Do Biospheric Values Moderate the Impact of Information Appeals on Pro-Environmental Behavioral Intentions?

Abstract

Information-based interventions are still the most commonly used intervention approach to motivate people to engage in pro-environmental behaviors. However, recently, researchers assume that the effectiveness of such information depends on the extent to which recipients endorse biospheric values (the so-called ‘values-as-moderator’ hypothesis). If this hypothesis is correct, it would have important implications for the use of information appeals: they could not motivate the entire population, but only those population groups with high biospheric values. Two experimental studies using case vignettes were conducted to investigate the effects of biospheric values prioritization, an informational and a socio-normative appeal, as well as the interaction between biospheric values and those two interventions on two different pro-environmental behavioral intentions: everyday ‘low-cost’ behavior of purchasing (organic) coffee (Study 1) and the more difficult ‘high-cost’ behavior of adopting a heating style that prevents rebound effects (Study 2). In Study 1 we additionally examined the influence of the degree of privacy in which a behavior takes place on the purchase intention. Both studies were conducted as online survey studies in Germany. A total of 800 participants took part in each study. We used moderated logistic regression models to examine the main and interaction effects of biospheric values and the two different appeals on participants’ pro-environmental behavioral intentions. Both studies showed a significant main effect of the informational appeal on pro-environmental behavioral intentions. In Study 1, but not in Study 2, the main effect of the socio-normative appeal was also significant, especially if the purchase decision took place in a public space. However, no consistent evidence of a meaningful effect of biospheric values was found: the main effect of biospheric values was statistically significant only in Study 1, and the postulated interactions between biospheric values and interventions were not statistically significant in either study. Further theoretical and practical implications of the reported results are discussed.

1. Introduction

“Did you know that producing 1 kg of beef causes 13 kg of CO₂ emissions?” Such information is frequently offered in attempts to persuade people to change their behaviors into a more environmentally friendly direction. The rationale for implementing such informational interventions is intuitively appealing: because people will likely fail to act pro-environmentally when they are unaware that their behavior has a detrimental impact on environmental quality, an obvious approach would be to provide information about the behavior’s negative environmental impact [1,2]. Consequently, information-based interventions are still the most frequently used intervention approach to motivate people to change not only environmentally harmful behaviors but also unhealthy or socially problematic behaviors [1,2]. But are information-based interventions actually effective in motivating people to change their behaviors? First research results on this question showed that providing ‘pure’ information alone is not effective [3,4]. According to the existing literature, this finding is not surprising because information has behavioral consequences only if the recipient of the information is motivated to do something with that knowledge [5]. As a consequence of this insight, research subsequently focused on the question of which motivating factors influence the effectiveness of informational interventions on environmentally friendly behaviors/behavioral intentions [1].

Above all, some research findings indicate that personal values are a central motivating factor for environmentally conscious intentions and actions [6,7]. Values reflect relatively stable, general, and desirable goals that are important in and guide peoples’ lives [8]. Values function as standards by which actions, groups, and individuals are evaluated. All basic human values are endorsed, at least to some extent, by all individuals across the world. Yet, individuals differ in how much they endorse and prioritize each value over others. Furthermore, values only indirectly guide peoples’ behaviors [9]. For example, the Value-Belief-Norm (VBN) theory [10] further specifies how values affect behavior via behavior-specific beliefs and norms. In the environmental domain, four types of values appear particularly important: egoistic, hedonic, altruistic, and biospheric values [6,7]. Egoistic and hedonic values focus on goals that primarily benefit oneself: acquiring possessions and status, and pleasure and reducing effort, respectively. Individuals who strongly endorse these two values are typically less likely to engage in pro-environmental actions, because such actions often have some individual costs. On the other hand, altruistic and biospheric values reflect goals that benefit a greater cause than oneself: nature and the environment, or others and society. The more individuals endorse these two values, the more likely they will engage in pro-environmental actions, as such actions typically support these values [6,7]. Particularly, people with strong biospheric values are more likely to be intrinsically motivated to engage in climate actions, as doing so is in line with what they find important. Intrinsically motivated people act without being forced to do so and without receiving external rewards for doing so because they feel good when they act in accordance with their values and experience intrinsic rewards [11,12,13,14]. Because intrinsic motivation comes from within a person, it is self-sustaining and long-lasting, making it a solid source of consistent climate action [6,15].

Based on these findings, it is argued that whether recipients of information appeals consider environmental protection an important personal goal in their lives is crucial for the effectiveness of these appeals (the ‘values-as-moderator hypothesis’) [1]. In other words, it is assumed that the effectiveness of informational appeals is moderated by the extent to which recipients endorse biospheric values (BioVs) [1]. Informational appeals typically explicate the negative environmental consequences of behaviors that people consider convenient and comforting. Following this line of thinking, such impact information should elicit cognitive dissonance if recipients personally care about environmental quality, because it highlights a discrepancy between what recipients freely choose to do and what they personally consider important [1]. However, people who do not consider protecting the environment an important personal goal should not respond to such an informational appeal because they are not experiencing a discrepancy between their values and behavior. Hence, informational appeals should not have an impact on people with weak BioVs. Besides an experimental study conducted by the founders of this assumption themselves [1], we found two other experimental studies providing empirical support for the moderating function of BioVs on the effectiveness of informational appeals. One study found that the strength of BioVs moderates the effectiveness of an informational appeal on participants’ intention to buy bottled water [1]. In another study it was examined whether provision of information on the environmental impact of street lighting affects the acceptability of reduced street lighting levels [16]. As predicted, low lighting levels were seen as more acceptable when information on the environmental impact of street lighting was provided to individuals that strongly endorsed BioVs. Another research group found a moderation effect of BioVs on the power of an informational appeal to save paper [2].

In recent years, the values-as-moderator hypothesis has been extended to socio-normative appeals, that is, appeals that provide information about the BioVs of others. It is assumed that people are more likely to engage in climate actions when they believe other group members strongly endorse BioVs and engage in climate actions. This seems to be the case particularly when members strongly identify with the relevant group and do not strongly endorse BioVs themselves [17,18,19]. These results suggest that communicating that most other group members care deeply about nature and the environment and act accordingly can encourage individuals to take climate action, especially those who are not strongly motivated themselves.

If the values-as-moderator hypothesis is true, it would have important practical implications for the design of appeals to motivate people to show more pro-environmental behavior. In light of this hypothesis, neither informational nor socio-normative appeals would be able to motivate the entire population. Instead, an approach that tailors such appeals to the values prioritized by an individual would be necessary for their persuasive effectiveness: whereas informational appeals should be persuasive for persons with high BioVs prioritization, socio-normative appeals should be persuasive for persons with low BioVs prioritization. Consequently, it was already suggested to use modern technology, for example smart meters, to assess with a short questionnaire people’s value prioritization and then tailor the delivered appeal to their value profiles [2]. However, deriving practical implications with such serious consequences from research results is legitimate only if their validity is based on strong empirical evidence. In our view, however, the empirical evidence for the values-as-moderator hypothesis is currently not very strong. There are only a handful of experimental studies, mainly conducted by one research group from the Netherlands, that aim to explain very simple and low-cost behaviors, such as using bottled water, saving paper, or accepting reduced street lighting. Furthermore, the assumption derived from the values-as-moderator hypothesis that neither informational nor socio-normative appeals are able to motivate the entire population does not match the results of meta-analyses, which point to the general effectiveness of informational and socio-normative appeals: a meta-analysis summarized the results of 92 studies that evaluated the effectiveness of information interventions appealing to the environmental protection motive (e.g., goal setting, self-contrast feedback, environmental information, and public commitment) in influencing energy-saving behaviors [20]. Overall, an average mean effect size of d = +0.50 was determined for these appeals. Based on the results of 91 studies evaluating the effectiveness of socio-normative appeals, an overall average small-to-medium effect size was found (d = +0.32) [21]. These meta-analytic results provide more support for the view that practitioners might use informational and socio-normative measures as effective interventions without considering people’s value prioritization.

In the following, we report the results of two studies aimed at replicating the test of the values-as-moderator hypothesis in two different contexts: the everyday ‘low-cost’ behavior of purchasing (organic) coffee and the more difficult ‘high-cost’ behavior of adopting a heating style that prevents rebound effects after the energetic renovation of an apartment. For each behavioral domain, we developed an informational appeal highlighting the negative environmental consequences of buying conventional coffee or an energy-wasting heating method versus the positive environmental consequences of buying organic coffee or an energy-saving heating method and compared the effect of these appeals to a no-appeal control condition (CG). Furthermore, we also developed a socio-normative appeal inducing participants to imagine that they work in a very environmentally conscious company, where the head of the company and all the fellow workers put a lot of effort into limiting their negative impact on the environment. We compared the effects of the socio-normative appeal with a CG without explicit socio-normative information. Additionally, we asked participants in Study 1 to imagine the target behavior (purchasing coffee) either for their own household vs. for breaks with colleagues at work. This third intervention element was motivated by the assumption that the private versus public nature of the target behavior may be a second moderator of the impact of socio-normative appeals: if neither the enactment nor the consequences of a behavior are known to others, social norms may lose their motivational power [22,23,24]. Consequently, publicness increases, whereas privateness decreases the effect of socio-normative appeals. By randomly assigning participants to experimental conditions using quotas that ensured that an equal number of high BioVs and low BioVs participants were assigned to each appeal condition, we were able to compare the effects of value-congruent (information appeals for participants with high BioVs and socio-normative appeals for participants with low BioVs) and value-incongruent (informational appeals for participants with low BioVs and socio-normative appeals for participants with high BioVs) appeals. In both studies, the importance that participants assigned to BioVs was assessed using a short scale that we explain in the Methods Section of this paper. The novelty of our studies is to test independently and with a sophisticated experimental design the circulating assumptions about the massive impact of BioVs on the effectiveness of interventions to promote pro-environmental behaviors, and furthermore to examine the effects of the level of publicity of an action. In more detail our studies aim to provide an experimental test of the following six hypotheses:

H1: 

Participants’ BioVs are significantly associated with their pro-environmental behavioral intention: the higher their BioVs, the stronger is their pro-environmental behavior intention [6,7].

H2: 

In line with the reported meta-analytical results, we hypothesized that the intervention of an ‘informational appeal’ would have an independent impact on participants’ pro-environmental behavioral intention in comparison with the CG (main effect of environmental appeal) [20].

H3: 

In line with the reported meta-analytical results, we hypothesized that an intervention providing socio-normative information would have an independent impact on participants’ pro-environmental behavioral intention in comparison with the CG (main effect of socio-normative information) [21].

H4: 

In line with the values-as-moderator hypothesis, for participants with high BioVs prioritization the informational appeal should have a stronger effect on participants’ intention to choose the more pro-environmental behavioral intention (interaction effect of BioVs importance and the informational appeal in comparison with the CG) [1,2,16].

H5: 

In line with the values-as-moderator hypothesis, for participants with low BioVs prioritization the socio-normative appeal should have a stronger effect on participants’ intention to choose the more pro-environmental behavioral intention (interaction effect of BioVs importance and the socio-normative appeal in comparison with the CG) [17,18,19].

H6 (only tested in Study 1):

The effect of an intervention providing socio-normative information is stronger in the public than in the private context (interaction effect of context and appeal type) [22,23,24].

2. Study 1—Testing our Hypotheses in the Domain of Purchasing a Conventional Versus Organic Coffee Brand

Study 1 was conducted in April 2021 as an online study. Pro-environmental behavioral intention was operationalized as the decision to purchase either an organic or a conventional coffee brand. We examined how this ‘low-cost’ daily-life decision would be affected by participants’ personal BioVs, an informational appeal, and a socio-normative appeal. Our main theoretical interest lay in the interaction effects between these intervention elements and individual BioVs expression. Further, to test the assumption that a public versus private context is a second moderator of the impact of the socio-normative appeal, we instructed participants to make two coffee choices, one for the private context of their home and one for the public context of the company they should imagine they work for.

2.1. Materials and Methods

2.1.1. Participants and Procedure

A German panel provider invited participants over 18 years of age who are registered in their database to take part in the online survey. Only those participants who declared to drink coffee at least sometimes were forwarded to the survey. To ensure data quality, each participant had to pass two control items, one attention-check item (“To guarantee that you are reading the items with full attention, please tick 7 [1] in this line”), and one condition-specific multiple-choice item to check if the provided information was received correctly (for example, in the socio-normative intervention and the corresponding CG the participants were asked: “What facts have been given about the company in which you are employed? Please list at least two facts”). This led to the exclusion of data from 234 participants who failed to pass at least one of the two items. We aimed to have 800 participants, 200 in each subsample. We included data from 804 participants in the analysis. Quota functions were used in the survey programming to ensure that the proportion of men and women in each condition was equal and to ensure that the same number of participants with low and high BioVs expressions would be assigned to each experimental condition. Due to a mistake in the defined rules for the quota specification, the subsamples did not turn out as evenly as intended. The average age in the final sample was 55 years, with a range of 18 to 91 years.

At the beginning of the survey, all participants were assigned to either the high or low BioVs group according to their individual BioVs expression. Later, all participants read the following introduction: “Please imagine you are in the following situation: You are an employee in a company. In the team you work with it is customary that each person brings a packet of coffee once a quarter. Your coffee at home is also used up, so you go to the supermarket in the afternoon to run your coffee errands”.

Subsequently, participants in the high and low BioVs groups were randomly and evenly assigned to one of the interventions or CGs (see

Table 1. Number of participants in the experimental conditions of Study 1.

	Environ Appeal	Socio-Norm Appeal	Environ Control	Socio-Norm Control	Total
High BioVs	84	87	117	101	389
Low BioVs	85	86	128	116	415
N	169	173	245	217	804

Note. BioVs = biospheric values, environ = environmental, socio-norm = socio-normative.

). Depending on the assigned condition, the participants got to read additional information. Subsequently, they were asked to answer twice to a coffee-choice item. In the last step participants answered a specific manipulation check for either the informational appeal and informational CG or the socio-normative appeal and socio-normative CG. Several other constructs were measured that are beyond the scope of this paper, therefore we do not explain them in detail.

2.1.2. Assessment of Biospheric Values Importance

Participants completed the four-item BioVs subscale of the E-PVQ [25] (item example: “It is important to the person to protect the environment”). Participants indicated on a 7-point response scale (1 = not like me; 7 = very much like me) how similar to themselves they perceived each person portrayed by the E-PVQ. The internal consistency of the four E-PVQ items was high (Cronbach’s α = 0.91), and items were averaged into a BioVs score (M = 5.54, SD = 1.20). The cut-off value of mean = 6.00 was used with the intention to assign participants as equally as possible either to the group of high BioVs or to the group of low BioVs. We based the cut-off on the European Social Survey 2018 Round 9 [26], as it contains the only representative data on BioVs in German society that we could find through a literature search [27].

In the frame of the European Social Survey the mean of the German sample was 5 on a 6-point Likert scale. Because we wanted to maintain the 7-point Likert scale used in the E-PVQ while ensuring that equal numbers of participants with low and high BioVs would participate in the study, we set a cut-off at 6, also the second-highest point in the scale used.

2.1.3. Experimental Conditions

Within the high and low BioVs groups, participants were randomly assigned to one of the following four experimental conditions: informational appeal, informational control, socio-normative appeal, or socio-normative control.

Participants in the informational appeal condition were instructed to imagine that next to the coffee shelf in the supermarket two posters were placed. They were asked to read the posters with their full attention. One poster highlighted the massive negative impact of conventional coffee farming on nature, the other one highlighted the positive impact that the choice for organic coffee might have on the environment. Both posters are documented in the Supplementary File in Figure S1. Participants assigned to the informational CG did not receive any additional information besides the case vignette.

As a manipulation check, participants assigned to the informational appeal or informational CG completed three items measuring their problem awareness. The items to measure problem awareness [28] were adapted to the context of conventional coffee (for example: “Conventional (non-organic) coffee cultivation is associated with a large environmental impact”). Response options ranged from 1 = I totally agree to 7 = I do not agree at all. Cronbach’s α was 0.89, and the items were averaged to form a problem awareness score (M = 4.42, SD = 1.40).

Participants in the socio-normative intervention condition were instructed to imagine that the company they work for was characterized by a strong environmental social norm. For this purpose, they read the following paragraph: “The company where you are employed and for which you want to buy one of the coffee packs is known to attach great importance to environmental protection: the company building is completely supplied with green electricity, a fixed percentage of the annual profit is donated to climate protection projects, and there are vegetarian and vegan options in the canteen every day. For years, you yourself have been an important member of a team that received an internal company award for its special commitment to environmental protection”.

Participants in the socio-normative CG were not instructed to imagine that the company they work for has any special interest in environmental protection. For this purpose, they read the following paragraph: “The company where you are employed and for which you want to buy one of the coffee packs is very large. The company building is located near the city center. The company is easily accessible both by car and by public transport. In the canteen there is a wide range of meals every day. You yourself have been working in a team for years that last month received an internal company award for its special creativity”.

As a manipulation check, participants assigned to the socio-normative appeal or the socio-normative CG were asked to answer three items measuring the perceived environmental social norm of the imagined firm. The items [29] were adapted to the context of organic coffee (for example: “My fellow workers would expect me to buy organic coffee”). Response options ranged from 1 = I totally agree to 7 = I do not agree at all. Cronbach’s α for the scale was 0.87 (M = 5.09, SD = 1.88).

2.1.4. Dependent Variable: Choice of a Coffee Brand

Participants were asked to choose one coffee brand from a picture showing seven different coffee brands that are popular in Germany with additional price information. Four of the seven presented brands were conventional coffees, three brands were organic coffees. The picture with the seven coffee brands was presented twice to each participant, first with the instruction to choose a coffee for their own household (coffee choice private) and second with the instruction to choose a coffee for the company (coffee choice public). This manipulation aims to introduce a private vs. public choice context. Prior to data analysis participants’ chosen coffee brands were dichotomized. The decision for an organic coffee brand was evaluated as pro-environmental behavior. The picture presenting the seven coffee brands is documented in the Supplementary File in Figure S3.

2.2. Results

Most of the analyses were conducted with IBM SPSS Version 29.0.0.0, and the diagrams were created in Excel. R Version 4.3.0 [30] was used to calculate the sensitivity analyses and to cross-check all other conducted analyses.

2.2.1. Manipulation Check

In the informational appeal condition the problem awareness scores were significantly higher (M = 4.87, SD =1.39) than in the informational CG (M = 4.12, SD = 1.32), t(412) = −5.54, p < 0.001, d = 0.55. In the socio-normative appeal condition the perceived social norm scores were significantly higher (M = 5.09, SD = 1.40) than in the socio-normative CG (M = 3.04, SD = 1.78), t(388) = −12.68, p < 0.001, d = 1.29. These results show that both interventions clearly worked as intended.

2.2.2. Test of Hypothesis

For testing H1, which states that the importance participants assign to BioVs is associated with their pro-environmental behavior intentions, we conducted two binominal logistic regression analyses (one for the private and one for the public coffee-choice context) with the dichotomized coffee brand choice as dependent variables (0 = conventional coffee, 1 = organic coffee) and the participants’ continuous BioVs expression as a predictor. As shown in

Table 2. Logistic regression analyses to predict the coffee choice from biospheric values in Study 1.

Model	Exp(b)	df	Wald	95%-C.I.	d *	RN²	df	χ2	p
DV: Private						0.045	1	26.19	<0.001
~BioVs	1.41	1	23.97	1.231, 1.624	0.19
DV: Public						0.046	1	27.662	<0.001
~BioVs	1.41	1	25.56	1.231, 1.603	0.19

Note. N = 804; private = coffee choice private; public = coffee choice public; BioVs = biospheric values; * d = odds ratio (Exp(b)) transformed to Cohens d, using an effect size calculator [31]; R_N² = Nagelkerke R².

, in the private context as well as the public context the importance that participants assigned to BioVs was significantly associated with choosing an organic coffee brand. In both contexts the effect size was d = 0.19. Thus, Study 1 supports H1.

To test H2 (the main effect of informational appeal) and H3 (the main effect of socio-normative appeal), we conducted four binominal logistic regression analyses with the dichotomized coffee brand choices (two analyses for the private and two for the public context) as dependent variables and each of the two interventions (versus their respective CG) as independent predictors. As shown in Figure 1 and

Table 3. Logistic regression analyses to predict the coffee choice from the interventions in Study 1.

Model	Exp(b)	df	Wald	95%-C.I.	d *	RN2	df	χ2	p
DV: Private						0.116	1	36.66	<0.001
~info	3.59	1	35.22	2.352, 5.469	0.70
DV: Public						0.101	1	31.03	<0.001
~info	3.34	1	29.95	2.168, 5.142	0.66
DV: Private						0.079	1	22.03	<0.001
~soc	2.96	1	21.17	1.863, 4.690	0.60
DV: Public						0.313	1	103.17	<0.001
~soc	9.42	1	89.00	5.909, 15.003	1.24

Note. Private = coffee choice private; public = coffee choice public; info = informational appeal vs. informational control (N = 414); soc = socio-normative appeal vs. socio-normative control (N = 390); * d = odds ratio (Exp(b)) transformed to Cohens d, using an effect size calculator [31]; R_N² = Nagelkerke R².

, in both choice contexts the informational appeal as well as the socio-normative appeal were significant independent predictors of choosing an organic coffee brand.

For the informational appeal in the private and public context as well as for the socio-normative appeal in the private context the respective effect sizes were similar (ds = 0.60 to 0.70). However, as postulated by H6, the socio-normative appeal’s effect size was much larger in the public context (d = 1.24) than in the private context (d = 0.60). To compare the effectiveness of the interventions in the two choice contexts, we conducted a mixed-model 2 × 2 × 2 ANOVA with the context (private vs. public) as a within-subjects factor and two between-subjects factors: Factor 1, assignment to either the information condition (information appeal and information appeal CG) or the socio-normative condition (socio normative appeal and socio-normative CG), and Factor 2, assignment to either one of the interventions (information appeal and socio-normative appeal) or one of the CGs (information appeal CG or socio-normative appeal CG).

The 2 × 2 × 2 ANOVA results showed a significant interaction between the context and Factor 1, F(1, 800) = 45.82, p < 0.001, η² = 0.05, between the context and Factor 2, F(1, 800) = 117.93, p < 0.001, η² = 0.13, and a significant three-way interaction, F(1, 800) = 40.27, p < 0.001, partial η² = 0.05. The significant three-way interaction indicates that the context indeed had an effect on the effectiveness of the interventions: as shown in Figure 1, the informational intervention affected behavioral choices more strongly in the private than in the public context, and the socio-normative appeal affected behavioral choices more strongly in the public than in the private context. Taken together, these results clearly support H6.

In the last step, we tested H3 and H4, which were derived from the values-as-moderator hypothesis. H3 and H4 postulate an interaction effect of participants’ value profiles (high vs. low BioVs importance) and each of the intervention types, with opposite signs: high (vs. low) BioVs scores should increase the effectiveness of the informational appeal (H3) but decrease the effectiveness of the socio-normative appeal (H4). To test H3 and H4, we conducted two moderated binominal logistic regression analyses within each context (private and public) with the informational appeal vs. the informational CG and the socio-normative appeal vs. the socio-normative CG as independent predictors and the BioVs expression as moderator, using SPSS Process Model 1 [32]. All variables were mean-centered prior to analyses. As shown in

Table 4. Moderation analyses to test the ‘values-as-moderator’ hypothesis in Study 1.

Model	Exp(b)	df	Z	95%-C.I.	p	RN2	df	χ2	p
DV: Private						0.17	3	56.80	<0.001
~Info × BioVs	1.19	3	0.86	−0.2278, 0.5846	0.39
DV: Public						0.16	3	51.19	<0.001
~Info × BioVs	1.47	3	1.79	−0.0365, 0.8069	0.07
DV: Private						0.11	3	32.24	<0.001
~Socio × BioVs	1.01	3	0.06	−0.4099, 0.4380	0.95
DV: Public						0.16	3	51.19	<0.001
~Socio × BioVs	0.76	3	0.23	−0.7174, 0.1774	0.25

Note. Private = coffee choice private; public = coffee choice public; info = informational appeal vs. informational control (N = 414); socio = socio-normative intervention vs. socio-normative control (N = 390); BioVs = biospheric values; R_N² = Nagelkerke R².

, none of the interaction terms between the intervention types and the BioVs group were statistically significant. Thus, Study 1 does not support either H3 or H4, i.e., it does not provide evidence for the values-as-moderator hypothesis. We conducted simulation-based sensitivity analyses with R based on 5000 replications for each of the models including non-significant interaction effects. The corresponding data and software code including a detailed explanation can be found on https://github.com/peterpuetz2020/anicker_et_al (accessed on 29 March 2024). Using a required power of 0.80 and an alpha of 0.05, the minimum detectable effect size ranged from an odds ratio of 1.9 for the informational appeal vs. the informational CG on the coffee choice in the private context to 2.2 for the socio-normative appeal vs. the socio-normative CG on the coffee choice in the private context. We consider an odds ratio of 1.5 as practically relevant. Such an odds ratio means that the effect of the intervention, measured in odds, increases by 50% if BioVs increase by one unit. On a probability scale, for example, this means an increase in the intervention effect from 30 percentage points to 41 percentage points. As we find no significant interaction effects and a practically relevant effect is below the minimum detectable effect sizes, our study lacks power to conclude that there is no relevant interaction effect.

2.3. Discussion

The aim of Study 1 was to test six hypotheses derived from the literature on the effectiveness of informational appeals to increase peoples’ pro-environmental behavior intentions. Like many studies before [6,7,15,16,33], our results support the role of BioVs as an independent direct predictor of organic coffee purchases (H1). Study 1 also supports H2 and H3, which postulate that the informational and socio-normative appeals have a main effect on participants’ pro-environmental behavior intention. The impact of the informational appeal was somewhat stronger in the private context and the impact of the socio-normative appeal was much stronger in the public context. This finding supports H6 and, in line with the literature [22,23], shows that greater privacy may decrease the effect of socio-normative interventions. Of special theoretical interest is the finding that Study 1 provides no support for H4 and H5, which were derived from the values-as-moderator hypothesis: neither the effectiveness of the environmental appeal nor that of the socio-normative appeal was moderated by the importance people assign to BioVs. Given the fact that the minimum detectable effect sizes according to the sensitivity analyses were higher than the practically relevant effect sizes, the lack of a significant effect in respect to our moderation hypotheses cannot be interpreted as evidence in favor of the null hypothesis.

3. Study 2—Testing Our Hypotheses in the Domain of Energy-Saving Heating Styles

To test whether our findings are valid in a different pro-environmental behavior domain, we replicated Study 1 within the area of individual heating behavior. Other than shopping for groceries, heating behavior takes place in the private area of one’s own home, so this behavior can be seen as a more private behavior. Additionally, because an energy-efficient heating style entails consistently performing a set of behaviors (e.g., ventilating, responding to temperature changes or being away), this behavior is significantly more effortful than the simple act of purchasing organic coffee examined in Study 1 and may therefore be characterized as a high-cost behavior. Study 2 was conducted in April 2021 as an online study. The target variable consisted of participants’ intention to perform a heating style that prevents rebound effects after the energetic renovation of an apartment, that is, the conscious adaptation of one’s heating behaviors to the new situation. Rebound stands for the empirical observation that people’s everyday usage behavior often fails to make use of the full energy-saving potential of modern technology [34]. Studies in the field of energy-efficient building renovations [35] show that the technologically possible energy savings are not fully exploited due to such rebound effects: after the energetic renovation of a building, people often increase the average room temperature, ventilate by opening the windows for a long time, and use the air conditioning more often. To test whether Hypotheses H1 to H5 are supported in this context, Study 2 follows the procedures used in Study 1, but with one exception: Study 2 examines the effect of an informational appeal and a socio-normative appeal only in a private context, because we considered heating as a private behavior in general. Thus, Study 2 does not test H6.

3.1. Materials and Methods

3.1.1. Participants and Procedure

As part of Study 2, a German panel provider invited participants over the age of 18 who are registered in its database to take part in the online survey. Only those participants who declared to live in a rented flat or house were forwarded. As in Study 1, each participant had to pass two control items; this led to the exclusion of 191 cases. We aimed to include data from 800 participants, with 200 in each subsample. Of the 806 participants whose data were included in the data analysis, 50.6% were female and the median age was 53 years, with a range from 18 to 85 years. The same procedure as in Study 1 was used to assign participants to a high vs. low BioVs group. The number of participants per condition is shown in

Table 5. Number of participants in the experimental conditions in Study 2.

	Informational Appeal	Socio-Normative Intervention	Informational Control	Socio-Normative Control	Total
High BioVs	77	85	97	88	347
Low BioVs	110	102	129	118	459
N	187	187	226	206	806

Note. BioVs = biospheric values score.

. We specified quotas in the survey Qualtrics software, Version May 2022 of Qualtrics, copyright © 2024 Qualtrics, to ensure that the same number of participants with low and high BioVs expressions would be assigned to each experimental condition. Due to a mistake in the defined rules for the quota specification, the subsamples did not turn out as evenly as intended.

All participants read a vignette that describes the living situation in a rented apartment with very poor thermal insulation on the outside walls, a technically outdated heating system, and drafty windows. It was pointed out that, due to the poor energetic condition of the apartment, heating costs are very high—around 900 Euros per winter. Participants were asked to imagine themselves in this dwelling situation and then to report their intention to perform a list of 12 heating energy-saving behaviors (see

Table 6. Means of behavior intentions regarding energy-saving in Study 2.

Items	Pre-Renovation		Post-Renovation
	M	SD	M	SD
Investing 20 Euros in an easily attachable door seal to avoid draughts from the entrance area of the apartment.	5.93	1.49	5.23	1.90
Investing 10 Euros in a film that can be easily attached behind the radiator, which ensures that the heat radiates into the room and not into the wall.	5.75	1.68	4.86	2.12
Investing 50 Euros in electronic thermostats for the entire apartment, so that the temperature in the rooms is automatically controlled.	4.29	2.06	4.18	2.17
Ventilating the apartment twice a day with 10 min of shock ventilation (wide open window), instead of airing for a long time with the window in the tilt position.	6.24	1.25	6.19	1.29
If it is very cold outside, tilting the window longer rather than (briefly) shock ventilate it (reverse-coded).	1.92	1.53	1.93	1.60
Keeping the doors between the rooms closed so that heat is not lost.	5.59	1.61	5.25	1.76
Making sure that the windows are not open when the heating is turned up high.	6.49	1.03	6.38	1.12
Dressing warmer instead of turning up the heat further.	4.64	1.92	3.95	1.92
Tilting the window in the bedroom during the day because of the good air, even if the heating is running there (reverse-coded).	1.49	1.10	1.68	1.39
Setting the heating to a comfortable temperature so you don’t have to wear a thick sweater indoors in winter (reverse-coded).	4.13	1.75	4.48	1.79
Making sure that the radiators are not obstructed by furniture.	6.43	0.99	6.29	1.20
Venting the radiators if there are signs of air in them.	6.22	1.33	6.24	1.33

Notes. N = 806; the response scale went from 1 (low intention) to 7 (high intention).

).

All participants then read a second vignette, which described the living situation in the same apartment after a comprehensive energy-saving renovation. The high costs from the first vignette were highlighted again and then the improvements were listed: the heating system was now modern and efficient, the exterior walls were thermally insulated, and triple-glazed windows had been installed. And best of all: after the energetic renovation, the rent had hardly increased. Then it was pointed out that, due to the energetic renovation, the heating costs of the apartment had fallen to less than half of the previous costs, so that the participants saved more than 500 Euros per year. The complete text of the pre-renovated and renovated flat vignettes is documented in the Supplementary File, Figures S4 and S5.

3.1.2. Assessment of Biospheric Values Importance

To assess BioVs, the same four E-PVQ items [25] as in Study 1 were used. In Study 2 the internal consistency of these four E-PVQ items was again high (Cronbach’s α = 0.91), so they were averaged into a BioVs index (M = 5.43; SD = 1.25). Following the findings from the ESS (2018) regarding the relatively high BioVs expression in the German population, we decided again to use the cut-off value mean = 6.00 to assign participants to the high vs. low BioVs groups. In comparison with the reported mean, it turned out that the cut-off value may have been set too high for this sample. Implications will be discussed below in the Limitations Section of the General Discussion.

3.1.3. Experimental Conditions and Manipulation Checks

Within the high and low BioVs groups, participants were randomly assigned to one of four experimental conditions: informational appeal, informational control, socio-normative appeal, or socio-normative control. The informational appeal consisted of a fictitious letter from the property owner, which participants imagined receiving, after the energy-efficient renovation of their apartment. The letter informed about the role of heating as a main consumer of energy. It stressed that the adoption of heating energy-saving behavior was one of the most impactful behaviors individuals could adopt to contribute to environmental protection. The letter concluded with five recommendations regarding which simple behaviors and cheap investments could save up to 36% of the heating energy used every day. The complete text of the letter is documented in the Supplementary File, and Figures S6 and S7. Participants assigned to the informational CG did not get any additional information besides the two case vignettes. Participants in the socio-normative appeal condition were instructed to imagine that the company they work for was characterized by a strong environmental social norm. They read almost the same instructions as in Study 1, except that the sentence mentioning the intention to buy coffee was removed. The paragraph began with the following sentence: “The company where you are employed is known to attach great importance to environmental protection: (…)”, followed by the same details as in Study 1.

Participants in the socio-normative CG were not instructed to imagine that the company they work for has any special interest in environmental protection. For this purpose, they read the same paragraph as did the participants in the socio-normative CG in Study 1, except that the sentence mentioning the intention to buy coffee was removed.

Depending on their assigned condition, participants completed a manipulation check that consisted of similar items as in Study 1, however, adapted to the context of heating energy-saving. In the informational appeal conditions, three items assessed problem awareness (e.g., “Energy consumption for heating is associated with a large environmental impact”); they showed high internal consistency (alpha = 0.91) and were averaged to form a problem-awareness index (M = 4.83, SD = 1.24). In the socio-normative conditions, three items assessed perceived social norms (e.g., “My colleagues would expect me to save heating energy where possible”); they showed high internal consistency (alpha = 0.83) and were averaged to form a social-norm index (M = 4.40, SD = 0.89).

3.1.4. Dependent Variable ‘Reduction of Heating Energy-related Rebound Effect’

After reading the intervention or control text, all participants again reported their intention to perform each of the 12 heating energy-saving behaviors, this time imagining being in the new dwelling situation. This ‘double-vignette’ method was developed and tested in earlier research work [36]. We operationalized the heating-related rebound effect (the dependent variable) by subtracting the mean across the 12 energy-saving behavioral intentions completed post-renovation from the mean across the same 12 behavioral intentions completed pre-renovation. The resulting difference score is used as an indicator of the heating-related rebound effect on an intentional level. A higher difference score represents a stronger rebound effect.

3.2. Results

Most of the analyses were conducted with IBM SPSS Version 29.0.0.0, and the diagrams were created in Excel. R Version 4.3.0 [30] was used to calculate the sensitivity analyses and to cross-check all other conducted analyses.

3.2.1. Rebound Effects

Table 6 shows the means and standard deviations of the 12 items used after reading the pre-renovated-flat and renovated-flat vignette for assessing participants’ intention to perform heating energy-saving behavior. The pre-renovation Cronbach’s α of the 12-item scale was 0.73, and the post-renovation Cronbach’s α was 0.75. After the renovation of the flat, participants reported a significantly lower intention to heat in an environmentally friendly way than before the renovation (Mpre = 5.67, SDpre = 0.76, Mpost = 5.37, SDpost = 0.86), t(802) = 14.27, p < 0.001; d = 0.58). Thus, as expected, an overall rebound effect was observed.

Table 7. Means of the heating-related rebound effect by the experimental condition in Study 2.

Condition	M	SD
Informational appeal	0.18	0.55
Informational control condition	0.41	0.64
Socio-normative intervention	0.30	0.52
Socio-normative control condition	0.26	0.59

presents the mean heating-related rebound effect by the experimental condition.

3.2.2. Manipulation Check

Unexpectedly, the mean of the problem awareness index was not significantly higher in the informational appeal condition (M = 4.90, SD = 1.19) than in the informational CG (M = 4.78, SD_CG = 1.28), t₍₄₁₁₎ = 0.95, p = 0.17. Thus, reading the informational appeal did not increase participants’ problem awareness. As intended, however, the mean of the social-norm index was higher in the socio-normative appeal condition (M = 4.54, SD = 0.86) than in the socio-normative CG (M = 4.27, SD = 0.89), t₍₃₉₁₎ = 3.15, p < 0.002, d = 0.32.

3.2.3. Test of Hypotheses

To test H1, which stated that the importance a person assigns to BioVs is a direct independent predictor of the intention to perform an energy-saving heating style, we conducted a linear regression analysis with the heating-related rebound effect as the dependent variable and the continuous BioVs scores as a predictor. As shown in

Table 8. Linear regression analyses to predict the heating-related rebound effect in Study 2.

Model	b	β	T	95%-C.I.	d	RN2	df	F	p
DV: HRE						0.00	1	0.28	0.60
~BioVs	−0.01	−0.02	−0.53	−0.041, 0.024	-
DV: HRE						0.03	1	14.1	<0.001
~Info	−0.22	−0.18	−3.76	−0.340, −0.106	0.37
DV: HRE						0.00	1	0.54	0.45
~Socio	0.04	0.04	0.76	−0.068, 0.153	-

Note. Info = informational appeal vs. informational control (dummy coded as 0 and 1, N = 413); socio = socio-normative intervention vs. socio-normative control (dummy coded as 0 and 1, N = 393); BioVs = biospheric values; HRE = heating-related rebound effect.

, Study 2 does not support H1: even though the β coefficient −0.02 indicates that participants with higher BioVs tended to show lower rebound effects than participants with lower BioVs, this effect was not significant, p = 0.60. Thus, the BioVs score did not have a significant direct impact on the intention to perform a rebound-preventing heating style.

For testing whether the informational (H2) as well as the socio-normative intervention (H3) would significantly decrease the heating-related rebound effect, we also conducted linear regression analyses (Table 8). These showed that the informational appeal significantly decreased the reported rebound effect as predicted, β = −0.18. With d = 0.37, this effect may be characterized as small to medium-sized. The results of Study 2 thus support H2. However, the socio-normative intervention had no significant effect, and its regression coefficient unexpectedly had a positive sign, β = +0.04. Thus, the results of Study 2 do not support H3.

In a last step, to test again H4 and H5, which represent the values-as-moderator hypothesis, we conducted two moderated regression analyses, one for the informational appeal condition and one for the socio-normative intervention condition (versus their respective CGs). We used SPSS Process [32] Model 1 with the reported rebound effect as the dependent variable, assignment to the intervention vs. CG as a dummy-coded predictor, and BioVs scores as a moderator. All variables that defined products were mean-centered prior to analyses. As

Table 9. Moderation analyses to test the ‘values-as-moderator’ hypothesis in Study 2.

Model	b	t	95%-C.I.	p
DV: Rebound
~Info × BioVs	−0.03	−0.55	−0.1189, 0.0671	0.58
DV: Rebound
~Socio × BioVs	−0.01	−0.17	−0.0968, 0.0811	0.86

Note. Info = informational appeal vs. informational control (dummy coded as 0 and 1, N = 413); socio = socio-normative intervention vs. socio-normative control (dummy coded as 0 and 1, N = 393); BioVs = biospheric values.

shows, neither of the interaction terms between intervention and BioVs was significant in these analyses. Thus, Study 2 also does not provide evidence for the impact of BioVs priority on the effectiveness of an environmental and socio-normative appeal, as postulated by the values-as-moderator hypothesis. To test the meaningfulness of these non-significant interaction effects, we conducted simulation-based sensitivity analyses with R based on 5000 replications for both models. The corresponding data and software code including a detailed explanation can be found on https://github.com/peterpuetz2020/anicker_et_al (accessed on 29 March 2024). Using a required power of 0.80 and an alpha of 0.05, the minimum detectable effect size was about 0.13 for the interaction in both models. We consider an effect size of 0.05 as practically relevant, which would mean that the effect of the respective intervention on the heating-related rebound effect increases by 0.05 if BioVs increase by one unit. As we find no significant interaction effects and a practically relevant effect is below the minimum detectable effect sizes, our study lacks power to conclude that there is no relevant interaction effect.

3.2.4. Discussion

The aim of Study 2 was to replicate the test of hypotheses H1 to H5 by using as a dependent variable the intention to perform a more complex high-cost pro-environmental behavior that is conducted in a private context. Therefore, we used the double-vignette method developed by Schmidt et al. [36] for eliciting a heating-related rebound effect. In line with the findings [36], Study 2 showed that such a rebound effect can indeed be elicited within a lab context. Considering that the dependent variable was only the self-reported heating intention, we definitely consider the reported rebound effect (M_Dif = 0.3) to be practically relevant. It can be assumed that the effect on actual behavior in the private sphere of one’s own home would be larger than on the intention reported in our research, because social desirability would no longer have any influence.

The results do not support that BioVs are a direct independent predictor of the pro-environmental heating intention. Thus, the support for H1 that we had obtained in Study 1, which is in line with numerous previous studies, did not replicate in Study 2.

Although the manipulation check of the informational appeal indicated that receiving the informational appeal did not lead to a significant increase in problem awareness, this intervention nonetheless caused a significant decrease in the size of the rebound effect. Thus, the support for H2 that we had obtained in Study 1 was replicated in Study 2, although the effect size in Study 2 for the environmental appeal (d = 0.37) was smaller than the respective effect sizes found in Study 1 (d = 0.66 and 0.70). One potential conclusion from these findings is that informational appeals are more effective at influencing intentions associated with lower behavioral costs than at influencing intentions associated with higher behavioral costs.

Although the manipulation check of the socio-normative appeal indicated that receiving the socio-normative appeal led to a significant increase of the perceived social norm in Study 2, this increase did not affect the size of the rebound effect significantly. Thus, the support for H3 that we had obtained in Study 1 did not replicate in Study 2. Apparently, the increase in the perceived social norm did not carry over to participants’ imagined heating intention in the private context of their own rented flat. This lack of a normative effect is consistent with the results in Study 1, showing, in line with the literature [22,23], that greater privacy may decrease the effect of socio-normative interventions, as heating is considered to be a private behavior.

Replicating the null findings from Study 1, Study 2 also did not provide empirical support for H4 and H5, which represented the values-as-moderator hypothesis: neither the effectiveness of the informational appeal nor the effectiveness of the socio-normative intervention was moderated by the importance people assign to BioVs. Given the fact that the minimum detectable effect sizes according to the sensitivity analyses were higher than the practically relevant effect sizes, the lack of a significant effect in respect to our moderation hypotheses cannot be interpreted as evidence in favor of the null hypothesis.

4. General Discussion

The starting point of our research was the values-as-moderator hypothesis, which assumes that the effectiveness of informational and socio-normative appeals depends on how important BioVs are to the target person. On the one hand, this hypothesis is based on the assumption that BioVs are the central motivator for adopting pro-ecological behaviors. On the other hand, if this hypothesis is correct, it would have serious practical consequences: informational appeals would only be a meaningful intervention for people with high BioVs and socio-normative appeals would only be a meaningful intervention for people with low BioVs. Because of their grave practical consequences, we consider independent replication of both assumptions to be essential.

The assumption that BioVs are the central motivating factor for pro-environmental behavior [6,7,15,16,32] is only partially supported by our studies. Only in Study 1, where the dependent variable was the low-cost intention to purchase organic coffee, we found a weak main effect of BioVs on this intention. In Study 2, where the dependent variable was the high-cost intention to reduce rebound effects after energy renovation, we did not find a main effect of BioVs on this intention.

However, in contrast to the values-as-moderator hypothesis and in line with the available meta-analytic findings [20], both studies showed a main effect of informational appeals on the studied intentions. Furthermore, the impact of the informational appeal was stronger in the case of the low-cost organic coffee purchase intention than in the case of the high-cost rebound reduction intention. For the main effect of the socio-normative appeal the finding was similar: In the case of the low-cost organic coffee purchase intention the main effect of the socio-normative appeal was significant, whereas in the case of the high-cost rebound effect reduction intention the socio-normative appeals had no significant main effect. Thus, the fact that heating-related behaviors are normally performed in a private context may be one reason for the ineffectiveness of the socio-normative appeals in Study 2. Summarizing, our empirical results suggest that it is not peoples’ value profiles, but the perceived effort and costs associated with a behavior as well as the public vs. private nature of the context in which the intended behavior will be performed that are of more significance for the effectiveness of socio-normative appeals: in public contexts and for low-cost intentions the impact of socio-normative appeals is stronger than in private contexts or for high-cost intentions. Another point we may discuss is that the reference group (colleagues at work) may have been less relevant in Study 2 than in Study 1. A more relevant group might have been, e.g., other tenants in rented apartments.

From our point of view, the most important finding of the two studies, however, is that they do not support the values-as-moderator hypothesis. Neither in Study 1 nor in Study 2, the interaction effects of BioVs and the two different appeals were significant. Thus, our results do not support the assumption underlying the values-as-moderator hypothesis that the effectiveness of informational and socio-normative appeals depends on the importance participants assign to BioVs. Instead, our results are more in line with the view that informational and socio-normative appeals influence the behavior of people regardless of their value priorities. These results argue against the high influence attributed to values in activating behavior change [1]. In summary, according to our findings, the target behavior and the behavioral context seem to be of much greater importance for the effectiveness of behavioral interventions than individual value expression. Our findings of context-dependent effectiveness of different types of interventions argue for the approach of multimodal interventions that combine socio-normative triggering information with environmental awareness-raising information. Thus, multimodal interventions may be the best option if the goal is to influence the broadest possible range of behaviors toward more environmentally conscious actions.

An important limitation of our study is that we measured only behavioral intentions and not real behavior. Furthermore, despite the advantages of conducting experiments with random assignment to conditions, the experiments have a quite artificial character. Another limitation is that the mean BioVs scores were overall high in both the ‘high’ and the ‘low’ BioVs subgroups. This confirms the assessment that most people around the world regard BioVs as important [7]. But from a methodological point of view, the relatively small difference between the ‘high’ and ‘low’ BioVs subgroups may be a reason why the interaction of BioVs and the interventions was not significant. Another methodological deficit of our studies relates to the fact that, despite the relatively large sample size (>800 in each study), our studies had too little statistical power to detect minimum effect sizes we considered to be practically relevant. In addition, the participants may have guessed the aim of the studies and thereby might have been influenced to respond in a particular way, especially in Study 2.

In the future, conducting theory-driven intervention studies using objective behavioral measures to evaluate the effects of informational interventions in field experiments would be the best strategy to provide reliable evidence of the effectiveness of this type of intervention. In our view, conducting meta-analyses on the direct effect of BioVs on pro-environmental intentions and behaviors, as well as on the existence of interactions between BioVs and interventions, would be a good way to address the above-mentioned problem of statistical power and provide a solid evidence base.