o k´
erd´
esekben, Magy.
Tudomany (2018), https://doi.org/10.1556/2065.179.2018.2.4.
[24] D. Whittington, What have we learned from 20 Years of stated preference
research in less-developed countries? Annu. Rev. Resour. Econ. 2 (2010)
209–236, https://doi.org/10.1146/annurev.resource.012809.103908.
[25] M. Ahlheim, O. Fr¨
or, A. Heinke, N.M. Duc, P.V. Dinh, Labour as a Utility Measure
in Contingent Valuation Studies: How Good Is it Really? University of
Hohenheim, Center for Research on Innovation and Services (FZID), 2010.
https://ideas.repec.org/p/zbw/fziddp/132010.html. April 15, 2021.
[26] R. Rai, L.D. Bhatta, M. Nepal, P. Shyamsunder, Is willingness-to-pay sensitive to
the mode of payment? A choice experiment in the koshi basin of Nepal, in: Fifth
World Congress of Environmental and Resource Economists. June - 2 July 2014,
Istanbul, Turkey | HimalDoc, 2014. https://lib.icimod.org/record/30161. April
15, 2021.
[27] R.K. Rai, H. Scarborough, Nonmarket valuation in developing countries:
incorporating labour contributions in environmental benets estimates, Aust. J.
Agric. Resour. Econ. 59 (2015) 479–498, https://doi.org/10.1111/1467-
8489.12071.
[28] M. Tilahun, L. Vranken, B. Muys, J. Deckers, K. Gebregziabher, K. Gebrehiwot,
H. Bauer, E. Mathijs, Rural households’ demand for frankincense forest
conservation in tigray, Ethiopia: a contingent valuation analysis, Land Degrad.
Dev. 26 (2015) 642–653, https://doi.org/10.1002/ldr.2207.
[29] T. Lankia, M. Neuvonen, E. Pouta, T. Siev¨
anen, Willingness to contribute to the
management of recreational quality on private lands in Finland, J. For. Econ. 20
(2014) 141–160, https://doi.org/10.1016/j.jfe.2014.04.001.
[30] Y.-S. Eom, D.M. Larson, Valuing housework time from willingness to spend time
and money for environmental quality improvements, Rev. Econ. Househ. 4
(2006) 205–227, https://doi.org/10.1007/s11150-006-0008-1.
[31] I. Kastner, E. Matthies, Investments in renewable energies by German households:
a matter of economics, social inuences and ecological concern? Energy Res.
Social Sci. 17 (2016) 1–9, https://doi.org/10.1016/j.erss.2016.03.006.
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
17
[32] C.C. Michelsen, R. Madlener, Homeowners’ preferences for adopting innovative
residential heating systems: a discrete choice analysis for Germany, Energy Econ.
34 (2012) 1271–1283, https://doi.org/10.1016/j.eneco.2012.06.009.
[33] S. Rouvinen, J. Matero, Stated preferences of Finnish private homeowners for
residential heating systems: a discrete choice experiment, Biomass Bioenergy 57
(2013) 22–32, https://doi.org/10.1016/j.biombioe.2012.10.010.
[34] B. Mills, J. Schleich, Residential energy-efcient technology adoption, energy
conservation, knowledge, and attitudes: an analysis of European countries,
Energy Pol. 49 (2012) 616–628, https://doi.org/10.1016/j.enpol.2012.07.008.
[35] P. Mariel, D. Hoyos, J. Meyerhoff, M. Czajkowski, T. Dekker, K. Glenk, J.
B. Jacobsen, U. Liebe, S.B. Olsen, J. Sagebiel, M. Thiene, Environmental
Valuation with Discrete Choice Experiments: Guidance on Design,
Implementation and Data Analysis, Springer International Publishing, 2021,
https://doi.org/10.1007/978-3-030-62669-3.
[36] M. Achtnicht, R. Madlener, Factors inuencing German house owners’
preferences on energy retrots, Energy Pol. 68 (2014) 254–263.
[37] Z. Yu, S.A.R. Khan, P. Ponce, A.B. Lopes de Sousa Jabbour, C.J. Chiappetta
Jabbour, Factors affecting carbon emissions in emerging economies in the context
of a green recovery: implications for sustainable development goals, Technol.
Forecast. Soc. Change 176 (2022), 121417, https://doi.org/10.1016/j.
techfore.2021.121417.
[38] S.A.R. Khan, P. Ponce, Z. Yu, Technological innovation and environmental taxes
toward a carbon-free economy: an empirical study in the context of COP-21,
J. Environ. Manag. 298 (2021), 113418, https://doi.org/10.1016/j.
jenvman.2021.113418.
[39] W. Hu, D. Wang, How does environmental regulation inuence China’s carbon
productivity? An empirical analysis based on the spatial spillover effect, J. Clean.
Prod. 257 (2020), 120484, https://doi.org/10.1016/j.jclepro.2020.120484.
[40] H. Wang, H. Cui, Q. Zhao, Effect of green technology innovation on green total
factor productivity in China: evidence from spatial durbin model analysis,
J. Clean. Prod. 288 (2021), 125624, https://doi.org/10.1016/j.
jclepro.2020.125624.
[41] S. Yin, Z. Gong, L. Gu, Y. Deng, Y. Niu, Driving forces of the efciency of forest
carbon sequestration production: spatial panel data from the national forest
inventory in China, J. Clean. Prod. 330 (2022), 129776, https://doi.org/
10.1016/j.jclepro.2021.129776.
[42] P. Ponce, J. ´
Alvarez-García, J. Medina, M. de la C. del Río-Rama, Financial
development, clean energy, and human capital: roadmap towards sustainable
growth in am´
erica latina, Energies 14 (2021) 3763, https://doi.org/10.3390/
en14133763.
[43] G. Peng, F. Meng, Z. Ahmed, J. Ol´
ah, E. Hars´
anyi, A path towards green
revolution: how do environmental technologies, political risk, and environmental
taxes inuence green energy consumption? Front. Environ. Sci. 10 (2022). htt
ps://www.frontiersin.org/articles/10.3389/fenvs.2022.927333. (Accessed 30
August 2023).
[44] Q. Wang, F. Zhang, R. Li, Revisiting the environmental kuznets curve hypothesis
in 208 counties: the roles of trade openness, human capital, renewable energy and
natural resource rent, Environ. Res. 216 (2023), 114637, https://doi.org/
10.1016/j.envres.2022.114637.
[45] O.D.M. Tuegeh, M. Harangi-R´
akos, A.S. Nagy, Industry 4.0 and human resource
in Indonesia: a systematic literature review, Econ. Ann.-ХХI. 190 (2021)
171–180, https://doi.org/10.21003/ea.V190-16.
[46] A.S. Nagy, B. Bittner, O.D.M. Tuegh, J.R. Tumiwa, AUGMENTED REALITY
IMPROVING CONSUMER CHOICE CONFIDENCE DURING COVID-19, Issues Inf.
Syst., 2022, https://doi.org/10.48009/2_iis_2022_126.
[47] J. Popp, M. Harangi-R´
akos, K. Pet˝
o, A. Nagy, Bioenergy: risks to food-, energy-
and environmental security, Appl. Stud. Agribus. Commer. 7 (2013) 121–130,
https://doi.org/10.19041/APSTRACT/2013/4-5/17.
[48] V. Vida, I. Sz˝
ucs, Pork production and consumption issues from the perspective of
the religion and the World’s growing population, Appl. Stud. Agribus. Commer.
14 (2020) 121–128, https://doi.org/10.19041/APSTRACT/2020/1-2/16.
[49] M. Ringel, Fostering the use of renewable energies in the European Union: the
race between feed-in tariffs and green certicates, Renew. Energy 31 (2006)
1–17, https://doi.org/10.1016/j.renene.2005.03.015.
[50] D. Ürge-Vorsatz, G. Miladinova, L. Paizs, Energy in transition: from the iron
curtain to the European Union, Energy Pol. 34 (2006) 2279–2297, https://doi.
org/10.1016/j.enpol.2005.03.007.
[51] W. Abrahamse, L. Steg, C. Vlek, T. Rothengatter, The effect of tailored
information, goal setting, and tailored feedback on household energy use, energy-
related behaviors, and behavioral antecedents, J. Environ. Psychol. 27 (2007)
265–276, https://doi.org/10.1016/j.jenvp.2007.08.002.
[52] S.Z. Attari, M. Schoen, C.I. Davidson, M.L. DeKay, W. Bruine de Bruin, R. Dawes,
M.J. Small, Preferences for change: do individuals prefer voluntary actions, soft
regulations, or hard regulations to decrease fossil fuel consumption? Ecol. Econ.
68 (2009) 1701–1710, https://doi.org/10.1016/j.ecolecon.2008.10.007.
[53] F. Belaïd, Understanding the spectrum of domestic energy consumption:
empirical evidence from France, Energy Pol. 92 (2016) 220–233, https://doi.org/
10.1016/j.enpol.2016.02.015.
[54] W. Abrahamse, L. Steg, Social inuence approaches to encourage resource
conservation: a meta-analysis, Global Environ. Change 23 (2013) 1773–1785,
https://doi.org/10.1016/j.gloenvcha.2013.07.029.
[55] H. Allcott, R.L. Sweeney, The role of sales agents in information disclosure:
evidence from a eld experiment, Manag. Sci. 63 (2017) 21–39, https://doi.org/
10.1287/mnsc.2015.2327.
[56] K. Gram-Hanssen, Households’ Energy Use – Which Is the More Important:
Efcient Technologies or User Practices?, 2011. http://www.ep.liu.se/ecp/057/
vol3/032/ecp57vol3_032.pdf. March 10, 2017.
[57] S. Bouzarovski, S. Petrova, R. Sarlamanov, Energy poverty policies in the EU: a
critical perspective, Energy Pol. 49 (2012) 76–82, https://doi.org/10.1016/j.
enpol.2012.01.033.
[58] C. Morton, C. Wilson, J. Anable, The diffusion of domestic energy efciency
policies: a spatial perspective, Energy Pol. 114 (2018) 77–88, https://doi.org/
10.1016/j.enpol.2017.11.057.
[59] C. Cattaneo, Internal and external barriers to energy efciency: which role for
policy interventions? Energy Efc 12 (2019) 1293–1311, https://doi.org/
10.1007/s12053-019-09775-1.
[60] J. Schleich, C. Faure, M.-C. Guetlein, G. Tu, Household Preferences for New
Heating Systems: Insights from a Multi-Country Discrete Choice Experiment,
Fraunhofer Institute for Systems and Innovation Research (ISI), 2020. https://id
eas.repec.org/p/zbw/sisi/s052020.html. April 15, 2021.
[61] A. Jaffe, R. Stavins, The energy-efciency gap: what does it mean? Energy Pol. 22
(1994) 804–810, https://doi.org/10.1016/0301-4215(94)90138-4.
[62] J. Schleich, X. Gassmann, C. Faure, T. Meissner, Making the implicit explicit: a
look inside the implicit discount rate, Energy Pol. 97 (2016) 321–331, https://
doi.org/10.1016/j.enpol.2016.07.044.
[63] T.D. Gerarden, R.G. Newell, R.N. Stavins, Assessing the energy-efciency gap,
J. Econ. Lit. 55 (2017) 1486–1525.
[64] L.W. Davis, G.E. Metcalf, Does Better Information Lead to Better Choices?
Evidence from Energy-Efciency Labels, National Bureau of Economic Research,
Inc, 2014. https://ideas.repec.org/p/nbr/nberwo/20720.html. (Accessed 28
December 2019).
[65] S. Houde, How consumers respond to product certication and the value of
energy information, Rand J. Econ. 49 (2018) 453–477, https://doi.org/10.1111/
1756-2171.12231.
[66] N. Hamza, R. Gilroy, The challenge to UK energy policy: an ageing population
perspective on energy saving measures and consumption, Energy Pol. 39 (2011)
782–789, https://doi.org/10.1016/j.enpol.2010.10.052.
[67] A. Ramos, X. Labandeira, A. L¨
oschel, Pro-environmental households and energy
efciency in Spain, Environ. Resour. Econ. 63 (2016) 367–393, https://doi.org/
10.1007/s10640-015-9899-8.
[68] E. L´
opez-Bernab´
e, S. Foudi, I. Galarraga, Mind the map? Mapping the academic,
citizen and professional stakeholder views on buildings and heating behaviour in
Spain, Energy Res. Social Sci. 69 (2020), 101587, https://doi.org/10.1016/j.
erss.2020.101587.
[69] K. Willis, R. Scarpa, R. Gilroy, N. Hamza, Renewable energy adoption in an
ageing population: heterogeneity in preferences for micro-generation technology
adoption, Energy Pol. 39 (2011) 6021–6029, https://doi.org/10.1016/j.
enpol.2011.06.066.
[70] M. Poniatowska-Jaksch, Energy consumption in central and eastern europe (CEE)
households in the platform economics, Energies 14 (2021) 1002, https://doi.org/
10.3390/en14041002.
[71] F. Carlsson, P. Frykblom, C. Johan Lagerkvist, Using cheap talk as a test of
validity in choice experiments, Econ. Lett. 89 (2005) 147–152, https://doi.org/
10.1016/j.econlet.2005.03.010.
[72] S. Klain, T. Sattereld, K.M.A. Chan, K. Lindberg, Octopus’s garden under the
blade: boosting biodiversity increases willingness to pay for offshore wind in the
United States, Energy Res. Social Sci. 69 (2020), 101744, https://doi.org/
10.1016/j.erss.2020.101744.
[73] J. Loomis, WHAT’S to KNOW about hypothetical bias in stated preference
valuation studies?: hypothetical bias in stated preference valuation studies,
J. Econ. Surv. 25 (2011) 363–370, https://doi.org/10.1111/j.1467-
6419.2010.00675.x.
[74] J. Ladenburg, S. Olsen, Augmenting Short Cheap Talk Scripts with a Repeated
Opt-Out Reminder in Choice Experiment Surveys, University of Copenhagen,
Department of Food and Resource Economics, 2010. https://econpapers.repec.
org/paper/foiwpaper/2010_5f09.htm. (Accessed 26 January 2022).
[75] J.J. Louviere, D.A. Hensher, J.D. Swait, Stated Choice Methods: Analysis and
Applications, Cambridge University Press, Cambridge, UK; New York, NY, USA,
2000. http://site.ebrary.com/lib/academiccompletetitles/home.action. April 15,
2021.
[76] K.J. Lancaster, A new approach to consumer theory, J. Polit. Econ. 74 (1966)
132–157. https://www.jstor.org/stable/1828835.
[77] K.E. Train, Discrete Choice Methods with Simulation, second ed., Cambridge
University Press, 2001 https://doi.org/10.1017/CBO9780511805271.
[78] D.A. Hensher, J.M. Rose, W.H. Greene, Inferring attribute non-attendance from
stated choice data: implications for willingness to pay estimates and a warning for
stated choice experiment design, Transportation 39 (2012) 235–245, https://doi.
org/10.1007/s11116-011-9347-8.
[79] N. Saxena, T.H. Rashidi, V.V. Dixit, S.T. Waller, Modelling the route choice
behaviour under stop-&-go trafc for different car driver segments, Transport.
Res. Part Policy Pract 119 (2019) 62–72, https://doi.org/10.1016/j.
tra.2018.11.004.
[80] O. Mandrik, A. Yaumenenka, R. Herrero, M.F. Jonker, Population preferences for
breast cancer screening policies: discrete choice experiment in Belarus, PLoS One
14 (2019), e0224667, https://doi.org/10.1371/journal.pone.0224667.
[81] A. Radley, M. van der Pol, J.F. Dillon, Application of a discrete choice experiment
approach to support the design of a hepatitis C testing service in primary care, Int.
J. Drug Pol. 65 (2019) 1–7, https://doi.org/10.1016/j.drugpo.2018.12.008.
[82] Z. Kallas, E. Varela, M. ˇ
Candek-Potokar, C. Pugliese, M. Cerjak, U. Tomaˇ
zin,
D. Karolyi, C. Aquilani, M. Vitale, J.M. Gil, Can innovations in traditional pork
A. Bai et al.
Energy Strategy Reviews 50 (2023) 101192
18
products help thriving EU untapped pig breeds? A non-hypothetical discrete
choice experiment with hedonic evaluation, Meat Sci. 154 (2019) 75–85, https://
doi.org/10.1016/j.meatsci.2019.04.011.
[83] J. Wang, J. Ge, Y. Ma, Urban Chinese consumers’ willingness to pay for pork with
certied labels: a discrete choice experiment, Sustainability 10 (2018) 603,
https://doi.org/10.3390/su10030603.
[84] F.J. van Rijnsoever, A. van Mossel, K.P.F. Broecks, Public acceptance of energy
technologies: the effects of labeling, time, and heterogeneity in a discrete choice
experiment, Renew. Sustain. Energy Rev. 45 (2015) 817–829, https://doi.org/
10.1016/j.rser.2015.02.040.
[85] X. Zhao, Q. Cai, S. Li, C. Ma, Public preferences for biomass electricity in China,
Renew. Sustain. Energy Rev. 95 (2018) 242–253.
[86] D. McFadden, Conditional logit analysis of qualitative choice Be, in: P. Zarembka
(Ed.), Front. Econom. Econ. Theory Math. Econ., Academic Press, New York,
1973, pp. 105–142. https://eml.berkeley.edu/reprints/mcfadden/zarembka.pdf.
[87] M. Ben-Akiva, D. Mcfadden, K. Train, J. Walker, C. Bhat, M. Bierlaire, D. Bolduc,
A. Boersch-Supan, D. Brownstone, D.S. Bunch, A. Daly, A. De Palma, D. Gopinath,
A. Karlstrom, M.A. Munizaga, Hybrid choice models: progress and challenges,
Market. Lett. 13 (2002) 163–175, https://doi.org/10.1023/A:1020254301302.
[88] A. Bujosa, A. Riera, R.L. Hicks, Combining discrete and continuous
representations of preference heterogeneity: a latent class approach, Environ.
Resour. Econ. 47 (2010) 477–493, https://doi.org/10.1007/s10640-010-9389-y.
[89] M. Ben-Akiva, D. McFadden, T. G¨
arling, D. Gopinath, J. Walker, D. Bolduc,
A. B¨
orsch-Supan, P. Delqui´
e, O. Larichev, T. Morikawa, A. Polydoropoulou,
V. Rao, Extended framework for modeling choice behavior, Market. Lett. 10
(1999) 187–203, https://doi.org/10.1023/A:1008046730291.
[90] R. Hurtubia, M.H. Nguyen, A. Glerum, M. Bierlaire, Integrating psychometric
indicators in latent class choice models, Transport. Res. Part Policy Pract. 64
(2014) 135–146, https://doi.org/10.1016/j.tra.2014.03.010.
[91] S. Etzioni, R.A. Daziano, E. Ben-Elia, Y. Shiftan, Preferences for shared automated
vehicles: a hybrid latent class modeling approach, Transport. Res. C Emerg.
Technol. 125 (2021), 103013, https://doi.org/10.1016/j.trc.2021.103013.
[92] U. Liebe, P. Mariel, H. Beyer, J. Meyerhoff, Uncovering the nexus between
attitudes, preferences, and behavior in sociological applications of stated choice
experiments, Socio. Methods Res. 50 (2021) 310–347, https://doi.org/10.1177/
0049124118782536.
[93] P. Mariel, D. Hoyos, A. Artabe, C.A. Guevara, A multiple indicator solution
approach to endogeneity in discrete-choice models for environmental valuation,
Sci. Total Environ. 633 (2018) 967–980, https://doi.org/10.1016/j.
scitotenv.2018.03.254.
[94] B. Mao, C. Ao, J. Wang, B. Sun, L. Xu, Does regret matter in public choices for air
quality improvement policies? A comparison of regret-based and utility-based
discrete choice modelling, J. Clean. Prod. 254 (2020), 120052, https://doi.org/
10.1016/j.jclepro.2020.120052.
[95] A. Motz, Security of supply and the energy transition: the households’ perspective
investigated through a discrete choice model with latent classes, Energy Econ. 97
(2021), 105179, https://doi.org/10.1016/j.eneco.2021.105179.
[96] E. Strazzera, R. Atzori, D. Meleddu, V. Statzu, Assessment of renaturation
measures for improvements in ecosystem services and ood risk mitigation,
J. Environ. Manag. 292 (2021), 112743, https://doi.org/10.1016/j.
jenvman.2021.112743.
[97] E. Owusu-Sekyere, Y. Mahlathi, H. Jordaan, Understanding South African
consumers’ preferences and market potential for products with low water and
carbon footprints, Agrekon 58 (2019) 354–368, https://doi.org/10.1080/
03031853.2019.1589544.
[98] A. Motz, Consumer acceptance of the energy transition in Switzerland: the role of
attitudes explained through a hybrid discrete choice model, Energy Pol. 151
(2021), 112152, https://doi.org/10.1016/j.enpol.2021.112152.
[99] E. Owusu-Sekyere, A. Abdulai, H. Jordaan, H. Hansson, Heterogeneous demand
for ecologically sustainable products on ensuring environmental sustainability in
South Africa, Environ. Econ. Pol. Stud. 22 (2020) 39–64, https://doi.org/
10.1007/s10018-019-00246-6.
[100] B. Salak, K. Lindberg, F. Kienast, M. Hunziker, How landscape-technology t
affects public evaluations of renewable energy infrastructure scenarios. A hybrid
choice model, Renew. Sustain. Energy Rev. 143 (2021), 110896, https://doi.org/
10.1016/j.rser.2021.110896.
[101] D. Bolduc, M. Ben-Akiva, J. Walker, A. Michaud, Hybrid choice models with logit
kernel: applicability to large scale Models1, in: M.E.H. Lee-Gosselin, S.T. Doherty
(Eds.), Integr. Land-Use Transp. Models, Emerald Group Publishing Limited,
2005, pp. 275–302, https://doi.org/10.1108/9781786359520-012.
[102] D. Bolduc, N. Boucher, R. Alvarez-Daziano, Hybrid choice modeling of new
technologies for car choice in Canada, transp, Res. Rec. J. Transp. Res. Board.
2082 (2008) 63–71, https://doi.org/10.3141/2082-08.
[103] R.A. Daziano, D. Bolduc, Incorporating pro-environmental preferences towards
green automobile technologies through a Bayesian hybrid choice model, Transp.
Transp. Sci. 9 (2013) 74–106, https://doi.org/10.1080/18128602.2010.524173.
[104] A. Daly, S. Hess, B. Patruni, D. Potoglou, C. Rohr, Using ordered attitudinal
indicators in a latent variable choice model: a study of the impact of security on
rail travel behaviour, Transportation 39 (2012) 267–297, https://doi.org/
10.1007/s11116-011-9351-z.
[105] I. Sarman, S. Scagnolari, R. Maggi, Acceptance of life-threatening hazards among
young tourists: a stated choice experiment, J. Trav. Res. 55 (2016) 979–992,
https://doi.org/10.1177/0047287515612595.
[106] Ngene User Manual, 2018, p. 241.
[107] S. Hess, D. Palma, Apollo: a Flexible, Powerful and Customisable Freeware
Package for Choice Model Estimation and Application Version 0.2.4 User Manual,
2021. www.ApolloChoiceModelling.com. http://www.apollochoicemodelling.
com/les/Apollo.pdf.
[108] S. Hess, D. Palma, Apollo: a exible, powerful and customisable freeware package
for choice model estimation and application, J. Choice Model. 32 (2019), 100170,
https://doi.org/10.1016/j.jocm.2019.100170.
[109] P. Mariel, J. Meyerhoff, S. Hess, Heterogeneous preferences toward landscape
externalities of wind turbines – combining choices and attitudes in a hybrid
model, Renew. Sustain. Energy Rev. 41 (2015) 647–657, https://doi.org/
10.1016/j.rser.2014.08.074.
[110] S. Yilmaz, P. Cuony, C. Chanez, Prioritize your heat pump or electric vehicle?
Analysing design preferences for Direct Load Control programmes in Swiss
households, Energy Res. Social Sci. 82 (2021), 102319, https://doi.org/10.1016/
j.erss.2021.102319.
[111] P.C. Boxall, W.L. Adamowicz, Understanding heterogeneous preferences in
random utility models: a latent class approach, Environ. Resour. Econ. 23 (2002)
421–446, https://doi.org/10.1023/A:1021351721619.
[112] W.H. Greene, D.A. Hensher, A latent class model for discrete choice analysis:
contrasts with mixed logit, Transp. Res. Part B Methodol. 37 (2003) 681–698,
https://doi.org/10.1016/S0191-2615(02)00046-2.
[113] J.E. Cavanaugh, A.A. Neath, The Akaike information criterion: background,
derivation, properties, application, interpretation, and renements, WIREs
Comput. Stat. 11 (2019), e1460, https://doi.org/10.1002/wics.1460.
[114] D. McFadden, The choice theory approach to market research, Mark. Sci. 5 (1986)
275–297, https://doi.org/10.1287/mksc.5.4.275.
[115] N. Boogen, C. Daminato, Large Sample Survey - D1.3, 2019. http://www.penny-p
roject.eu/wp-content/uploads/2017/05/PENNY_D1.3_updated_nal.pdf.
[116] E. van der Werff, L. Steg, K. Keizer, The value of environmental self-identity: the
relationship between biospheric values, environmental self-identity and
environmental preferences, intentions and behaviour, J. Environ. Psychol. 34
(2013) 55–63, https://doi.org/10.1016/j.jenvp.2012.12.006.
[117] S.H. Schwartz, J.A. Howard, A normative decision-making model of altruism, in:
J.P. Rushton, R.M. Sorrentino (Eds.), Altruism Help. Behav., Erlbaum., Hillsdale,
1981, pp. 89–211.
[118] A.M. Ruepert, K. Keizer, L. Steg, The relationship between Corporate
Environmental Responsibility, employees’ biospheric values and pro-
environmental behaviour at work, J. Environ. Psychol. 54 (2017) 65–78, https://
doi.org/10.1016/j.jenvp.2017.10.006.
[119] H. Steinmetz, M. Knappstein, I. Ajzen, P. Schmidt, R. Kabst, How effective are
behavior change interventions based on the theory of planned behavior?: a three-
level meta-analysis, Z. Psychol. 224 (2016) 216–233, https://doi.org/10.1027/
2151-2604/a000255.
[120] O. Fül¨
op, A. Lehoczki-Krsjak, Energiaszeg´
enys´
eg Magyarorsz´
agon, Statisztikai
Szle. 92 (2014) 820–831.
[121] A. Bai, E. Durk´
o, K. Tar, J.B. T´
oth, I. L´
az´
ar, L. Kapocska, A. Kircsi, B. Bart´
ok,
R. Vass, J. P´
enzes, T. T´
oth, Social and economic possibilities for the energy
utilization of tomass in the valley of the river Hern´
ad, Renew. Energy 85 (2016)
777–789, https://doi.org/10.1016/j.renene.2015.06.069.
[122] C. Hughes, S. Natarajan, C. Liu, W.J. Chung, M. Herrera, Winter thermal comfort
and health in the elderly, Energy Pol. 134 (2019), 110954, https://doi.org/
10.1016/j.enpol.2019.110954.
[123] J. P´
apay, Konvencion´
alis ´
es nem konvencion´
alis k˝
oolaj-´
es f¨
oldg´
azkitermel˝
o
elj´
ar´
asok, Magy. Tudomany 176 (2015) 1285–1294. ´
evfolyam 2015/11. pp.
1285–1294., Magy. Tud. 176.
[124] H. Xu, S. Cao, X. Xu, The development of highway infrastructure and CO2
emissions: the mediating role of agglomeration, J. Clean. Prod. 337 (2022),
130501, https://doi.org/10.1016/j.jclepro.2022.130501.
A. Bai et al.