Recycled CO₂ in Consumer Packaged Goods: Combining Values and Attitudes to Examine Europeans’ Consumption Intentions

Abstract

The main objective of this study was to investigate European consumers’ intentions to purchase cosmetics and detergents with green ingredients made from recycled CO₂. Aiming to better understand both moral and practical criteria of consumers’ intentions, a combination of the Values-Beliefs-Norms and the Theory of Planned Behaviour models served as the basis of this study’s theoretical framework. The combination was extended with risk perception about global warming, scepticism and media influence. Online interviews were conducted with stratified samples based on gender and age distributions in France, Germany, Greece and Spain. Structural equation modelling and moderation analyses were employed to analyse the data. The results indicated that consumption intentions are generated by consumers’ biospheric values and a sequence of risk perception, awareness of consequences, and ascription of responsibility while they are directly determined (in declining order) by perceived behavioural control, personal norms, attitudes and subjective norms. Subjective norms indicated additional indirect impacts on consumption intentions through personal norms and ascription of responsibility. Moderation also indicated that the relationship between perceived behavioural control and consumption intentions is stronger in consumers, who are less sceptical towards ecological claims on packaging, while the relationship between personal norms and consumption intentions is stronger in consumers, who are less influenced by advertisements. Theoretical, managerial and social implications were derived from the results.

1. Introduction

“The end of the fossil fuel era is assured. The only questions are: Will we move fast enough to limit the worst of climate chaos? And will the transition to renewables be fair, just and equitable? It is up to all of us to ensure the answer to both these questions is—yes”

[1]. UN Secretary-General’s video message to the International Energy Agency’s 50th Anniversary Celebration in Paris, February 2024.

Extreme weather events (drought, fires, etc.) require urgent national, European, and global policies [2,3]. Global warming is a problem attributed to human activities, and stakeholders (businesses, governments, and academia) should engage in social climate science and work towards solutions [4,5]. Climate change is attributed to carbon emissions, undisputedly a top priority issue on the global environmental agenda [6]. In the drive towards decarbonising societies, the role of individual consumption behaviour has been repeatedly emphasised [7,8]. It is a common belief that the chemical industry is responsible for large amounts of carbon emissions [3], and therefore, any environmentally friendly change in the production of consumer packaged goods (CPGs) is welcome.

This study focuses on examining cosmetics and detergents that include chemicals made from recycled CO₂. Capturing and recycling CO₂ reduces its amount in the atmosphere; CO₂ accounts for 75% of all carbon gases [9] responsible for global warming and climate change [10]. This paper presents part of a project to design a reactor for use in a green process that can produce innovative fluid ingredients, namely Glycolic acid, n-Valeraldheyde, and Limoxal^TM. These chemicals are important, as they are used in the production of fragrances embedded in all cosmetics and detergents [https://suncochem.eu (accessed on 30 October 2024)].

Previous attempts to explain consumers’ engagement in pro-environmental purchasing behaviour have extensively used two theoretical models, namely the Theory of Planned Behaviour (TPB) by Ajzen [11] and the Value-Belief-Norms (VBN) Theory by Stern et al. [12]. Review papers have gathered numerous articles, in which TPB was expanded or modified [7,13,14,15,16]. Several publications involving VBN expansions appear in the pro-environmental literature [17,18,19]. Even fewer efforts have applied and compared the two theoretical models or parts of them [20,21,22]. A study of the literature [23] indicates that fully combining VBN and TPB perspectives can significantly contribute to a deeper understanding of consumer intentions, as the former model focuses mainly on values while the latter focuses on attitudes. Additionally, the extension of such a combination with external variables originating from the marketing environment is expected to broaden the field of the overall examination.

Hence, this study aimed to examine the power of a combined VBN-TPB theoretical framework, which is modified by Leiserowitz’s [24] climate change risk perception and encompasses the moderating role of scepticism [25] and media influence [26] to examine European consumers’ intentions to purchase cosmetics and detergents with ingredients made from recycled CO₂.

In an effort to fulfil the above aim, consumer surveys with probability sampling and adequate sample sizes were conducted in France, Germany, Greece and Spain with a structured questionnaire. Characteristics of the sample and descriptive statistics of the variables were assessed employing classical statistics. Data were analysed using structural equation modelling (SEM), which was judged to be the most appropriate method for the combined VBN-TPB model [27]. To reveal the role of the additional variables, moderation analysis was employed.

Overall contribution of this study concerns the very first examination of consumers’ intentions to purchase cosmetics and detergents that contain ingredients made from recycled CO₂, with the goal of reducing its amount in the atmosphere for the purposes of climate change mitigation. Further, the theoretical contribution of this paper mainly concerns the simultaneous examination of values and attitudes in a combined VBN-TPB theoretical framework. The theoretical contribution is also provided as VBN was modified by replacing the general ecological worldviews with specific perceptions about risks and threats due to global warming. In addition, the overall theoretical framework was extended by the moderating effect of variables relevant to advertising.

2. Theoretical Background

Regarding the guidance of academic theory, in the previous attempts to explain consumers’ engagement in pro-environmental purchasing behaviour, there is extensive use of two well-known theoretical models, namely the Theory of Planned Behaviour (TPB) by Ajzen [11] and the Value-Belief-Norms (VBN) Theory by Stern et al. [12]. TPB incorporates three predictors of Intentions, namely attitudes, subjective norms, and perceived behavioural control, while VBN suggests a chain of values (altruistic, egoistic, and biospheric), beliefs (Dunlap’s and Van Liere’s [28] new environmental paradigm, awareness of consequences and ascription of responsibility), all affecting in sequence the personal norms as the main predictor of pro-environmental behavioural variables.

Fishbein and Ajzen [29] introduced intentions as the closer path to actual behaviour, understanding them as a person’s willingness to engage in a particular action. They firstly introduced the Theory of Reasoned Action (TRA), in which attitudes and subjective norms were assumed to be the predictors of intentions. Later, Ajzen [30] presented TPB—as an extension of TRA—which includes not just the volitional factors (attitudes) or influences of a close social circle (subjective norms) but also the non-volitional aspect of means and opportunities (perceived behavioural control). TPB’s founder argued that individuals behave according to their attitudes in specific situations [11]. Ajzen ([11], pp. 181–182) stated that TPB “… has been designed to predict and explain human behaviour in specific contexts”, whereas “intention can find expression in behaviour only if the behaviour in question is under volitional control” and thus “to the extent that a person has the required opportunities and resources, and intends to perform the behaviour, he or she should succeed in doing so”.

On the other hand, VBN is a morally oriented theory. Stern et al. [12] introduced VBN, inspired by Schwartz’s [31,32] moral norm-activation theory, and proposed that any social movement is developed based on a chain of human values and specific beliefs about consequences and responsibilities, which activate personal norms that obligate individuals to support the movement’s goals. Stern et al. ([33], p. 1614) stated that when one believes that a particular condition has harmful consequences and that one is responsible for them, “he or she will experience a sense of moral obligation (a personal moral norm) to take action to prevent the expected harm”. Stern et al. ([12], p. 85) adopted Dunlap’s and van Liere’s [28] NEP as “a view that human actions have substantial adverse effects on a fragile biosphere” that “measures awareness of very general adverse consequences of environmental conditions”. In VBN, Stern et al. [12] placed NEP in the second link of the chain, assuming that it is, overall, influenced by values.

TPB has always been considered a rational approach theory, in which individuals are seen as self-interested consumers [7]. VBN was perceived from a different viewpoint, namely under the assumption that pro-environmental behavioural activities are driven by citizens’ values and beliefs, which shape their moral obligations towards public interests [34]. However, it would be extremely arbitrary and misleading to conclude that morality is excluded from the TPB conceptualisation or that attitudes are discredited in VBN. Ajzen [11,35] suggested adding moral obligation, when applying TPB in ethical contexts, as one’s personal obligation to act-or refuse to act-in favour of a particular cause that would certainly affect one’s intentions. On the other hand, it should be noted that Stern et al.’s [33] work began not with the denial of attitudes but with the intention of exploring their deeper meaning and their consequent formulation. In the long-standing efforts to gain a deeper understanding of intentions and behaviour, there have been some attempts to extend TPB with values or norms [8,36,37,38] and VBN with attitudes [39,40]. There are also some attempts that have applied modified frameworks containing variables from both theories in several subjects (environmental behaviour [41]; collaborative consumption [42]; organics [43]; environmentally sustainable products [20,21]; road transportation [22]; natural food [44]).

The fact is that each and every single theory, when applied alone, leaves a considerable percentage of variance unexplained in the under-examination behaviour [18,44,45]. Therefore, it is argued that investigating both rational and moral motives, which has escaped the attention of academics [23,44] so far, should not be neglected anymore. In this study, a full combination of both theories was employed with one modification and two additional variables. In VBN, a decision was made to replace NEP [28] with risk perception [24]. This decision was made as NEP concerns general environmental beliefs [28] while risk perception is a contemporary index particularly focusing on global warming and climate change and thus, it is suggested to be more closely related to the topic of this study. Further, during the design stage of this study, it was rather rationally deemed that consumers’ response to consumer packaged goods (CPGs) related to innovations, such as CO₂ recycling, should definitely depend on aspects of advertisement and further promotional tools. Therefore, the conceptual framework was extended by two exogenous variables, namely scepticism [25] and media influence [26], representing the negative and positive perceptions of advertising, respectively.

Hence, the theoretical framework of this study (Figure 1) incorporates the three types of personal values, namely egoistic (EV), altruistic (AV), and biospheric (BV), the risk perception of climate change, the more specific beliefs about adverse conditions in the natural environment, namely awareness of consequences (AC) and the relevant humanity’s duty, namely ascription of responsibility (AR), leading to personal norms (PN) for pro-environmental action. The variables representing the three predictors of TPB, namely attitudes (Att), subjective norms (SN) and perceived behavioural control (PBC), are also included and, along with personal norms (PN), are assumed to be able to directly predict consumption intentions (CI). Scepticism and media influence were added and treated as moderators. It is to be noted that Stern’s [34] suggestion about the sequential relationships in VBN was followed in the formulation of hypotheses in the overall combined framework of this study.

With regards to the methods followed for the analysis of the results of this study, structural equation modelling (SEM) was chosen for data analysis, as it is considered the ideal technique for models incorporating multiple regression equations. SEM requires two models, the measurement model and the structural model. First, in the measurement model analysis, the validity of all constructs in the model is examined. Then, the fit of the data to the theoretical model is tested by the goodness-of-fit (GOF) estimates. Second, SEM provides the structural model, in which multiple hypothesised relationships are tested simultaneously. Both VBN and TPB are constructed to examine a series of variables in a sequence, where a given variable is assumed to influence the next one, and so on. This logic is even more present in the theoretical framework of this study, in which the two models are combined, and therefore, a network of relationships should be examined in order to assess the overall configuration of intentions.

3. Literature Review and Hypotheses

With regard to the replacement of the new environmental paradigm (NEP) [28,46] with the climate change risk perception index [24], it is noted that the results of some recent studies concerning NEP are rather limited [43,47]. It was also noticed that there have been a few previous studies in which NEP was replaced by attitudinal or other constructs (e.g., [48,49]). Therefore, employing a contemporary scale of beliefs concerning global warming, which is attributed to carbon gases, should be ideal for a topic specifically focusing on carbon gas reduction.

3.1. Values

Egoistic values are customarily assumed to prevent people from participating in any type of ecologically conscious behaviour. Indeed, in some VBN applications, egoistic values have been found either unrelated to environmental beliefs [22,40,47,50,51,52] or negatively related to them [42,53,54]. In this study, egoistic values are hypothesised to be negatively related to risk perception. Therefore, the following hypothesis was set:

H₁. 

Egoistic Values influence Risk Perception negatively.

From the opposite point of view, altruistic values are at the core of norms conceptualisation, and thus, they are expected to positively influence any type of ecologically conscious behaviour [50,55,56,57]. In several VBN applications, altruistic values have indicated a positive effect on pro-environmental beliefs [22,40,42,47,51,52,53,54,58,59]. In this study, it is assumed that altruistic values will have a positive effect on the respondents’ perceptions of risks that are attributed to climate change and threaten not just themselves but all human kind. Therefore, the following hypothesis was set:

H₂. 

Altruistic Values influence Risk Perception positively.

The very semantic concept of biospheric values expresses deep-rooted feelings about the natural environment [57]. Biospheric values have been found to indicate an exclusive [20] or a stronger—than other values—impact in a VBN model [22,52,60,61,62]. Therefore, a relevant hypothesis was set in this study:

H₃. 

Biospheric Values influence Risk Perception positively.

3.2. Beliefs

Many times, beliefs have been conceptualised as the intermediate path to link values with intentions and behaviours [56,63,64]. This view has been verified by a considerable number of previous studies [40,65,66]. Stern and his colleagues [33,67], based on Schwartz’s model [31,32], included in their model two types of beliefs, namely awareness of consequences and ascription of responsibility. In this study, the construct of consequences focuses on the harmful effects of conventional production and distribution, such as the exhaustion of fossil fuels, energy sources and carbon emissions, which are responsible for a significant part of climate change. Ascription of responsibility examines, in essence, whether one feels jointly responsible or not for carbon emissions and global warming. It is hypothesised that awareness of consequences is affected by Risk Perception while ascription of responsibility is affected by awareness of consequences. Accordingly, the following two hypotheses were set:

H₄. 

Climate change Risk Perception positively influences Awareness of Consequences.

H₅. 

Awareness of Consequences positively influences Ascription of Responsibility.

3.3. Personal Norms

Personal norms reflect one’s feelings of moral obligation to act upon one’s beliefs and values in order to repair damages for which he holds responsibility [12]. Personal norms differ from social norms, in the sense that they force towards change and not towards support of social status [12]. There have been studies indicating that the impact of ascription of responsibility on personal norms is the strongest of all relationships in the VBN chain [40,47]. The question is whether the respondents’ responsibility attribution impacts their sense of moral obligation. At this point, in the value and belief sequence, the formation of personal norms is examined overall. This study specifically examines whether taking personal responsibility influences the formation of personal norms favourable to CPGs containing green chemical ingredients in order to mitigate climate change. The proposed hypothesis is that:

H₆. 

Ascription of responsibility positively influences Personal Norms.

3.4. Intentions

According to Stern [30], personal norms are considered to be the closer predisposition to intentions and behaviours, especially altruistic, such as pro-environmental behaviour. Indeed, there have been many studies adopting a VBN model to verify the strength of personal norms in predicting pro-environmental intentions [20,47,68,69] and behaviour [40,42,43,51,53,54,56,58,59]. In this study, it is mandatory to examine behavioural intentions and not actual behaviour as the new, green CPGs are still in the research and development stage. Therefore, the following hypothesis was set:

H₇. 

Consumption Intentions are influenced positively by Personal Norms.

As mentioned in the Introduction, this study aimed to combine both VBN and TPB, and for that reason, the TPB predictors are assumed to influence consumption intentions, along with personal norms. TPB has been extensively employed in the examination of pro-environmental behavioural intentions [13,14,15,45]. With regards to cosmetics, there have been some studies examining natural or organic products in a variety of places ([70], US; [71], Indonesia; [72], Malaysia; [73], Taiwan; [74], Indonesia; [75], South Africa). Although there are discrepancies in their results, they seem to agree on the particularly strong influence of attitudes on purchase intentions compared to the lower impacts of subjective norms and perceived behavioural control. Studies about detergents are undoubtedly rare. In Indonesia, Arli et al. [76] applied a modified TPB model and found that subjective norms are the stronger predictor of consumers’ intentions to prefer environmentally friendly household products.

The TPB is a rational theory of causal relationships. It assumes that the stronger an individual’s attitudes, the influence of other people, and the perception of the existence or absence of barriers associated with consumer behaviour, the stronger an individual’s intentions to engage in that behaviour will be [11,77]. Particularly relevant to attitudes, they concern a person’s positive or negative emotional tendency towards a certain action or preference [29]. There have been some previous studies in cosmetics [74,75] or other subjects [52] that indicated a predictive impact of attitudes on intentions. In this study, it is assumed that when consumers hold positive attitudes towards cosmetics and detergents containing recycled CO₂, they are more likely to form positive intentions to buy them in an effort to contribute to climate change mitigation. Accordingly, the following hypothesis was set:

H₈. 

Consumption Intentions are positively influenced by Attitudes.

Subjective norms concern a person’s perceptions about what other—important or close to him—people (such as family, friends, colleagues, etc.) would like him to do or avoid doing or what these people actually do themselves [11]. These perceptions collectively form a type of social pressure capable of influencing an individual’s decision-making [7]. Some previous studies in cosmetics [74,75] indicated a predictive impact of subjective norms on intentions. In this study, it is assumed that when consumers think that significant others would like them to choose green cosmetics and detergents, then they are more likely to form positive intentions to buy them, and thus, the following hypothesis was set:

H₉. 

Consumption Intentions are positively influenced by Subjective Norms.

The crucial accession of control in TPB refers to a person’s perceptions about circumstances that make it easier or more difficult to adopt a particular behaviour or not [78]. These factors might concern money availability, convenience, time, facilities or barriers relevant to certain choices. There have been some previous studies in cosmetics [72,73] and other subjects [52] in which perceived behavioural control demonstrated an influence on intentions. In this study, it was hypothesised that when consumers believe that there are no barriers for them, they are more likely to choose new, green cosmetics and detergents. They would do so if there is a relevant, attractive opportunity in the market. Accordingly, the following hypothesis was set:

H₁₀. 

Consumption Intentions are positively influenced by Perceived Behavioural Control.

3.5. Scepticism and Media Influence

Furthermore, there have been many previous suggestions that there is a large plexus of variables, some of which might be able to add explanatory power in a theoretical framework of either VBN or TPB [20,48,51,73,75,79]. ElHaffar et al. [14] noted that there are many indications in some review studies about mediation and moderation techniques that are considered valuable tools for revealing significant portions of variance in any given dependent variable. In this study, two variables, namely scepticism [25] and media influence [26], were selected for utilisation as moderators in the combined VBN and TPB framework. According to Kenny [80], a moderator can change the sign or, more often, the strength in a relationship between an independent and a dependent variable. In this study, it was chosen to examine whether scepticism [25] and media influence [26] would indicate a moderation effect in the relationships between consumption intentions and each one of its alleged direct predictors, namely personal norms, attitudes, subjective norms, and perceived behavioural control.

Regarding scepticism, Zarei and Maleki [81] expressed surprise that it has been neglected by both academics and practitioners [82] in investigating consumers’ pro-environmental behaviour. Indeed, it has always been interesting to examine those variables that rather inhibit consumers’ pro-environmental choices [83] given the attitude–intentions/behaviour gap [84,85]. In their qualitative study, Gleim et al. [86] found distrust indications in the “greenness” of organisations and products. However, in their quantitative study, they preferred to follow Soh et al. [87] and examine “trust in ethical advertising”, while they were followed by Osburg et al. [88] in a study about organic and fair-trade products. In fact, a relatively increasing trend to examine trust is observed (e.g., [44,89]). However, trust usually concerns a reversal of the scepticism perspective, i.e., consumers’ positive attitudes towards a supply chain. Scepticism should be viewed as a crucial, powerful, negative barrier regarding preferences for ecological products [81,90]. In practice, scepticism concerns consumers’ reactions to the sight of eco-labels or eco-ads regarding an automatic denial to believe them [25], a type of disbelief possibly generated by ignorance or suspiciousness about greenwashing. Mohr et al. [25] also argued that consumers customarily feel sceptical about ecological products as they might think they should be more expensive and less qualitative and effective. Indeed, Leonidou and Skarmeas [82] found that scepticism can lead to lower assessments of eco-products. Goh and Balaji [91] and Luo et al. [90] found a negative effect of scepticism on consumers’ intentions to prefer green products or services. In this study, the proposed hypothesis is:

H₁₁. 

Scepticism moderates the relationships between Consumption Intentions and each one of its predictors, namely Personal norms, Attitudes, Subjective Norms, and Perceived Behavioural Control.

With regards to the role of media, Stern et al. [33] had previously explained that, among several social interactions, reports and information broadcasted by the mass media should be considered able to formulate beliefs, and thus, they should be included when examining the generation of intentions or behaviour from human values. More recent suggestions have been made that media can strengthen ecological concerns due to relevant information delivery [92] and that in a highly competitive market, the role of advertisement is crucial in driving consumers’ attention [93]. Nonetheless, there have been opposite implications that due to consumers’ scepticism, advertisements do not play a significant role in increasing consumers’ pro-environmental attitudes and behaviour [90]. With relevance to media influence, the following hypothesis was set:

H₁₂. 

Media Influence moderates the relationships between Consumption Intentions and each one of its predictors, namely Personal Norms, Attitudes, Subjective Norms, and Perceived Behavioural Control.

4. Methodology

With regard to data collection, surveys were conducted in four European countries using electronic interviews. The interviews were undertaken by a research agency. A stratified sampling method was adopted [94,95]. Gender and age were used as strata in the samples of the four European countries, with large enough sizes according to the population of each country. The procedure resulted in the following useable questionnaires: France 510, Germany 574, Greece 308, and Spain 454 (in total 1846). The questionnaire (initially developed in English) was translated via the TRAP method (Translation, Review, Adjudication, Pretesting, and Documentation) into French, German, Spanish and Greek.

With regards to the measurement of the variables, for VBN, the following variables were entered in the questionnaire: egoistic values (EV), altruistic values (AV), and biospheric values (BV) with four items each, all adopted from Steg et al. [56] and measured on a 6-point importance scale. It was decided to adopt the first sub-measure of Leiserowitz’s [24] climate change risk perception index, as this particularly focuses on examining people’s perceptions of global warming. This sub-measure includes three items (RiskPer1) and was measured on a 6-point rating scale. Awareness of consequences (AC) had five items, ascription of responsibility (AR) four items, and personal norms (PN) seven items and all were measured on a 6-point Likert scale (no midpoint). The phrasing of the items of the last three variables was based on Steg et al.’s [56] measures, while some of them were modified according to the needs of the specific topic of this study. The decision to utilise a 6-point scale was made in an effort to avoid the respondents’ tendency to choose neutral points.

For TPB, the following variables were included: Attitudes (Att) with five items, measured on a 6-point semantic differential scale; subjective norms (SN), perceived behavioural control (PBC), and consumption intentions (CI), with four items each, were measured on a 6-point Likert scale. The TPB variables were all originally developed for this study. The TACT methodology [96] was followed for CI, while due to the principle of compatibility, all other measures of the theoretical model were developed accordingly. Target is “CPGs (cosmetics and detergents) containing ingredients made from recycled CO₂”, Action is “consumers’ intentions to prefer them”, Context is “any point of sale”, and Time is “when available in the market”. The attentive development stage followed instructions provided by Churchill [97,98], Robinson et al. [99], and Spector [100]. It included preliminary small-scale surveys with student samples in order to collect items for the initial pool, which was purified using item analysis tools (Cronbach’s alpha and item-to-total correlation). The inventory was tested using thorough pilot techniques with experts, academics, and consumers.

Scepticism (Sc) by Mohr et al. [25] consists of six items, for example: “Most environmental claims on package labels or in advertising are intended to mislead rather than to inform consumers”; media influence (MI) by Bearden et al. [26] consists of five items, for example: “I mainly prefer brands that are advertised on TV, radio and the internet”. Both these additional variables were measured on 6-point Likert scales.

Five demographic characteristics were also included in the questionnaire, namely gender, age, level of formal education, annual family income, and occupation.

5. Results

Initially, the data collected from the European consumers were examined to detect potential outliers. The results of the Mahalanobis D²/df test [27] indicated that 12 cases should be excluded from further analyses. Thus, the total sample was 1834 respondents (FR: 503, DE: 570, GR: 308, ES: 453).

5.1. Demographics

Demographic data are presented in

Table 1. Demographic characteristics.

	France		Germany		Greece		Spain
	n	%	n	%	n	%	n	%
Total	503	100	570	100	308	100	453	100
Gender
Men	242	48.9	285	50.0	149	48.4	226	49.9
Women	261	51.1	285	50.0	159	51.6	226	49.9
Other							1	0.2
Age
18–24 years old	61	12.1	67	11.8	36	11.7	60	13.2
25–34 years old	98	19.5	118	20.7	59	19.2	67	14.8
35–44 years old	108	21.5	107	18.8	63	20.5	95	21.0
45–54 years old	104	20.7	101	17.7	73	23.7	95	21.0
55–64 years old	63	12.5	97	17.0	56	18.2	85	18.8
65 years or older	68	13.5	80	14.0	21	6.8	51	11.3
No answer	1	0.2
Education
Primary school	11	2.2	9	1.6	16	5.2	25	5.5
Secondary school	193	38.4	52	9.1	114	37.0	147	32.5
Vocational training	71	14.1	336	58.9	61	19.8	112	24.7
University	126	25.0	116	20.4	80	26.0	106	23.4
Masters’	83	16.5	43	7.5	31	10.1	51	11.3
Ph.D.	10	2.0	9	1.6	6	1.9	12	2.6
No answer	9	1.8	5	0.9
Annual Income
up to 5000 €	31	6.2	23	4.0	49	15.9	21	4.6
between 5001 €–15,000 €	54	10.7	67	11.8	84	27.3	73	16.1
between 15,001 €–25,000 €	94	18.7	78	13.7	81	26.3	132	29.1
between 25,001 €–35,000 €	111	22.1	100	17.5	37	12.0	109	24.1
between 35,001 €–45,000 €	87	17.3	88	15.4	9	2.9	55	12.1
between 45,001 €–55,001 €	55	10.9	67	11.8	6	1.9	22	4.9
55,001 € and more	40	8.0	108	18.9	2	0.6	25	5.5
No answer	31	6.2	39	6.8	40	13.0	16	3.5
Occupation
Professional/Entrepreneur/Farmer	75	14.9	38	6.7	33	10.7	46	10.2
Private employee	103	20.5	251	44.0	99	32.1	125	27.8
Public employee	74	14.7	35	6.1	45	14.6	58	12.8
Unemployed	52	10.3	18	3.2	48	15.6	43	9.5
Houseperson	22	4.4	37	6.5	13	4.2	35	7.7
Retired	88	17.5	111	19.5	31	10.1	58	12.8
Student	34	6.8	32	5.6	26	8.4	49	10.8
Other	45	8.9	37	6.5	7	2.3	32	7.1
No answer	10	2.0	11	1.9	6	1.9	7	1.5

separately for each country. They were tested using χ² with the corresponding populations of the four countries. No statistically significant differences at p < 0.05 were found for gender and age, which were also the stratification variables. Regarding education, however, it should be noted that in the French and Greek samples, tertiary graduates are over-represented to a degree [101].

5.2. Structural Equation Modelling (SEM)

SEM was employed because it is suggested to be the most appropriate technique for simultaneously analysing a series of gridded relationships [27], as mentioned in the Introduction. The analysis was performed with AMOS v20 in two phases: the measurement model analysis and the structural model analysis. In the measurement model analysis, the measurement properties of all variables in the theoretical model were examined simultaneously in terms of unidimensionality and validity. In addition, common method variance (CMV) was also estimated to ensure the absence of common method bias, namely that there is no tendency in the sample to respond with a particular pattern for the majority of questions. The structural model analyses initially tested each theoretical model separately (TPB and VBN) and then their combination simultaneously. In the end, multigroup (with SEM) moderation analyses were performed to test the hypothesised (H₁₁ and H₁₂) impact of the additional marketing variables on the relationships revealed in the previous step.

5.2.1. Measurement Model

The unidimensionality of all constructs was examined in terms of (a) factor loadings, (b) cross-loadings, and (c) error covariances in all the items in the variables. The results led to the elimination of three items due to low factor loadings in the awareness of consequences construct (AC1, AC2, AC5) and two items due to cross-loadings in the attitudes construct (Att1, Att5).

The construct validity of all variables was examined with the assessment of convergent, discriminant and nomological validity [27]. The initial measurement model estimations resulted in the exclusion of egoistic and altruistic values from the final measurement model analysis due to a lack of nomological validity, i.e., the respective hypothesised relationships with Risk Perception were not statistically significant. Therefore, hypotheses H₁ and H₂ were rejected.

Values satisfying the convergent validity criteria for all variables were obtained: (a) factor loadings higher than 0.70 (

Table 2. Measurement model: GOFs, factor loadings, means and reliability.

Goodness of Fit		χ2	df	Sig.	χ2/df	TLI	CFI	RMSEA
		1855.635	518	p < 0.001	3.582	0.971	0.974	0.038
Variables/items, Range, Means, Reliability								Factor Loadings
Biospheric Values (BV) (range: 4–24, Mean: 19.61, α: 0.934, construct reliability: 0.934)
Bio1	Protecting the environment: preserving nature							0.897 ***
Bio2	Preventing pollution							0.897 ***
Bio3	Respecting the earth: live in harmony with other species							0.880 ***
Bio4	Unity with nature: fitting into nature							0.856***
Risk Perception 1 (RiskPer1) (range: 3–18, Mean:13.72, α: 0.914, construct reliability: 0.914)
RP1	How concerned are you about global warming?							0.858 ***
RP2	How serious of a threat do you believe global warming is to nonhuman nature?							0.906 ***
RP3	How serious are the current impacts of global warming around the world?							0.886 ***
Awareness of Consequences (AC) (range; 2–12, Mean; 9.00, α: 0.813, construct reliability: 0.823)
AC3	The exhaustion of fossil fuels is a problem							0.742 ***
AC4	The exhaustion of energy sources is a problem							0.924 ***
Ascription of Responsibility (AR) (range: 4–24, Mean: 16.11, α: 0.895, construct reliability: 0.898)
AR1	I am jointly responsible for CO2 emissions							0.812 ***
AR2	I feel jointly responsible for the exhaustion of energy sources							0.889 ***
AR3	I feel jointly responsible for global warming							0.899 ***
AR4	Not only the government and industry are responsible for high levels of CO2 emissions, but me too							0.708 ***
Personal Norms (PN) (range: 7–42, Mean: 26.54, α: 0.936, construct reliability: 0.933)
PN1	I feel personally obliged to buy CPGs containing green chemical ingredients							0.816 ***
PN2	Regardless of what others do, I feel morally obliged to buy CPGs containing green chemical ingredients							0.804 ***
PN3	I feel guilty when I do not buy CPGs containing green chemical ingredients							0.772 ***
PN4	I feel morally obliged to use ecological products instead of regular products							0.835 ***
PN5	When I buy a new CPG, I feel a moral obligation to prefer one that contains green chemical ingredients							0.882 ***
PN6	People like me should do everything they can to buy CPGs containing green chemical ingredients							0.843 ***
PN7	I would be a better person if I consumed CPGs containing green chemical ingredients							0.758 ***
Attitudes (Att) (range: 3–18, Mean: 12.76, α: 0.876, construct reliability: 0.827)
At2	Undesirable-Desirable							0.924 ***
At3	Unwise (Foolish)/Wise							0.701 ***
At4	Negative/Positive							0.714 ***
Subjective Norms (SN) (range: 4–24, Mean: 14.06, α: 0.929, construct reliability: 0.929)
SN1	My family members think I should buy CPGs containing green chemical ingredients							0.881 ***
SN2	My friends think I should buy CPGs containing green chemical ingredients							0.903 ***
SN3	Important people who influence my behaviour think I should buy CPGs containing green chemical ingredients							0.854 ***
SN4	Persons who are significant to me do buy CPGs containing green chemical ingredients for themselves							0.866 ***
Perceived Behavioural Control (CI) (range: 4–24, Mean: 16.10, α: 0.879, construct reliability: 0.883)
PBC1	Selecting a CPG containing green chemical ingredients is completely up to me.							0.725 ***
PBC2	I am confident that if I want to buy a CPG containing green chemical ingredients, I can buy it.							0.878 ***
PBC3	There are no obstacles for me if I want to select a CPG with green, chemical ingredients							0.839 ***
PBC4	I am confident that I can easily find a CPG containing green chemical ingredients if I want to buy it							0.786 ***
Consumption Intentions (CI) (range: 4–24, Mean: 16.59, α: 0.880, construct reliability: 0.877)
CI1	I will buy CPGs containing green chemical ingredients if they are of similar quality to the regular products							0.766 ***
CI2	I will buy CPGs containing green chemical ingredients if they are of similar price to the regular products							0.730 ***
CI3	I am seriously thinking to buy CPGs containing environmentally friendlier ingredients as soon as I run out of the products I am currently using							0.834 ***
CI4	I will definitely switch to a brand of a CPG that contains green chemical ingredients							0.868 ***

α: Cronbach’s alpha, *** p < 0.001.

), (b) construct reliability (CR) higher than 0.80 indicating exemplary reliability [27,102], and (c) average variance extracted (AVE) higher than 0.50 (

Table 3. Measurement model: validity.

	AVE	Correlations Squared Correlations
		BV	RiskPer1	AC	AR	PN	Att	SN	PBC
Biospheric Values (BV)	0.779
Risk Perception (RiskPer1)	0.781	0.678 *** 0.460
Awareness of Consequences (AC)	0.702	0.481 *** 0.231	0.552 *** 0.305
Ascription of Responsibility (AR)	0.690	0.404 *** 0.163	0.604 *** 0.365	0.443 *** 0.196
Personal Norms (PN)	0.667	0.394 *** 0.155	0.552 *** 0.305	0.401 *** 0.161	0.716 *** 0.513
Attitudes (Att)	0.618	0.396 *** 0.157	0.479 *** 0.229	0.331 *** 0.110	0.453 *** 0.205	0.539 *** 0.291
Subjective Norms (SN)	0.768	0.323 *** 0.104	0.425 *** 0.181	0.301 *** 0.091	0.530 *** 0.281	0.744 *** 0.554	0.539 *** 0.291
Perceived Behav. Control (PBC)	0.655	0.401 *** 0.161	0.456 *** 0.208	0.351 *** 0.123	0.483 *** 0.233	0.607 *** 0.368	0.477 *** 0.228	0.634 *** 0.402
Consumption Intentions (CI)	0.642	0.465 *** 0.216	0.548 *** 0.300	0.404 *** 0.163	0.531 *** 0.282	0.722 *** 0.521	0.607 *** 0.368	0.724 *** 0.524	0.754 *** 0.569

*** p < 0.001.

). All AVE values were higher than the squared correlation of all two-contract combinations, indicating evidence of discriminant validity. Finally, statistically significant, in the hypothesised direction, relationships were indicated for all pairs of variables, leading to a nomological validity assessment (Table 3). These results verified the overall validity of the measurement model.

5.2.2. Common Method Variance

When empirically examining issues related to socially desirable behaviours, it is imperative to check the acquired data for the existence of bias. The common practice in this matter requires examining common method variance (CMV). In the beginning, Harman’s test was carried out by conducting exploratory factor analysis (EFA). The variance extracted in the 1st factor (43.63%) was lower than 50%, indicating the absence of CMV and, for that reason, unbiased data [103].

After that, data were collected in four different countries, Steenkamp and Maydeu-Olivares’s [104] procedure should be performed in order to check whether there were common method effects in the multinational data. Two confirmatory factor analyses (CFA) were conducted, and their goodness-of-fit (GOF) values were compared. In the first, the standard CFA, a multigroup (groups are the four countries) measurement model was run. In the second, the random intercept CFA, all analysis items were loaded in both their respective variable and in an unmeasured common variable. The multigroup standard CFA provided acceptable GOF values (CFI = 0.974, RMSEA = 0.035, SRMR = 0.030). Then, the multigroup random intercept CFA was conducted, and the GOF values (CFI = 0.982, RMSEA = 0.032, SRMR = 0.022) were not substantially improved (ΔCFI ≤ 0.010, ΔRMSEA ≤ 0.015, and ΔSRMR ≤ 0.030). Therefore, it is concluded that CMV is not a serious issue in the multinational data of this study.

5.2.3. Structural Model

The analysis of the structural model included (a) the GOF values to assess the fit of the model to the data, (b) the examination of the structural relationships of the model to test the hypotheses and (c) the assessment of the variance explained in the dependent variable, i.e., consumption intentions.

First, structural models were run for each TPB and VBN model separately (

Table 4. Structural models and moderation.

GOFs	Structural Models			Moderation
	VBN	TPB	Combined	Scepticism				Media Influence
				Unconstrained Model		Constrained Model		Unconstrained Model		Constrained Model
χ2	1520.528	526.507	2901.845	3596.031		3602.362		3723.124		3729.583
sig.	0.000	0.000	0.000	0.000		0.000		0.000		0.000
df	242	80	541	1082		1083		1082		1083
χ2/df	6.283	6.581	5.364	3.324		3.326		3.441		3.444
TLI	0.959	0.971	0.950	0947		0947		0.943		0.943
CFI	0.964	0.978	0.955	0.952		0.952		0.949		0.948
RMSEA	0.054	0.055	0.049	0.036		0.036		0.037		0.037
Paths	Structural relationships (β)			Structural relationships (β)		Critical Ratios	Δχ2 test	Structural relationships (β)		Critical Ratios	Δχ2 test
				Below the Mean	Above the Mean			Below the Mean	Above the Mean
BV→RiskPer1	0.684 ***		0.683 ***	0.736 ***	0.624 ***			0.687 ***	0.705 ***
RiskPer1→AC	0.628 ***		0.597 ***	0.635 ***	0.571 ***			0.553 ***	0.653 ***
AC→AR	0.541 ***		0.382 ***	0.427 ***	0.358 ***			0.343 ***	0.452 ***
AR→PN	0.724 ***		0.435 ***	0.432 ***	0.436 ***			0.418 ***	0.413 ***
PN→CI	0.736 ***		0.252 ***	0.187 ***	0.333 ***	0.986		0.299 ***	0.165 ***	−2.640	6.549
Att→CI		0.222 ***	0.186 ***	0.168 ***	0.211 ***	0.166		0.194 ***	0.140 ***	−1.200
SN→CI		0.318 ***	0.185 ***	0.205 ***	0.159 ***	−1.780		0.186 ***	0.210 ***	0.190
PBC→CI		0.444 ***	0.407 ***	0.460 ***	0.340 ***	−2.603	6.331	0.369 ***	0.487 ***	1.832
SN→AR			0.450 ***	0.324 ***	0.544 ***			0.412 ***	0.480 ***
SN→PN			0.535 ***	0.537 ***	0.539 ***			0.541 ***	0.501 ***
Squared Multiple Correlations (R2)	0.541	0.697	0.711	0.703	0.726			0.696	0.725

β: standardised regression weights, *** p < 0.001.

). The obtained GOF values indicated that both models fit the data well. It is noticed that the square multiple correlation (R²) obtained in CI in the TPB model is considerably higher than the one in the VBN model (Table 4).

Then, the combined model (VBN and TPB) was run several times. Initially, both VBN and TPB predictors, namely PN, Att, SN, and PBC, were modelled to impact CI directly. Although GOF values were acceptable (χ² = 3969.720, df = 543, p < 0.001, χ²/df = 7.311, TLI = 0.928, CFI = 0.935, RMSEA = 0.059), the explanation of CI (R² = 0.694) was a bit lower than the one obtained by the TPB alone (R² = 0.697). This finding possibly indicates interrelationships among the variables of the two models, as has been found in some previous studies [22,41,43]. Indeed, the examination of the modification indices (regression weights) revealed relationships unforeseen in the first run of the combined model. These relationships were the structural paths from SN to AR and PN. Therefore, the relevant arrows were entered in the final run of the combined structural model, which obtained better GOF values and higher squared multiple correlations (R² = 0.711). These results supported hypotheses H₃, H₄, H₅, H₆, H₇, H₈, H₉ and H₁₀.

5.2.4. Moderation

To reveal any moderating effects of scepticism (Sc) and media influence (MI) in the relationships between each of PN, Att, SN, PBC and CI, two multigroup moderation analyses were conducted. The sample was split into two groups for each moderation analysis based on the scores obtained, including one group with respondents who obtained scores below the Mean and one group with those who obtained scores above the Mean, in each of the moderators.

With regard to scepticism, in the multigroup moderation analysis of the unconstrained model (all relationships were free to vary across groups), it was observed that the critical ratio of the difference in the path from PBC to CI was the only one outside the confidence interval (±1.96). The relevant path was constrained to be invariant across the two groups, and the Δχ² test verified that the GOFs of the constrained model were significantly worse than those of the unconstrained model. The latter indicates that the path from PBC to CI is statistically significantly different across the two groups (below and above the Mean in Sc). The same procedure was applied with the moderation analysis of media influence, which revealed that the path from PN to CI is statistically significantly different between the groups of those with scores below the Mean in MI and those with scores above the Mean in MI.

In the relationship between PBC and CI, a higher coefficient was found in the group of respondents who obtained scepticism scores below the Mean. In the relationship between PN and CI, a higher coefficient was found in the group of respondents who obtained media influence scores below the Mean (Table 4).

Therefore, hypotheses H₁₁ and H₁₂ are partially supported.

6. Discussion

The results provide evidence in favour of the proposed theoretical framework of this study. The evidence is considered reliable due to the rigorous methodology followed for variable measurement and sampling. As expected, the full combination of VBN and TPB, and thus the simultaneous consideration of both ethical and rational perspectives, contributed to a broader and deeper understanding of specific consumer intentions. These intentions are complex and multi-layered, as are all pro-environmental behaviours, while they have been unexplored to date. In the combined model, the proportion of variance explained in consumption intentions is higher than in the separate theoretical models. The decision to replace NEP with Climate Change Risk Perception is argued to be a remarkably effective choice in the design of this study. Risk perception indicated the strongest evidence of impact (BV to RiskPer1 and RiskPer1 to AC in Table 4) in the sequence of all relationships in the combined theoretical framework (Figure 1). It is to be noted that the levels of concern about global warming were found to be considerably higher than those of intentions, indicating that—as Leiserowitz [24] had predicted—high levels of risk and threat perceptions do not necessarily translate into environmentally friendly consumer choices.

Only biospheric values demonstrated a statistically significant and positive relationship with Risk Perception, while egoistic and altruistic values were excluded from the measurement model, in line with Ünal et al. [61] and Gkargkavouzi et al. [41]. However, it cannot be claimed that an individual being an egoist or an altruist is unrelated to one’s pro-environmental intentions, especially when there are other recent studies, in which altruistic [43,52,59,88] or egoistic values [20] indicated significant evidence of impact.

Further, the coefficient in the relationship between awareness of consequences and ascription of responsibility is notable, as it is distinctly lower than the other coefficients (Table 4). This highlights the point that the awareness of the harm to the climate due to carbon emissions does not necessarily mean an awareness of our own responsibility for mitigating it. On the other hand, it seems that when someone actually takes responsibility, this can trigger the formation of personal norms (Table 4), which in turn were found to be the direct and strong predictor of consumption intentions when applying VBN alone.

Regarding TPB, it should be firstly underlined that the ability of this model to explain the variance in intentions was found to be much stronger than that of VBN, in line with the results by Gkargkavouzi, et al. [41] and Zhang, et al. [21]. Perceived behavioural control showed stronger evidence of an effect on consumption intentions (PBC to CI), significantly higher than the impact of attitudes and subjective norms, either in the TPB model alone or in the combined model (Table 4). It seems that PBC holds crucial meaning as it essentially expresses consumers’ perceptions about their own confidence and convenience in preferring a green CPG if they want to do so. PBC deserves extensive research attention as a direct or indirect predictor in the grid of variables that lead to intentions, as Odou and Schill [105] have also suggested. Nevertheless, subjective norms are of particular interest, too, as it was found that they can quite strongly influence the last two of the intentions’ antecedents in VBN, namely the ascription of responsibility and personal norms (SN to AR and SN to PN in Table 4). It appears that the opinions and practices of close and important people regarding green CPGs positively and effectively influence respondents. Subjective pressure seems to affect the ascription of responsibility more strongly than awareness of consequences can. It also seems that personal norms are highly influenced by other people’s norms, too.

With regards to the role of the additional variables, it is to be discussed that the causal relationship between perceived behavioural control and intentions is stronger in the group with low scepticism scores, while the causal relationship between personal norms and intentions is stronger in the group with low media influence scores. In other words, for non-sceptics, the barriers or facilitators of finding and purchasing such products has a stronger influence on their intentions to choose them, while sceptics do not seem to care so highly about control. Further, the effect of personal norms (as shaped by beliefs and values) is stronger on the intentions of those who are not much influenced by media compared to those who are more affected by advertising.

7. Limitations and Future Research Suggestions

The variance residuals imply that some other factors must also interfere with the complex procedure of consumers’ decision-making. This is currently acceptable, to an extent, as there are no consumer goods with ingredients containing recycled CO₂ on the market yet. Nonetheless, future research should employ a plethora of variables that might add to our knowledge.

Another limitation of this study is that differences between each country’s population have been neglected. Adding to previous suggestions [48,106] about cross-country examinations, it is argued that future studies should focus on country discrepancies, possibly valorising culture variables, too.

Further, although best practice was followed in translating the questionnaire, there may be some weaknesses in the accuracy of wording from country to country. In future duplications of this theoretical framework, there should be plenty of room for better treatment of wording on some variables (e.g., attitudes that provided rather weak evidence of impact). Besides the acceptable evidence of CMV, there is still a social desirability effect in the responses, and thus, improvements regarding its removal should always be taken care of in any future duplication of this study. Eventually, any relevant study in other European or overseas countries will definitely deepen our understanding of citizens’ intentions to contribute to climate change mitigation through their consumption preferences.

8. Conclusions

The results validated the theoretical framework of this study, which examined VBN and TPB in combination in an effort to obtain a broader and deeper understanding of Europeans’ intentions to buy CPGs that contain ingredients made from recycled CO₂ for the purposes of climate change mitigation. The combined model was modified by climate change risk perception, which was employed to replace NEP, while scepticism and media influence were added to the framework. It was found that TPB is more powerful than VBN in explaining consumption intentions regarding new, innovative green cosmetics and detergents, whereas the combined model was found to be capable of explaining a higher percentage of variance (71.1%) than the separate TPB (69.7%) or VBN (54.1%) models. Perceived behavioural control was found to be the stronger predictor of consumption intentions, with personal norms in second place among all the TPB and VBN variables. Inter-relations among the variables of the two models were estimated, as subjective norms were found to impact the ascription of responsibility and personal norms, too. Moderation analysis indicated that perceived behavioural control over choosing the new green CPGs influences consumption intentions more strongly when scepticism is low, and that personal norms have a stronger impact on consumption intentions when media influence is low.

9. Implications

The results of this study offer grounds for both theoretical and managerial implications. It can be implied that it is fruitful to draw on both moral and practical antecedents in examining pro-environmental intentions and behaviours. It can also be implied that closer to the physical environment antecedents, namely values (biospheric) and attitudes (risk perception), are better drivers of pro-environmental intentions than general pro-environmental beliefs are. It is also implied that consumers’ distrust in advertising (scepticism and low influence of media) deserves attention, maybe in addition to the usual examination of trust [44,88]. Examination of the role of labelling and advertisement are implied able to provide new insights into this subject.

Regarding practical implications, chemical industries of CPGs interested in adopting the new, green ingredients may find useful information about their target group. This will be among consumers, who above all, think that it is easy for them to find and choose cosmetics and detergents containing recycled CO₂. These people hold higher biospheric values, risk perceptions, awareness of consequences, ascription of responsibility and personal norms than others, with the latter two traits being influenced by close and important other people. Marketing communication tactics should be developed accordingly. Careful design of advertising messages is essential to minimise scepticism as well as any negative impact of media on personal norms.

Public authorities have to endorse and implement education, training and research programmes concerning threats to humanity and the overall biosphere due to climate change. The dissemination of information should target increasing Europeans’ biospheric values and building confidence that it is easy and advantageous for consumers to prefer everyday products that reduce carbon emissions and, therefore, contribute to climate change mitigation.

Finally, merging theoretical and practical implications, it is argued that there are interactive and interdependent relationships between the consumers’ social psychological mechanism and practical marketing variables, such as product labelling or mass media advertising. This means that biospheric values, risk perceptions, perceived behavioural control and personal norms (already found able to motivate consumption intentions) are going to increase when the new, green products are actually delivered and easily found in the market, supported by honest, effective promotional techniques, in all types of media, which will be able to minimise scepticism. Taking into consideration that consumers are not capable of (nor are they of course required to be) expertly assessing the ecological footprint of a product, it is highly recommended that the overall communication effort should focus on informing and persuading the audience using messages that are easy to understand and remember.