Developing a Typology Based on Energy Practices and Environmental Attitudes

Abstract

The European Union is increasingly directing efforts and resources toward promoting the adoption of energy saving measures among citizens, but the effectiveness of these efforts cannot be ensured if the heterogeneity of citizens is neglected. This paper assesses whether Greek citizens present heterogeneity in terms of their energy behavior and investigates whether they can be classified based on the practices they apply to reduce energy costs and their environmental attitudes. To that end, the study collected a representative citizen sample, and k-means cluster analysis was performed. Results indicated significant heterogeneity and four clusters were identified: ‘Extreme heat savers and pro-environmental food consumers’ saved on heating but neglected practices regarding lighting, hot water use, and transport; ‘Heat savers and environmentally unaware’ applied energy saving measures only to lower heating costs; ‘Environmentally aware energy savers’ were the most engaged segment in the application of energy saving measures driven by pro-environmental attitudes; and ‘Mindful resources and transport users’ did not apply any measures for heating despite having positive attitudes in terms of electricity, water and transport use. Therefore, it is necessary to train citizens on pro-environmental practices, and in doing so, the recorded heterogeneity can guide the design of differentiated and effective approaches.

1. Introduction

In an effort to reduce residential energy consumption, the European Union has established the Building Performance Directive (2010/31/EU), the Ecodesign and Energy Labelling Directive (2010/30/EU), and the Energy Efficiency Directive (2012/27/EU), acknowledging not only the need for households to adopt energy saving practices but also the role of citizens as responsible energy consumers. The latter directive consists of the EU’s main policy tool to promote energy efficiency and was recently revised with the establishment of even more ambitious energy efficiency targets as part of the Clean Energy for all European Package and also as a means to conform to the union’s commitment to the Global Pledge [1]. In particular, the revised directive sets new rules for energy efficiency and involves a loftier objective of a minimum of 32.5% to be achieved by the year 2030, which is significantly higher than the previous target of 20%. It is now required from all member states to save, on average, 4.4% in annual energy consumption until the year 2030. Moreover, the total energy consumption across the EU should be less than 992.5 Mtoe for primary energy and less than 763 Mtoe for final energy. It is also important to note that the revised directive promotes the principle of ‘energy efficiency first’ as the guideline for the design of the energy policy that should be followed by all member states in energy and investment decision-making [1].

While technological progress and regulatory policies can reduce the consumption of industrial energy, building energy consumption is still a major challenge [2,3]. On a global level, the building sector is a major energy consumer, accounting for over 30% of total energy, while a large proportion of this energy is consumed for heating purposes in residential buildings [4]. Meanwhile, in terms of greenhouse gas emissions, the building sector accounts for 38% of global greenhouse gas emissions [5]. In the European Union, household heating accounts for 63.6% of total energy, resulting in around 26% of carbon dioxide emissions [6]. These figures not only show that more attention should be paid to the building sector but also bring forward the need to direct more research on understanding occupant behavior [7]. In other words, as occupant behavior is what determines how energy is used in buildings, research on occupant behavior may point to ways to improve occupant behavior and, in this way, mitigate excessive residential energy consumption [8,9]. According to Ouyang and Hokao [10], if occupant behavior is affected, it can lead to significant energy use reduction because occupant behavior is often associated with unnecessary energy consumption and the mindless use of resources.

Occupant behavior can take many forms, and, in general, it encompasses a wide array of practices, habits, and choices, such as the use and purchase of domestic appliances, the setting of heating and air-conditioning temperatures, as well as the use of lighting [11,12,13,14,15]. What makes the understanding of occupant behavior even more challenging is that there is notable differentiation among occupants [16]. This was evidenced by the study of Bajaj and James [17], where buildings with identical envelopes and locations exhibited a 300% difference in terms of actual energy use, indicating the strong differentiation in the use of home appliances. A similar conclusion was drawn by Li et al. [18], whose study revealed notable differences among apartments in the same residential building in Beijing. Specifically, the difference among apartments ranged from 0 to 14.3 kWh/m² (at an average of 2.3 kWh/m²), indicating stark differences in occupant behavior in terms of temperature set points and opening/closing window habits.

This variation may be associated with households’ engagement in specific heating practices as well as preferences for certain energy behaviors [19,20,21]. In relation to heating practices, these are closely linked to occupants’ presence and movement in spaces, preferences for comfort conditions, and interactions with heating systems of residential buildings [22]. In addition, heating practices are affected by occupants’ lifestyle, income, education level, household size, personal perceptions of thermal comfort, and knowledge about energy saving [23,24,25,26]. The heating energy consumption profiles of households developed by Laskari et al. [27] provided insights into the ways people heat their residences in different ways and at different times. For instance, it was shown that even if occupants decided to operate heating systems at prescribed temperature limits, these were not necessarily kept throughout the day. This raises implications for the effectiveness of standardized heating profiles and temperature settings as well as for the establishment of national building regulations and international standards, which seek to ensure the health and comfort of occupants. In other words, the existing standards do not necessarily translate into actual occupant choices, as occupants often exhibit an unpredictable nature, which directly affects the consumption of energy [27].

In an effort to understand the acute differentiation in occupant behavior, a considerable number of studies focused on the factors affecting and shaping the adoption of certain behaviors. The most significant factors involve sociodemographic characteristics, socio-cultural environments, psychological factors, general values and beliefs, the availability of energy options, willingness to change routines and habits, the local climate, the characteristics of the buildings, environmental and climate change knowledge and views and individual preferences and needs [13,16,22,28,29,30,31,32,33]. Acknowledging the diversity in these factors, Schweiker and Wagner [34] classified the factors affecting occupant behavior into four categories: physiological, individual, environmental, and spatial.

At this point, the European Union is increasingly directing efforts and resources toward advising citizens about the adoption of energy saving measures [35,36,37]. The efforts aiming at altering occupant behavior and promoting the adoption of energy saving measures have taken various forms such as behavior change programs, communication tools (labeling prompts, posters, leaflets, and lists), personalized tools (such as events, workshops, and web forums), as well as economic and regulatory tools (incentives and disincentives, levies, dynamic energy pricing, tax reductions, refunds, penalties) [38,39].

The problem, however, is that the effectiveness of these efforts cannot be ensured mainly because they often do not take into account the heterogeneity of occupants and the differentiated influence of the factors affecting occupant behavior [38,40]. For this reason, it is necessary to first address the heterogeneity that exists among occupants in order to ensure the effectiveness of behavior change programs. To address the issue of heterogeneity, this study leverages the technique of market segmentation, where a large population is classified into distinct subsets of individuals who exhibit the same priorities, needs, and preferences [41]. The most notable benefit of this approach is that it enables policymakers to address the issue of heterogeneity, thereby enabling the practical implementation of results [41,42]. That is because policymakers can choose two or more clusters as target groups and devise tailored strategies and tools to reach them [41,42]. Moreover, the approach used in this study involves environmental variables because such variables are good indicators of energy saving behaviors [33,43]. Using a large representative sample of Greek citizens, this study aims to identify occupant typologies based on the energy saving measures they apply and their environmental attitudes. More analytically, this research is interested in identifying the heterogeneity of Greek citizens and the different ways in which they can be classified according to the practices they apply and their environmental attitudes. To establish a more practical approach, this research will identify citizen groups based on their differences in terms of their attitudes toward the adoption of heating, electricity, water, and transport use, as well as product purchase and eating habits. In this way, policymakers will know the areas where policy efforts and resources need to be directed and make informed decisions.

2. Materials and Methods

2.1. Questionnaire Design and Sampling Method

This paper presents an analysis that is based on data gathered through a broader research project focusing on citizens’ views and attitudes toward renewable energy investments. The study area was Greece, a southeastern EU member state chosen for its significant renewable energy potential and its remarkable policy effort to decarbonize its energy mix [44]. Successful energy transition in Greece could enhance the country’s role as renewable energy producer in the EU and also help to avoid expensive fuel imports. To perform this study, a structured questionnaire was developed based on literature regarding citizens’ attitudes to renewable energy and citizen investors’ socioeconomic profiles, motives, barriers, and preferences while considering the relevant policy framework for renewable energy investments (such as the research works of Willis et al. [45]; Vasseur and Kemp [42]). The questionnaire consisted of eight thematic sections, and this paper reports the analysis based on the data gathered through the sections that examined respondents’ energy behavior and environmental attitudes. All items were measured using five-point Likert scales to ensure precision and facilitate extensive statistical analyses [46]. The questionnaire’s efficacy and accuracy were validated through a pilot study with 30 Greek citizens from diverse socioeconomic backgrounds. Based on the results of the pilot study, to ensure clarity and comprehension for respondents, a few phrases in the items were rephrased. Additionally, the response scale for one item was modified to make it easier to answer, and the sequence of three items was adjusted to enhance the questionnaire’s coherence and prevent confusion. Following Law 4521/2018, Article 23, the research obtained a permit from the Research Ethics Committee of the Democritus University of Thrace (Decision No. 3/09-12-2019), which oversees all research conducted by the university’s affiliates. Reliability testing was performed in this study in order to ensure the data quality in the questionnaire. Specifically, Cronbach’s alpha coefficient was used to examine the inner consistency of the items and the reliability of scales. In theory, reliability tests can be performed before or after the survey [47]. In the present study, the tests were carried out after the survey. For the items reported in this paper, the values of alpha coefficients were all higher than 0.8, showing that the items are reliable and the questionnaires are acceptable.

To recruit a representative sample of citizens, we used simple random sampling, which does not require a high level of knowledge about the population under study [46,48]. The study area involved all Greek households and used them as sampling units for convenience and cost-effectiveness. Each selected household was represented by one randomly chosen member, While the sample size was determined using the simple random sampling formula without replacement [48]. In line with research principles, a pre-sampling with 50 subjects had to be conducted to estimate the population proportions for each variable, ensuring the questionnaire’s accuracy [46,49]. ‘Gender’ was the variable that required the largest sample size, so it was used to determine the final sample size. All questionnaires were completed through face-to-face interviews with the respondents in order to make sure that all questions were fully understood and that there would be no incomplete responses. In total, 1536 respondents took part in the study, and to secure their anonymity, all respondents placed the questionnaires into envelopes which were destroyed after data coding.

2.2. Data Processing

Collected data were coded and entered into the Statistical Package for Social Sciences (SPSS, version 22). First, descriptive statistics was applied to all variables, and then the non-parametric Friedman test was applied to the multivariates ‘thermal comfort’ and ‘frequency of application of energy cost reduction measures’ in order to compare the values of three or more correlated groups of variables. The distribution of the Friedman test is χ² distribution with degrees of freedom (df) df = k − 1, where k is the number of teams or samples. The test classifies the values of variables for every subject separately and estimates the mean rank of classification values for each variable. Then, to achieve the aim of the study, further statistical procedures were followed, and the data were segmented with the use of factor analysis. The objectives of factor analysis are dimensionality reduction and the examination of the correlations between observed variables, as well as to summarize correlations in a smaller number of new variables [50]. Here, factor analysis with Varimax rotation was performed on multivariates ‘measures for reducing energy costs’ and ‘environmental attitudes’ because the aim was to build clusters based on these variables. This served to reduce the observed variables into latent factors while retaining most of the initial dataset’s information. Specifically, we used the Guttman-Kaiser criterion to select principal components with eigenvalues ≥1, followed by Varimax rotation for optimization. The extent to which the factors obtained during the analysis explain the variance depends on how well these factors represent the original observed data. Factor analysis produces several numerical variables; each is associated with a unique and calculated factor score. These scores can be used for subsequent analyses, such as identifying variables related to the factors [50].

Then, cluster analysis was conducted using the factors from factor analysis. There are various methods for estimating distances and combining observations within clusters. Hierarchical and divisive hierarchical analyses are important, but k-means clustering is best for large samples, making it suitable for this study [51]. The k-means method classifies observations into predefined clusters, maximizing inter-cluster differences and minimizing within-cluster variance [52,53]. This iterative method refines cluster assignments based on the proximity of observations to cluster centroids. In order to indicate the most suitable number of clusters, various clustering combinations were tested, including a two-, three-, and four-cluster solution. Then, to validate the clusters, the chi-squared test was performed to ensure that the clusters differ significantly in terms of demographics. The chi-squared test helped evaluate the association between the clusters and these variables. In addition, ANOVA (analysis of variance) was used to determine if the means of the clusters were notably different. Finally, the typology of citizens was established, and the identified types were named to reflect their characteristics.

3. Results

3.1. Respondents’ Characteristics, Thermal Comfort, and Energy Cost Reduction Measures

In the sample, female respondents outnumbered (by 51.6%) their male counterparts and, with regard to age, a considerable share of respondents were aged 41 to 50 years old (27.9%). In relation to their occupation, the respondent percentages that were reported to be employed in the private (21.2%) and public (19.9%) sectors were considerably higher in comparison to the other occupations. Regarding education level, a substantial share of respondents were university graduates (22.3%) as well as upper secondary school graduates (20.8%). As for respondents’ family status, around half of them were married (51%), while a considerable percentage reported having two children (28.3%). Most respondents reported residing in urban centers (64.1%). As for their annual income, 28.5% reported earnings between 10,001 and 20,000 Euros, and 20.1% reported earnings between 5001 and 10,000 Euros a year. Respondents were also asked to report the thermal comfort at their residence over the last winter. The majority of respondents (81.9%) reported that the temperature at their residence was often or always satisfactory. However, the temperature was not slightly warmer (68.1%) or much warmer (84.8%) than the preference of respondents. Results from the non-parametric Friedman test indicated that the temperature at respondents’ houses was mostly satisfactory (mean rank 4.26) and slightly colder than their preference 228 (mean rank 3.11) (

Table A1. Percentages and differences concerning respondents’ thermal comfort last winter

	Never	Seldom	Often	Always	Do Not Know/No Answer	Mean Ranks (Friedman Test) *
Much colder temperature than preferred	54.0	25.7	14.9	5.3	0.1	2.46
Slightly colder temperature than preferred	30.3	36.2	27.1	6.3	0.1	3.11
Satisfactory	7.0	10.6	46.2	35.7	0.5	4.26
Slightly warmer temperature than preferred	34.0	34.1	23.2	7.8	0.8	2.99
Much warmer temperature than preferred	67.5	17.3	9.8	4.7	0.7	2.18

* N = 1536, Chi-Square = 2092.577, df = 4, p < 0.001.

, Appendix A). Respondents reported, moreover, the frequency with which they applied measures at their residence in order to reduce energy costs over the last twelve months (

Table A2. Percentages and differences concerning the frequency with which respondents applied energy cost reduction measures.

	Never	Seldom	Often	Always	Do Not Know/No Answer	Mean Ranks (Friedman Test) *
Turning heat off at the house	31.5	24.5	33.7	8.9	1.3	3.29
Lowering the temperature on the thermostat	18.6	28.1	38.5	12.9	2.0	3.82
Switching off the lights	18.8	19.3	36.4	24.4	1.1	4.12
Turning heat off in some rooms of the house	42.6	20.1	23.4	12.4	1.4	3.16
Heating and using only one room during daytime	47.3	17.8	24.7	9.3	0.8	2.90
Saving on the use of hot water	23.0	27.9	30.3	17.2	1.6	3.71

* N = 1536, Chi-Square = 670.114, df = 5, p < 0.001.

, Appendix A). Although ‘often’ and ‘always’ were rated highly for all measures, certain measures stood out. Specifically, the majority of respondents (60.8%) reported switching off the lights often and always, while around half of the respondents reported lowering the temperature on the thermostat (51.4%) and saving on the use of hot water (47.5%). Significant but lower proportions of respondents turned the heat off in some rooms of the house (35.8%) and turned the heat on in only one room of the house (34%). The non-parametric Friedman test showed that switching off the lights was the most important energy cost reduction measure (mean rank 4.12), followed by lowering the temperature on the thermostat (mean rank 3.82) and saving on the use of hot water (mean rank 3.71). Meanwhile, heating and using only one room during the daytime was the least applied measure, with a mean rank of 2.90.

3.2. Profiling Citizens Based on the Applied Energy Saving Measures and Environmental Attitudes

In order to investigate the underlying structure of respondents’ energy saving measures and environmental attitudes, principal components analysis was carried out on the multivariates ‘energy cost reduction measures’ and ‘environmental attitudes’. Regarding the multivariate ‘energy cost reduction measures’, prior to applying factor analysis, Cronbach’s alpha value (0.804), the Keiser–Meyer–Olkin index (0.744), and Bartlett’s test of sphericity (Chi-Square= 3171.339, with df = 15 and p < 0.001) confirmed the suitability of the data for factor analysis. After conducting Varimax rotation, first-order factor analysis gave three factors (

Table 1. Factor analysis results with Varimax rotation of respondents’ measures for reducing energy costs and environmental attitudes.

Principal Components—Variables Loaded in Each Factor	Loadings	Eigenvalue	Variance (%)
Heating only some rooms of the house (PC1_1)		3.046	50.758
Heating and using only one room during daytime	0.900
Turning heat off in some rooms of the house	0.854
Lower temperature and turning heat off at the entire house (PC1_2)		0.924	15.394
Lowering the temperature on the thermostat	0.908
Turning heat off at the house	0.780
Reduction of lighting and hot water use (PC1_3)		0.868	14.460
Switching off the lights	0.890
Saving on the use of hot water	0.734
Total variance (%): 80.613
Kaiser–Meyer–Olkin = 0.744, Bartlett Chi-Square = 3171.339, df = 15, p < 0.001
Saving resources and recycling (PC2_1)		4.404	40.033
I am willing to switch off the lights when leaving a room or use energy efficient light bulbs	0.834
I am willing to recycle	0.821
I am willing to re-use or give my old clothes to the needy	0.798
I am willing to turn off the tap while brushing teeth or shaving	0.711
I am willing to buy products contained in recyclable packages and manufactured in environmentally friendly ways	0.508
Environmentally friendly transport (PC2_2)		1.702	15.468
I am willing to use the bicycle	0.822
Instead of the car I am willing to use public transport	0.755
I am willing to cover short distances on foot	0.706
Environmentally friendly eating habits and preference for near-made products (PC2_3)		0.960	8.731
I am willing to buy organic products	0.836
I am willing to buy products travelling short distance	0.637
I am willing to reduce the consumption of meat or cured meat products if this is good for the environment	0.613
Total variance (%): 64.232
Kaiser–Meyer–Olkin = 0.858, Bartlett Chi-Square = 6094.195, df = 55, p < 0.001

). Variables related to the reduction of heat in some parts of the house were loaded on the first factor. Specifically, the first factor (PC1_1) included the variables ‘Heating and using only one room during daytime’ and ‘Turning heat off in some rooms of the house’ and thus could be named ‘Heating only some rooms of the house’. The second factor (PC1_2) comprises the variables ‘Lowering the temperature on the thermostat’ and ‘Turning heat off at the house’ and thus could be named ‘Lowering temperature and turning heat off at the entire house’. Finally, the variables ‘Switching off the lights’ and ‘Saving on the use of hot water’ were loaded on the third factor (PC1_3) which can be named ‘Reduction of lighting and hot water use’.

Before the application of factor analysis on multivariate ‘environmental attitudes’, the eligibility of data was confirmed through Cronbach’s alpha value (0.837), the Keiser–Meyer–Olkin index (0.858), and Bartlett’s test of sphericity (Chi-Square = 6094.195, with df = 55 and p < 0.001). The performance of factor analysis with Varimax rotation gave three factors. Variables related to habits that affect the consumption of electricity and water, as well as recycling products and reusing old things, fell under PC1 (Table 1). Specifically, the variables ‘I am willing to switch off the lights when leaving a room or use energy efficient light bulbs’, ‘I am willing to recycle’, ‘I am willing to re-use or give my old clothes to the needy’, ‘I am willing to turn off the tap while brushing teeth or shaving’ and ‘I am willing to buy products contained in recyclable packages and manufactured in environmentally friendly ways’ were loaded on the first factor (PC2_1). Due to the content of these variables, the first factor can be labeled ‘Saving resources and recycling’.

On the other hand, the second factor (PC2_2) contains variables that are explicitly related to transport. That is, the variables ‘I am willing to use the bicycle’, ‘Instead of the car I am willing to use public transport’, and ‘I am willing to cover short distances on foot’ fell under the second factor, which may be labeled ‘Environmentally friendly transport’. Finally, the variables loaded on the third factor (PC2_3) are associated with the consumption of meat and organic products as well as the purchase of products that travel short distances (Table 1). More specifically, the variables ‘I am willing to buy organic products’, ‘I am willing to buy products travelling short distance’ and ‘I am willing to reduce the consumption of meat or cured meat products if this is good for the environment’ fell under PC3 which can be labeled ‘Environmentally friendly eating habits and preference for near-made products’.

In order to identify the structure underlying citizens’ environmental attitudes and measures for reducing energy costs, a second-order factor analysis was performed. Before the performance of second-order factor analysis, the necessary tests were carried out; the Keiser–Meyer–Olkin index had a value of 0.493, and Bartlett’s sphericity test rejected the null hypothesis (Chi-Square = 173.738, df = 15, p < 0.001). The analysis gave three factors. The first factor (P1) is named ‘Heating one area of the house and avoiding recycling and resource saving (P1)’ and involves the variables ‘Heating only some rooms of the house (PC1_1)’ and ‘Saving resources and recycling (PC2_1)’. The second factor (P2) is named ‘Mindful use of electricity, water and transport (P2)’ and includes the variables ‘Environmentally friendly transport (PC2_2)’ and ‘Reduction of lighting and hot water use (PC1_3)’. The third factor is named ‘Extreme energy saving and pro-environmental diet and product purchase (P3)’ and contains the variables ‘Lower temperature and turning heat off at the entire house (PC1_2)’ and ‘Environmentally friendly eating habits and preference for near-made products (PC2_3)’ (

Table 2. List of principal components—Results of second-order factor analysis.

Principal Components—Results of Second-Order Factor Analysis	Loadings	Eigenvalue	Variance (%)
Heating one area of the house and avoiding recycling and resource saving (P1)		1.281	21.358
Heating only some rooms of the house (PC1_1)	0.763
Saving resources and recycling (PC2_1)	−0.757
Mindful use of electricity, water and transport (P2)		1.158	19.292
Environmentally friendly transport (PC2_2)	0.746
Reduction of lighting and hot water use (PC1_3)	0.741
Extreme energy saving and pro-environmental diet and product purchase (P3)		1.076	17.926
Lower temperature and turning heat off for the entire house (PC1_2)	0.719
Environmentally friendly eating habits and preference for near-made products (PC2_3)	0.716
Total variance (%): 58.576
Kaiser–Meyer–Olkin = 0.493, Bartlett Chi-Square = 173.738, df = 15, p < 0.001

).

The method of k-means cluster analysis was then used in order to identify the association of each member to the cluster. An analysis of variance (ANOVA) was performed to test if respondents can be clustered according to their environmental attitudes and energy cost reduction measures. The existence of important differences enabled the acceptance of the proposed clustering.

Two-, three-, and four-cluster solutions were explored using the k-means procedure, which was used in order to detect natural groupings of respondents based on respondents’ environmental attitudes and energy reduction measures. The best solution was given by the solution of four clusters as it provided an optimal balance between within-cluster homogeneity and between-cluster heterogeneity.

As shown in

Table 3. Mean factor loadings by the four clusters.

	CL1 (17.8%) Extreme Heat Savers and Pro-Environmental Food Consumers	CL2 (27.1%) Heat Savers and Environmentally Unaware	CL3 (27.6%) Environmentally Aware Energy Savers	CL4 (27.4%) Mindful Resources and Transport Users	Cluster		Error		F	p-Value
					Mean Square	df	Mean Square	df
P1	−0.79557	0.71974	0.55499	−0.75406	253.142	3	0.506	1532	500.034	0.000
P2	−1.00981	−0.66810	0.61778	0.69679	277.252	3	0.459	1532	603.985	0.000
P3	0.72369	−0.64145	0.81267	−0.65411	258.409	3	0.496	1532	521.054	0.000

, there are discernible differences among the four clusters. Cluster 1 was the smaller cluster, representing 17.8% of the sample. Citizens in this cluster gave a high rating to the third factor, ‘Extreme energy saving and pro-environmental diet and product purchase’. They seem to be inclined to follow somewhat strict energy saving practices such as turning the heat off at the entire house and reducing meat consumption. In addition, this cluster stands out as ascribing less importance both to the first and second factor (‘Heating one area of the house and avoiding recycling and resource saving’ and ‘Mindful use of electricity, water and transport’). This cluster could be labeled ‘Extreme heat savers and pro-environmental food consumers’.

The second cluster represents 27.1% of the sample, and its members are characterized by their positive evaluation of the factor ‘Heating one area of the house and avoiding recycling and resource saving’ and their negative evaluation of the two other factors. They seem likely to follow the practices of the first factor, and thus this cluster can be labeled ‘Heat savers and environmentally unaware’.

The third cluster accounts for 27.6% of the sample, and it is the only cluster whose members give importance to all three factors. This cluster also presents the highest positive score for the third factor, ‘Extreme energy saving and pro-environmental diet and product purchase’. This cluster could be labeled ‘Environmentally aware energy savers’.

Finally, the fourth cluster represents 27.4% of the sample and presents positive scores for the second factor ‘Mindful use of electricity, water and transport’ and negative scores for the other two factors. Due to these scores, the fourth cluster could be labeled ‘Mindful resource and transport users’.

The analysis of the clusters’ profile is based on statistically significant differences between the four clusters (

Table A3. Chi-squared tests results of differences between clusters.

Variable	Scale	CL1 (17.8%)	CL2 (27.1%)	CL3 (27.6%)	CL4 (27.4%)
Sociodemographic variables
Gender	Male	42.3%	52.0%	53.8%	43.2%
	Female	57.7%	48.0%	46.2%	56.8%
Occupation	Public employee	25.9%	16.1%	21.0%	18.8%
	Private employee	24.1%	20.6%	22.2%	18.8%
	Freelancer	11.3%	12.7%	10.1%	13.8%
	Entrepreneur	2.9%	5.5%	5.2%	3.1%
	Homemaker	6.6%	5.8%	4.7%	7.3%
	Crop farmer	6.2%	7.0%	5.4%	5.2%
	Livestock farmer	1.1%	2.4%	2.1%	1.4%
	Retired	12.0%	13.2%	17.5%	19.2%
	Unemployed	11.3%	15.8%	9.2%	15.7%
Income	Less than 5000 Euros	8.0%	9.8%	13.7%	11.4%
	5001–10,000 Euros	21.9%	17.3%	20.5%	21.1%
	10,001–20,000 Euros	31.4%	24.0%	34.2%	25.2%
	20,001–30,000 Euros	5.8%	7.9%	8.0%	8.8%
	More than 30,000 Euros	3.3%	6.2%	2.1%	3.6%
Environmental reasons for investing in RES
Contribution to air pollution mitigation	Strongly disagree	0.7%	3.4%	0.7%	0.7%
	Disagree	2.6%	8.4%	1.9%	1.9%
	Neither agree or disagree	4.7%	16.8%	13.7%	5.0%
	Agree	34.3%	49.4%	39.2%	39.2%
	Strongly agree	57.7%	22.1%	44.6%	53.2%
Contribution to the mitigation of resources depletion	Strongly disagree	0.4%	2.4%	1.2%	0.7%
	Disagree	5.5%	7.0%	2.8%	2.9%
	Neither agree or disagree	3.3%	23.7%	16.5%	10.9%
	Agree	36.9%	46.5%	35.8%	36.8%
	Strongly agree	54.0%	20.4%	43.6%	48.7%
Flora protection	Strongly disagree	0.4%	1.0%	0.2%	0.5%
	Disagree	2.9%	7.9%	2.4%	2.9%
	Neither agree or disagree	7.3%	18.2%	13.9%	9.3%
	Agree	33.2%	42.9%	36.6%	37.8%
	Strongly agree	56.2%	30.0%	46.9%	49.6%
Fauna protection	Strongly disagree	0.4%	0.7%	0.7%	1.2%
	Disagree	2.9%	6.5%	2.1%	2.1%
	Neither agree or disagree	6.9%	18.5%	13.0%	8.8%
	Agree	31.0%	43.2%	36.3%	37.8%
	Strongly agree	58.8%	31.2%	47.9%	50.1%
Thermal comfort—House temperature at home in last winter
Much colder temperature than preferred	Never	61.3%	55.9%	38.4%	63.2%
	Seldom	26.3%	25.7%	22.6%	28.5%
	Often	10.2%	16.1%	25.7%	5.9%
	Always	2.2%	2.4%	13.0%	2.4%
Slightly colder temperature than preferred	Never	40.1%	34.85	19.3%	30.4%
	Seldom	32.8%	39.8%	30.2%	40.9%
	Often	25.9%	22.1%	36.8%	23.0%
	Always	1.1%	3.1%	13.4%	5.7%
Satisfactory temperature	Never	4.7%	8.9%	9.0%	4.8%
	Seldom	6.9%	11.5%	14.4%	8.3%
	Often	46.7%	41.5%	47.4%	49.4%
	Always	41.6%	37.4%	28.3%	37.5%
Slightly warmer temperature than preferred	Never	33.2%	27.8%	40.3%	34.4%
	Seldom	34.7%	37.2%	24.8%	40.1%
	Often	23.4%	25.7%	23.8%	20.2%
	Always	8.0%	9.1%	9.4%	4.8%
Much warmer temperature than preferred	Never	76.3%	54.9%	61.6%	80.3%
	Seldom	16.1%	21.8%	15.3%	15.7%
	Often	4.4%	19.2%	10.8%	2.9%
	Always	2.9%	3.6%	10.6%	1.0%
Punctuality in electricity bill payments in the last 12 months
Electricity bills were not paid		0.0%	1.4%	1.9%	2.6%
Electricity bills were paid late with great difficulty		8.4%	9.8%	9.9%	6.9%
Electricity bills were paid on time but with great difficulty		37.2%	38.8%	51.9%	41.8%
Electricity bills were paid on time without difficulty		54.4%	49.9%	36.3%	48.7%

, Appendix A). More analytically:

Cluster 1 ‘Extreme heat savers and pro-environmental food consumers’:

The first cluster represents the lowest percentage of the sample (17.8%) and here, citizens are mainly female (57.7%). This cluster has the largest percentages of citizens employed in the public (25.9%) and the private (24.1%) sectors but also the lowest percentage of entrepreneurs (2.9%). More than half of the citizens in this segment earn between 5001 and 20,000 Euros per year. Concerning the environmental reasons for investing in renewables, ‘Extreme heat savers and pro-environmental food consumers’ stand out as the group with the highest percentages of total agreement with the examined environmental reasons for investments. It is thus the cluster that is mostly environmentally driven, which is also reflected in the high importance they ascribed to the dimension ‘Extreme energy saving and pro-environmental diet and product purchase’. The majority of members in this group reported that the temperature at their residence was satisfactory (88.3%) and not higher than preferred (92.4%). Finally, this cluster is distinguished from the others because half of its members have paid electricity bills on time without difficulty (54.4%).

Cluster 2 ‘Heat savers and environmentally unaware’:

The second cluster represents about 27.1% of the sample, and male members outnumber their female counterparts by four percentage units. They attached more importance to the dimension that concerned turning on heat only in one area of the house but attached less importance to the other factors that concerned, inter alia, the mindful use of electricity, water, and transport as well as stricter energy saving practices and pro-environmental diet and product purchase. Regarding occupation, this segment presents the lowest percentage of public employees (16.1%), and it has higher shares of unemployed citizens (15.8%), crop farmers (7%), and entrepreneurs (5.5%), though all clusters present low percentages for these occupations. Compared to the other clusters, this cluster has lower percentages for annual income categories 5001–10,000 Euros (17.3%) and 10,001–20,000 (24%). Concerning the environmental reasons for investing in RES, this segment stands out as presenting the greatest percentages of disagreement with the examined environmental investment reasons. In terms of thermal comfort, this segment has the highest percentage of citizens who reported that temperature was often much higher than preferred (19.2%).

Cluster 3 ‘Environmentally aware energy savers’:

Cluster 3 accounts for 27.6% of the sample, and its members are mainly men (53.8%). This cluster is distinguished as the only one that gave importance to both environmental and energy saving dimensions. Most citizens are public and private employees, but this segment also presents the lowest percentage of unemployed citizens (9.2%) and freelancers (10.1%). In addition, it presents the highest share (34.2%) of citizens earning between 10,001 and 20,000 Euros a year. Concerning the environmental reasons for investing in renewables, the members of this cluster gave high ratings to all examined reasons. This cluster stands out as it presents the highest percentages of citizens reporting that temperature at their house was often (25.7%) and always (13%) much lower than preferred. It also has the highest percentage (36.8%) for slightly colder temperatures than desired. Finally, about half of the citizens in this segment paid their electricity bills on time but with great difficulty.

Cluster 4 ‘Mindful resources and transport users’:

The fourth cluster represents 27.4% of the sample and comprises more female than male citizens. This cluster gave more importance to the second dimension (‘Mindful use of electricity, water and transport’) and less importance to the first and third dimensions (‘Heating one area of the house and avoiding recycling and resource saving’ and ‘Extreme energy saving and pro-environmental diet and product purchase’). In terms of occupation, this cluster stands out as having the highest percentages of citizens who are pensioners (19.2%) as well as homemakers (7.3%). In addition, this cluster presents the lowest share of private employees (18.8%) compared to the other clusters. Much like the previous cluster, however, the majority of members of the fourth cluster earn between 5001 and 20,000 and gave high ratings to all environmental reasons for investing in renewable energy sources. Nearly half of the members (49.4%) of this cluster reported that last winter, the temperature was satisfactory at their residence. Citizens in the fourth segment either paid their electricity bills on time without difficulty (48.7%) (as citizens in Clusters 1 and 2) or paid them on time but with great difficulty (41.8%) (as citizens in Cluster 3).

4. Discussion

The typology developed in this study provides a new angle for understanding occupant behavior and captures a nuanced picture of the relationship between environmental attitudes and the adoption of energy cost reduction measures. More analytically, the study has pointed at strong heterogeneity among respondents, which confirms previous literature showing a high level of differentiation in occupants’ energy behavior [17,18,19,20,23,24,25]. However, this study adds to the existing literature in that it explains how environmental attitudes are associated with the adoption of energy saving measures. This typology can thus help policymakers understand the observed heterogeneity and the characteristics of occupants and, in this way, devise the most suitable strategies and policies in order to appeal to each cluster effectively. Pointing at a high level of heterogeneity, our findings support arguments in favor of differentiating the efforts aimed at affecting the energy behavior of occupants. Specifically, behavior change programs should be designed in line with this differentiation, while ‘one-size-fits-all’ approaches [54] ought to be avoided due to the recorded heterogeneity.

It is interesting to observe that of the four clusters, the cluster ‘Environmentally aware energy users’ is the only one that is positively affected by all examined factors, indicating that there is probably only one part of the population in the study area that is already applying energy saving measures driven by positive environmental attitudes. The sociodemographic profile of this cluster stands out as it is the cluster with the highest representation of high incomers, while the environmental reasons for investing in RES (renewable energy sources) were positively evaluated. It is also worth noting that the members of this cluster are more likely to apply measures that compromise their thermal comfort despite their economic status. Therefore, it is possible that this cluster is genuinely concerned about the environment and seeks to minimize its environmental footprint through the application of energy saving measures. As the members of this cluster seem to be already engaged in the application of energy saving measures in their households, it is not necessary to target them with energy behavior change programs. However, since this segment has evaluated positively the environmental reasons for investing in RES, it may be worthwhile to target them as a potential consumer base for sustainable products and services that align with their values, such as residential renewable energy systems and energy efficient appliances [54,55,56].

In contrast to ‘Environmentally aware energy users’, the other clusters require considerable policy attention and dedicated efforts to affect their energy behavior as they seem to be engaged only in specific pro-environmental practices while neglecting other equally important ones. Specifically, the cluster ‘Heat savers and environmentally unaware’ applies energy saving measures probably only to lower the household’s heating costs as it has exhibited the lowest level of pro-environmental attitudes. It is also a cluster that pays attention only to saving on heating rather than on other resources and, in general, demonstrates a reluctance to adopt environmentally friendly behaviors in terms of water use, transport, product purchase, and eating habits. As this cluster saves energy through heating practices, one may argue that this cluster is already contributing to residential energy saving targets and, therefore, there is no need to address it. However, this would be a mistaken approach because occupants who apply measures only to lower household energy costs are not necessarily engaged in energy saving and are thus likely to revert to their old habits once energy prices are reduced. This observation emphasizes the need to dedicate more research to the motives that drive occupants to adopt or reject energy saving practices—a need that has been noted in previous research [30,32].

The other two clusters, ‘Extreme heat savers and pro-environmental food consumers’ and ‘Mindful resources and transport users’ are markedly different and demonstrate significant room for improving their environmental practices. Specifically, ‘Extreme heat savers and pro-environmental food consumers’ are inclined to apply very strict heat saving measures such as turning the heat off at the entire house and have expressed positive environmental attitudes in terms of eating habits and the purchase of products. At the same time, however, they seem to be much less willing to adopt environmentally friendly behavior in terms of lighting, hot water use, and transport. As the members of this cluster have positive environmental attitudes in terms of heating, eating habits, and product purchase, it is possible that their reluctance to save on other resources and to choose more environmentally friendly options may be stemming from a low level of knowledge about the environmental and economic impact of habits related to the mindless use of electricity, hot water use, and environmentally unfriendly transportation. This explanation is corroborated by research showing that occupant practices are also affected by occupants’ knowledge about energy saving [23,24,25]. Therefore, this cluster requires an approach that will focus on awareness campaigns highlighting the environmental impact of water waste and car usage and that will be combined with economic incentives and policy enforcement. Some examples of the latter would be rebate programs, stricter regulations to limit water usage for unnecessary activities during drought periods, as well as the provision of incentives such as discounts on electricity and water bills for households that demonstrate significant reductions in usage.

It seems that ‘Mindful resources and transport users’ are the exact opposite of the cluster described above, as the members of this cluster have adopted measures with which they reduce the consumption of lighting and hot water use while they express positive environmental attitudes in terms of transport. However, this cluster stands out as the group that does not apply any measures regarding heating; interestingly, it includes the highest shares of women, retirees, and homemakers, and, therefore, it is possible that there is an association with thermal comfort. That is, women and old people have a lower tolerance to low temperatures, and for this reason, they may be much more reluctant to reduce heating in their houses [57]. This points to the need to address this cluster by disseminating information about the use of space heating controls instead of whole-house thermostats. It has been shown that using programmable space heating controls can ensure the secure and cost-effective operation of heating systems, which enable households to reach and maintain preferred temperatures and possibly save energy by heating their spaces only when it is necessary [58].

5. Conclusions

The present study has examined, for the first time, whether citizens as energy consumers can be classified on the basis of their environmental attitudes and the measures they apply to reduce energy costs. Our analysis has provided a novel citizen typology by identifying four segments, named as Extreme heat savers and pro-environmental food consumers, Heat savers and environmentally unaware, Environmentally aware energy savers, and Mindful resource and transport users (representing 17.8%, 27.1%, 27.6%, and 27.4% of the sample, respectively). The identified clusters demonstrate significant differentiation, highlighting an association between environmental attitudes and energy behavior. Consequently, citizens should not be seen as a homogeneous target group by policymakers seeking to promote more sustainable energy practices but rather as a highly heterogeneous group requiring careful and differentiated approaches. Policymakers should also pay attention to the fact that each cluster places different emphasis on energy saving practices and exhibits different environmental attitudes. Specifically, as ‘Heat savers and environmentally unaware’ apply energy saving measures only to lower heating costs, it is important to raise their awareness about sustainable practices in terms of water use, transport, product purchase, and eating habits. Policy attention should also be placed on ‘Extreme heat savers and pro-environmental food consumers’ as they save on heating but neglect other important saving practices in terms of lighting, hot water use, and transport.

Finally, the findings of this study must be considered in view of certain limitations that point to the directions for future research. Although this study was based on a representative citizen sample, the findings are applicable only to the study area. As such, they ought to be seen only in view of the wider economic and environmental context in Greece. In order to attain an EU-wide level, it is recommended that the study be replicated on citizens in other EU countries. Another limitation of this study was that it is based on reported responses rather than data derived through smart meters. Nevertheless, the combination of questionnaire administration along with the use of meters is an excellent direction for a future study on occupant behavior. Finally, it is recommended that this analysis be repeated after a period of time, as it is necessary to examine the possibility that behaviors evolve and the ways in which they evolve.