Simple and Smart: Investigating Two Heuristics That Guide the Intention to Engage in Different Climate-Change-Mitigation Behaviors

Abstract

Individuals can support climate-change mitigation in many ways, e.g., through private-sphere behaviors or the support of political measures. We assume that the common climate-change-mitigation heuristic of restriction does not sufficiently support impactful mitigation intentions and therefore introduce and investigate a new heuristic (optimization heuristic.) In a cross-sectional survey with N = 1427 participants (representative of the German population with regard to age, gender, education), we developed two scales to measure the heuristics of restriction and optimization. As individual climate-change-mitigation intentions, we recorded four types of private-sphere behavior, activism, and three forms of policy support. Further psychological variables (personal norm, biospheric value orientation) and sociodemographic variables were recorded. The factorial structure of all concepts was assessed by means of confirmatory factor analyses. Hierarchical regression analyses with the climate-change-mitigation intentions as the criterion were carried out. Results support the assumption of two related, yet distinct, climate-change-mitigation heuristics that were highly correlated with biospheric value orientation. We additionally computed measure of the dominance of the restriction heuristic. This variable had no correlation with biospheric values, and correlated with the intentions in the expected ways, indicating that individuals with a dominant restriction heuristic tend to show lower scores of impactful climate-change-mitigation intentions.

1. Introduction

Individuals can take on many roles in the great transformation toward sustainability [1,2] and contribute to climate change mitigation through different behaviors [3], including private-sphere behaviors, citizenship actions, policy support and acceptance, and committed activism. These behaviors differ in their impacts on the mitigation of climate change [4]. Given the urgency of the climate crisis and the fact that most countries have failed their carbon reduction targets (e.g., Germany in the household and transportation sectors), it is crucial to focus on behaviors that have the greatest impact on reducing carbon emissions [5,6].

Studies have suggested that individual curtailment behaviors (e.g., reducing one’s car use or turning down the heat) have a lower impact than efficiency behaviors, especially than infrequent ones, such as purchasing a more fuel-efficient or electric car, investing in a new heating system, or insulating one’s home [7,8]. These decisions have long-lasting and impact-relevant consequences, as they change the framework of everyday consumption.

Alongside high-impact efficiency behaviors in the private sphere, public-sphere behaviors (e.g., policy support or activism for climate-change mitigation) may be even more relevant for effective climate-change mitigation. For years, UNFCCC reports, NGOs (Agora Verkehrswende), and climate advisory councils (e.g., WBGU) have called out national and EU policies for being insufficient and have called for more ambitious policies [9,10]. Thus, public support for a more ambitious climate-change-mitigation policy might be one of the behaviors through which individuals can contribute the most to limiting climate change.

To achieve the Paris Climate Agreement goals and successfully mitigate climate change, radical changes are required on a systemic level, brought forth by measures such as carbon taxation, the dismantling of subsidies for fossil fuels, a rapid decarbonization of the energy system (e.g., through wind turbines and photovoltaic systems), an expansion of the use of hydrogen in the steel industry or of bio-based carbon in the chemical industry see, e.g., [11], and new large-scale technologies (e.g., CCU or CCS). According to the latest IPCC report, every available measure needs to be undertaken to have the chance of keeping global warming below 1.5 degrees [12].

However, in the current dominant public environmental discourse, efficiency strategies and technical solutions to climate change are sometimes viewed skeptically, while the necessary yet less impactful curtailment behaviors receive broader public approval.

We assume that this is a consequence of the common climate-change-mitigation heuristic, which focuses primarily on restriction. Against this background, we make an effort to capture and identify a new heuristic, which in contrast to the common restriction heuristic focuses on the more impactful mitigation behaviors (optimization heuristic).

1.1. Heuristics as the Cognitive Link between Values and Pro-Environmental Behavior

There is a long psychological research tradition related to what motivates individuals to engage in various climate-change-mitigation behaviors. Over the years, values have emerged as a central predictor of specific attitudes and behaviors [13]. Values refer to desirable trans-situational goals that vary in importance and that serve as guiding principles in the life of a person or other social entity [14]. In particular, social–altruistic and biospheric values have successfully predicted a broad range of pro-environmental and climate-change-mitigation behaviors, such as energy-conserving behaviors [15], intentions to visit green hotels [16], sustainable water consumption [17], purchase intentions for recycled products [18], and the acceptance of green transport policies [19].

It has been shown repeatedly, however, that values do not influence behaviors directly, but that their influence is mediated through a chain of cognitive and emotional variables. The value–belief–norm (VBN) theory focuses on several cognitive mediators between pro-environmental behavior and values [1]. According to the VBN theory, pro-environmental behavior (e.g., private-sphere behavior or policy support) is directly influenced by an individual’s personal norm (moral obligation to act in ways consistent with one’s values), which is activated by problem awareness and ascription of responsibility [20]. The next and highest-level mediator between values and individual problem awareness is a set of beliefs—we would call it a heuristic—the “new environmental paradigm” (NEP), which consists of assumptions about economic growth and the relation of mankind and nature.

Heuristics in psychology are defined as cognitive processes that help individuals come to a quick decision by ignoring (consciously or unconsciously) parts of the available information [21,22]. Gigerenzer [23] therefore coined the concept of a positive, “ecological rationality”, which builds on simple heuristics which make decision making quick and smart. Also sets of consistent beliefs can work as heuristics, in the sense that they help to structure and select information in complex situations and fasten decision making, like the new environmental paradigm supports the quick identification of oneself being responsible for environmental behaviors [1].

Heuristics have proven influential in different domains of pro-environmental decision-making, e.g., when deciding on which foods to choose based on their environmental impact [24], or regarding the acceptance of fully automated driving [25].

In this line of argumentation, we understand climate-change-mitigation heuristics as a set of conscious presumptions that underlie our judgments of the appropriateness of climate-change-mitigation behaviors.

1.2. Possible Origin and Implications of the Current Restriction Heuristic

The aim of the present paper is to investigate what heuristics underlie different individual climate-change-mitigation behaviors and whether the current heuristic predominant in the environmental discourse is sufficient to support the required high-impact behaviors, such as efficiency behaviors, policy support, or support for new climate-change-mitigating technologies.

Ever since the Club of Rome’s “Limits to Growth”, an unlimited demand for growth and an exploitation of nature by Western economic systems have been discussed as the main drivers of environmental crises [26]. This argumentation gave rise to the important environmental movements of the 1970s, which focused on opposing this system of exploitation by, for example, boycotting certain companies and establishing alternative, more modest lifestyles [27]. This heuristic of modesty and restriction was underpinned by the oil crises in 1973 and 1979/1980 and an aggravation of the resource problem. In the late 1980s, this heuristic was supplemented by a global justice claim referring to the fair distribution of opportunities for development [28]. The global justice perspective on environmental crises was solidified in the Rio Conference in 1992. Furthermore, the focus of the debates expanded from the overexploitation of resources to exceeding carrying capacity in terms of carbon emissions. In the spirit of the global justice perspective, the so-called budget approach posits that industrialized countries bear greater responsibility for contributing to and mitigating the climate crisis and must therefore accept cuts to their remaining carbon budget so as to avoid limiting the development prospects of the least developed countries. To this effect, in 1992, the Kyoto protocol required industrialized countries and economies that have contributed the most to the climate crisis to commit to reducing their greenhouse gas emissions.

Against this historical background, a generalized heuristic of restriction seems to have emerged, rooted in the need for restraint and in the moral obligation to renounce the need for resources as a means of repentance and reparation for guilt [29].

This heuristic of restriction seems viable for limiting and compensating for past overconsumption and remaining within planetary boundaries as well as for addressing the inequitable distribution of resources around the global. On a global macrolevel, sufficiency (i.e., the reduction of one’s consumption of resources in order to remain within planetary boundaries) seems unquestionable.

However, the question arises as to whether the restriction heuristic is also a sufficient heuristic on the level of individual climate-change-mitigation behavior, which embraces efficiency behaviors and supports new technologies and economic instruments in addition to sufficiency behaviors to achieve changes in the economic sector. Some of the newer individual climate-change-mitigation behaviors, such as compensation payments or the purchase of an electric vehicle even seem to be in potential conflict with the restriction heuristic. In the morally driven discourse of restriction and renunciation, efficiency behaviors can be perceived as a means of reducing one’s obligation to restrict oneself and might be misunderstood as indulgences, particularly when associated with the use of financial resources.

Accordingly, the heuristic of pure restriction and renunciation needs to be complemented by an optimization heuristic that supports the seeking of new options, beyond restriction, through which an individual can actively and most effectively contribute to climate-change mitigation.

1.3. Present Study

With the present study, we aimed to empirically investigate the assumption that two different heuristics may support different climate-change-mitigation behaviors, or rather the intention to engage in them. As described above, we assume that the heuristics guide judgments of the appropriateness of climate-change-mitigation behaviors as a set of conscious presumptions (see

Table 1. Hypothesized relations for the restriction and optimization heuristics with different types of climate-change-mitigation intentions.

Hypotheses	Type of Behavior (Intention)	Restriction Heuristic	Optimization Heuristic
H1.1, H1.2 & H1.3	Curtailment	++	+
H2.1, H2.2 & H2.3	Efficiency	+	++
H3.3, H3.2 & H3.3	New alternatives	+	++
H4.1, H4.2 & H4.3	Political consumption	+	++
	Activism	No hypothesis
H5.1, H5.2 & H5.3	Support for restrictive measures	++	+
H6.1, H6.2 & H6.3	Support for push measures	+	++
H7.1, H7.2 & H7.3	Support for pull measures	+	++

Note. + represents a hypothesized significant positive relationship, ++ indicates a hypothesized stronger positive relationship. Hypotheses HX.1 refer to the hypothesized relationship between the type of behavior and the restriction heuristic, hypotheses HX.2 refer to the hypothesized relationship between the type of behavior and the optimization heuristic, and hypotheses HX.3 refer to the hypothesized difference in the strength of the relationships.

for specific hypotheses H1.1 to H7.3). We assumed that the restriction heuristic would be more strongly related to the intention to show curtailment behaviors and support for restrictive policy measures. Applying the whole spectrum of individual behaviors involved in the transformation [3], we also made assumptions about new private-sphere behaviors (e.g., trying out new practices and products) and new political behaviors (e.g., political consumption or divestment or supporting policy instruments associated with new technologies). We expected that the intentions for these behaviors would be more strongly related to the optimization heuristic. We had no hypothesis regarding activism.

The personal norm for climate protection, see, e.g., [30,31], and a biospheric value orientation [32] have repeatedly turned out to be relevant predictors of climate-change-mitigation behaviors. We therefore expected significant relationships between the personal norm for climate protection and the two heuristics (H8.1 and H8.2) as well as between the biospheric value orientation and the heuristics (H9.1 and H9.2). However, because the restriction heuristic is considered to be rooted in a more traditional perspective on environmental protection, we expected the restriction heuristic to be more strongly related to the biospheric value orientation (H9.3).

A large body of research, including single surveys and integrative, cross-national studies, has investigated the relevance of sociodemographic variables for national or individual environmental concern, e.g., [33]. Gender and political orientation have repeatedly emerged as relevant predictors, with women and libertarians tending to be more concerned about environmental problems. Several studies have suggested an influence of education and income on environmental concern, see, e.g., [34]. Income has repeatedly emerged as a relevant driver of national and individual carbon footprints [35]. We therefore additionally investigated the relationships of the two heuristics with sociodemographic variables (age, education, and income) and with participants’ political orientation. However, we had no specific hypotheses regarding the relationship between sociodemographic variables and the two heuristics.

2. Materials and Methods

2.1. Data Collection and Participants

The present study was designed and conducted following the APA guidelines on the ethical conduct of research. According to German Law, survey studies do not require ethical approval when anonymity is secured, and no sensitive content is assessed. We commissioned Bilendi, an external panel provider company, to recruit a stratified sample for the German population. Data were collected online from 22 April to 19 May 2022, across Germany. Of the 4600 people who visited the link sent out by Bilendi, a total of 1548 individuals completed the online survey. Of these, 121 participants were excluded on the basis of answering time (people who took less than five minutes to complete the survey, which took, on average, 25 min to answer), missing values (more than 60% missing values), and implausible answers to open questions. The final sample comprised N = 1427 participants. Informed consent was obtained from all participants. The sample was representative of the German population with respect to age, gender, and education (50.2% female, 49.7% male, and 0.1% diverse). Ages ranged from 16 to 74 years (M = 47.18, SD = 15.94). See

Table A13. Descriptive Statistics for Sociodemographic Variables.

	n (%)	Min.	Max.	Mean	SD
Age	1427	16	74	47.18	15.94
Gender
Female	716 (50.2)
Male	709 (49.7)
Diverse	2 (0.1)
Average monthly household net income
Under EUR 1000	159 (11)
EUR 1000 to under EUR 2000	343 (24)
EUR 2000 to under EUR 3000	317 (22)
EUR 3000 to under EUR 4000	245 (17)
EUR 4000 to under EUR 5000	155 (11)
EUR 5000 to under EUR 6000	55 (4)
EUR 6000 and more	49 (3)
Highest level of education
(Noch) Kein allgemeiner Schulabschluss	27 (2)
Maximal Haupt-(Volks-, Grund-)schulabschluss	90 (6)
Haupt- (Volks-, Grund-)schulabschluss mit abgeschlossener Lehre/Berufsausbildung	353 (25)
Weiterführende Schule ohne Abitur (Realschulabschluss/Mittlere Reife/Oberschule)	448 (31)
Abitur, (Fach-)Hochschulreife ohne Studium	243 (17)
Studium (Universität, Hochschule, Fachhochschule, Polytechnikum)	266 (19)
Place of residency
Very urban	356 (25)
Rather urban	538 (38)
Rather rural	408 (29)
Very rural	125 (9)
Political orientation	1297	1	10	5.93	1.76

in Appendix B for more information on the sociodemographic features of the sample.

We checked the missing value structure in our data, as some psychological variables showed high percentages of missing data (nine variables with at least 10%). Little’s [36] MCAR test suggested that the missing data were on a continuum between missing at random and not missing at random, χ²(141161) = 143518.4, p < 0.001. Because the missingness of the data was also moderately to highly correlated with other variables, the estimation maximization method with NORM Version 2.03 was chosen as an appropriate method for imputation [37]. Only values on the psychological variables were imputed. The climate-change-mitigation intentions were only partly imputed as the participants had the option to choose not applicable for the respective items.

2.2. Measures

2.2.1. Climate-Change-Mitigation Heuristics

To capture the two heuristics, we developed a pool of twelve items. The items were evaluated on a 5-point Likert scale ranging from 1 (completely disagree) to 5 (completely agree). Participants were given the option don’t know/not applicable. The pool resulted from five informal interviews with experts from environmental psychology and political science.

The restriction heuristic scale (short: restriction) comprised five items, e.g., “We have put our planet through far too much in recent years, so now we have to pay the price and renounce” or “The Western lifestyle is the cause of climate change, and it is only fair that we are now affected by severe restrictions”. The optimization heuristic scale (short: optimization or optimization heuristic) comprised five items, e.g., “In order to quickly limit the climate crisis, each individual should primarily implement the measures that save a particularly large amount of CO₂ in their area” or “I am willing to invest money to limit the climate crisis”. Two items from the initial pool were excluded (“We as citizens of industrial nations have to lead the way in climate protection”; “I inform myself about ways in which I can reduce my carbon footprint to a particularly large extent”) from the initial item pool after empirically determining the item-factor configuration which distinguished the two heuristics best (the decision was based on the wording of the items and the modification indices extracted from the CFA. We focused on the modification indices, which indicated an improvement in model fit by adding a cross-loading to the model and thus indicating items that reflect both paradigms the most. After having eliminated Item OH6, we computed a new model and repeated the procedure). See Appendix A for a complete list of items.

We estimated the reliability of the two heuristic scales with McDonald’s omega (ω). The assessment of ω is generally preferred over the more common Cronbach’s α, as it does not assume at least essential τ-equivalence for the measurement model [38]. The reliabilities were sufficiently high for both heuristics (ω_restriction = 0.85 and ω_optimization = 0.83).

2.2.2. Biospheric Value Orientation and Personal Norm for Climate-Change Mitigation

To measure their biospheric value orientation (short: biospheric values), participants were asked to indicate the extent to which they considered three values (e.g., “Respecting the earth, harmony with other species”) as guiding principles in their lives on a 9-point Likert scale ranging from −1 (opposed to my values) through 0 (not important) to 7 (extremely important). Items were taken from Stern et al.’s Brief Inventory of Values [39]. Personal norm for climate-change mitigation (short: personal norm) was measured with three items on a 5-point Likert scale ranging from 1 (completely disagree) to 5 (completely agree) taken from Matthies and Merten [40] (e.g., “I feel obligated to save CO₂ in my everyday life”). See Appendix A for a complete list of items.

Both the biospheric value orientation (ω = 0.87) as well as the personal norm (ω = 0.93) showed sufficiently high reliability.

2.2.3. Intention to Engage in Different Climate-Change-Mitigation Behaviors

To measure the intention to engage in five categories of climate-change-mitigation behavior (climate-change-mitigation intention), participants were asked which of the behaviors they were planning to engage in during the next year. Answers were assessed on a 6-point Likert scale ranging from 1 (no, definitely not) to 5 (yes, definitely) to 6 (I have already done this). Participants were also given the option not applicable, which was then coded as a missing value. Items were taken from Matthies and Merten [40] and supplemented by newly developed items (view Appendix A for entire list of items). All items were preceded by “In the following year, do you plan to …”

The intention for curtailment behavior in the private sphere (short: curtailment) was measured with five items, e.g., “In the following year, do you plan to reduce the temperature in your home by 1 degree Celsius in winter?”. The intention for efficiency behavior in the private sphere (short: efficiency) was measured with six items, e.g., “In the following year, do you plan to replace your heating system with a more modern, climate-friendly system?”. The intention to try out new products and practices in the private sphere (short: new alternatives) was measured with three items, e.g., “In the following year, do you plan to take part in food sharing initiatives, i.e., give food to other people/institutions before it expires?”. The intention for political consumption and divestment (short: political consumption) was measured with four items, e.g., “In the following year, do you plan to offset your carbon emissions by making compensation payments to climate offset projects (e.g., via Atmosfair, myClimate or Primaklima)?”. The intention to engage in activism was measured with three items, e.g., “In the following year, do you plan to participate in environmental protests?”. See Appendix A for a complete list of items.

The reliability of the five categories was sufficiently high except for new alternatives (the items measuring new alternatives showed insufficient psychometric properties, and thus, the following results involving this construct should be interpreted accordingly) (ω = 0.62), with ω values ranging from 0.74 (curtailment) to 0.76 (efficiency).

To assess the support for restrictive policy measures (short: support for restrictive measures), participants were asked to indicate their support of five measures on a 5-point Likert scale ranging from 1 (strongly opposed) to 5 (strongly in favor), e.g., “Ban on plastic packaging”. Support for economic push measures (e.g., taxes; short: support for push measures) was assessed with two items, “Stronger increase in carbon prices” and “Higher taxes on particularly climate-damaging products (e.g., meat)”. Support for economic pull measures (e.g., subsidies; short: support for pull measures) was assessed with five items, e.g., “Greater funding for the expansion of local public transport”.

The ω values were 0.78 (support for restrictive measures), 0.81 (support for economic pull measures), and 0.77 (support for economic push measures).

2.2.4. Political Orientation and Sociodemographic Variables

Political orientation was measured with one item [41]: “Many people use the terms ‘left’ and ‘right’ to refer to different political attitudes. We have a scale here that goes from left to right. When you think of your political views, where would you rank those views on this scale?” Answers were assessed on a scale ranging from 1 (left) to 10 (right) and were recoded for the analyses. Higher values reflected a more left political orientation.

Furthermore, we assessed the following sociodemographic variables: participants’ age, gender, highest level of education, and average monthly household net income. Income was assessed with the question “What was your average monthly household income over the last 12 months?” (see

Table A20. Test statistics for univariate and multivariate normality tests for the climate-change-mitigation heuristics.

		W
Item level
Restriction heuristic (N = 1427)
	RH1	0.88
	RH2	0.92
	RH3	0.92
	RH4	0.85
	RH5	0.92
Optimization heuristic (N = 1427)
	OH1	0.88
	OH2	0.91
	OH3	0.92
	OH4	0.92
	OH5	0.90
		W	Henze-Zirkler	Mardia skewness	Mardia kurtosis
Scale level
Restriction heuristic		0.98	8.27	445.75	20.23
Optimization heuristic		0.98	4.98	326.94	12.70

Note. W = Shapiro–Wilk test statistic (perfect normality = 1.00; all test statistics were the same when Shapiro–Francia test was applied). All ps < 0.001.

in Appendix B).

2.3. Factor Structure of the Two Climate-Change-Mitigation Heuristics

We applied an exploratory factor analytical (EFA) approach to determine the most appropriate item-factor configuration of the heuristic scales. We first included all items and then eliminated the two items from the most appropriate model. Horn’s [42] parallel analyses with all items suggested the extraction of one to four factors (based on a principal component and a principal axis analysis). Thus, we tested different models by means of EFA (utilizing maximum likelihood (ML) estimators and the oblique promax rotation; see

Table A14. Exploratory Factor Analysis Models for the 12-Item Heuristic Scales.

	One-Factor Model		Two-Factor Model			Three-Factor Model				Four-Factor Model
	λF1	h2	λF1	λF2	h2	λF1	λF2	λF3	h2	λF1	λF2	λF3	λF4	h2
RH1	0.78	0.60	0.44	0.38	0.59	0.60	0.32		0.64				0.89	0.91
RH2	0.70	0.49		0.95	0.70		0.84		0.70		0.85			0.70
RH3	0.72	0.52		0.54	0.54	0.35	0.47		0.56		0.33		0.36	0.57
RH4	0.63	0.39	0.49		0.39	0.73			0.47	0.64				0.46
RH5	0.70	0.49		0.81	0.62		0.70		0.61		0.80			0.65
OH6	0.83	0.69	0.65		0.69	0.65			0.70	0.72				0.72
OH7	0.62	0.39	0.57		0.40			0.81	0.63			0.83		0.67
OH1	0.78	0.61	0.91		0.70	0.76			0.70	0.77				0.69
OH2	0.67	0.45	0.35	0.36	0.45		0.31	0.61	0.58			0.56		0.57
OH3	0.68	0.46	0.46		0.46				0.46	0.36				0.47
OH4	0.58	0.34	0.55		0.35	0.48			0.34	0.52				0.35
OH5	0.71	0.50	0.89		0.60	0.58		0.41	0.61	0.64		0.35		0.61
Explained σ2 (eigenvalue)	0.49 (5.94)		0.32 (3.85)	0.22 (2.65)		0.26 (3.15)	0.18 (2.13)	0.14 (1.73)		0.24 (2.83)	0.16 (1.91)	0.13 (1.51)	0.09 (1.12)

Note. N = 1427. λ = standardized factor loadings, h² = communality, Explained σ² = item variance explained by the respective factor. For the exploratory factor analyses, the oblique promax rotation and Maximum Likelihood estimation were applied. Factor loadings below 0.30 were not included in this table.

in Appendix B for an overview of the models). On the basis of the size of the estimated factor loadings, at least >0.30; [43,44], the existence of substantial cross loadings, the number of items loading on each factor (at least three), and the variance explained by the factors, we decided to test the one-factor and two-factor models with confirmatory factor analyses (CFA). (Strictly speaking, this approach was not confirmatory, but as the method and its constraints on the measurement model are commonly described as CFA, we will refer to these analyses as CFA.) Furthermore, we specified a model based on our hypothesized measurement structure. We estimated the model parameters with a robust ML (MLR) algorithm accounting for the lack of multivariate normality in the data [45,46]. A comparison of the model fit indices (

Table A15. Fit indices for different CFA models for the heuristic scales.

Model	CFI	RMSEA	SRMR	BIC	χ2 (df)
12 items
One-factor model (exploratory)	0.899	0.110	0.051	44871	789.4 *** (54)
Two-factor model (exploratory)	0.920	0.099	0.049	44683	639.9 *** (53)
Two-factor model (hypothesized)	0.927	0.094	0.043	44619	587.9 *** (53)
11 items
Two-factor model (hypothesized)	0.927	0.096	0.043	41601	490.4 *** (43)
10 items
Two-factor model (hypothesized)	0.939	0.093	0.040	37804	366.2 *** (34)
Two-factor model with error covariance (hypothesized)	0.962	0.075	0.036	37651	243.7 *** (33)

Note. N = 1427. χ² = scaled χ² test statistic, df = degrees of freedom, CFI = (robust) comparative fit index, RMSEA = (robust) root mean square error of approximation, SRMR = standardized root mean square residual, BIC = Bayesian information criterion. Robust maximum likelihood estimation was applied. The error covariance of the last model was between RH2 and RH5. *** p < 0.001.

in Appendix B) suggested the superiority of the hypothesized model. After the exclusion of the two items, the final CFA model with an additional error covariance between two items (both referring to the responsibility of Western countries) yielded sufficient fit indices [robust CFI = 0.962, robust RMSEA = 0.075, SRMR = 0.036; 47; see

Table A16. Confirmatory factor analysis model for the 10-item heuristic scales.

	λ
	Restriction Heuristic	Optimization Heuristic
RH1	0.83
RH2	0.70
RH3	0.78
RH4	0.64
RH5	0.69
OH1		0.81
OH2		0.66
OH3		0.69
OH4		0.59
OH5		0.74

Note. N = 1427. λ = standardized factor loadings. Robust maximum likelihood estimation was applied. Standardized error covariance between RH2 and RH5: r = 0.370, latent factor correlation: r = 0.88.

for the model].

This model was compared across gender (female; male), age groups (younger than 30 years; 30 to 39 years; 40 to 49 years; 50 to 59 years; 60 to 74 years), net household income groups (under 2000 EUR/month; 2000 to 3999 EUR/month; over 4000 EUR/month), and formal education levels (lower secondary school graduate who completed vocational training; secondary school graduate; at least high school graduate) to test for measurement invariance (MI) (To detect potential differences in the measurement structure, we applied multiple-group CFA (MG-CFA), which tests successive levels of MI (configural, metric, scalar, and residual) by constraining different parameters of the measurement model (to detect potential differences in the measurement structure, we applied multiple-group CFA (MG-CFA), which tests successive levels of MI (configural, metric, scalar, and residual) by constraining different parameters of the measurement model). The multiple-group CFA (MG-CFA) indicated that the heuristic scales achieved partial residual MI across all grouping variables with a freed intercept of item OH2 (see

Table A17. Heuristic scales measurement invariance models for grouping variables.

Gender (n = 1425)
	CFI	RMSEA	SRMR	BIC
Configural	0.968	0.068	0.032	37751
Metric	0.968	0.065	0.035	37702
Scalar	0.953	0.074	0.044	37754
Partial scalar	0.958	0.071	0.041	37726
Partial residual	0.956	0.068	0.042	37677
Age groups (N = 1427)
	CFI	RMSEA	SRMR	BIC
Configural	0.965	0.071	0.038	38387
Metric	0.967	0.063	0.047	38177
Scalar	0.952	0.071	0.055	38083
Partial scalar	0.962	0.064	0.051	38037
Partial residual	0.962	0.059	0.051	37799
Income (n = 1323)
	CFI	RMSEA	SRMR	BIC
Configural	0.963	0.074	0.035	35225
Metric	0.959	0.072	0.049	35152
Scalar	0.944	0.079	0.055	35153
Partial scalar	0.958	0.070	0.050	35076
Partial residual	0.952	0.069	0.050	34997
Education (N = 1427)
	CFI	RMSEA	SRMR	BIC
Configural	0.961	0.075	0.036	37965
Metric	0.958	0.073	0.051	37889
Scalar	0.945	0.078	0.056	37879
Partial scalar	0.953	0.072	0.053	37831
Partial residual	0.950	0.070	0.052	37735

Note. CFI = (robust) comparative fit index, RMSEA = (robust) root mean square error of approximation, SRMR = standardized root mean square residual, BIC = Bayesian information criterion. Robust maximum likelihood estimation was applied. For the models with partial invariance, the constraint on the intercept of OH2 was freed. The error covariance between RH2 and RH5 was not fixed across groups.

in Appendix B for the relevant fit indices). Thus, even analyses with manifest scale scores were possible [47,48,49].

2.4. Further Measures

The CFA testing the hypothesized measurement model of the biospheric values, personal norm, and the climate-change-mitigation intentions showed sufficient model fit. The measurement model of the support for different policy measures was adapted based on EFA. See Appendix C for an extensive description of the measurement models and their MI. Furthermore, all scales were at least partial residual invariant across the groups defined by sociodemographic variables.

3. Results

Most analyses were conducted with R version 4.0.3 [50] and the following R packages: dplyr [51], EFAtools [52], lavaan [53], MVN [54], nortest [55], pastecs [56], psych [57], and utils [50]. The regression models were estimated with IBM SPSS Statistics 27 [58]. For the Pearson and Filon’s z-tests to compare correlation coefficients, we used a tool developed by Hemmerich [59].

We refrained from computing an extensive power analysis, as our sample size (N = 1427) exceeded even conservative rules of thumb for sufficient sample size for factor analytical procedures, 15 subjects per variable; [60]. Furthermore, the statistical power of finding a significant result in our sample at α ≤ 0.05 for a true Pearson correlation of 0.10 was estimated to be 0.97 [61]. The likelihood of finding a small effect of R² = 0.02 [62] that is valid in the total population in a linear regression model with two predictors, for our sample size, and α ≤ 0.05 was predicted to be 0.999 [63].

3.1. Descriptive Statistics

The means of the relevant constructs can be found in

Table 2. Descriptive Statistics for Scale Scores (Based on Mean Factoring).

	Min.	Max.	Mean	SD	SE
Heuristics (N = 1427)
Restriction heuristic	0.89	5.14	3.39	0.93	0.02
Optimization heuristic	0.89	5.16	3.37	0.85	0.02
Biospheric values (N = 1427)	−1	7	5.03	1.59	0.04
Personal norm (N = 1427)	0.90	5.12	3.48	1.09	0.03
Stern’s behavioral roles (intention)
Curtailment (n = 1423)	1	6	3.98	1.08	0.03
Efficiency (n = 1416)	1	6	2.78	1.08	0.03
New alternatives (n = 1419)	1	6	3.51	1.03	0.03
Political consumption (n = 1411)	1	6	2.96	1.08	0.03
Activism (n = 1409)	1	6	2.65	1.15	0.03
Support for different measures (N = 1427)
Support for restrictive policy measures	1	5.94	3.19	0.95	0.03
Support for economic push measures	0.45	5.19	2.55	1.23	0.03
Support for economic pull measures	1	5.20	3.72	0.83	0.02

Note. SE = standard error. Values below or above the scale range are due to the values being imputed. Allowing values outside the scale can counteract a potential bias.

. Descriptively, the mean values of both heuristics and the personal norm were on a comparable level (M_restriction = 3.39, M_optimization = 3.37, M_{personal norm} = 3.48). For the different behavioral roles in climate-change mitigation, the intention to show curtailment behavior (M = 3.98) seems to have been the highest in our sample followed by the intention to try out new products and practices (M = 3.51), the intention for political consumption (M = 2.96), the intention for efficiency behavior (M = 2.78), and the intention for activism (M = 3.65). Furthermore, in our sample we observe the tendency that the support for economic pull measures (M = 3.72) was higher than for restrictive policy measures (M = 3.19) which in turn found more acceptance than economic push measures (M = 2.55).

Table A18. Descriptive statistics for scale scores (based on mean factoring).

	Min.	Max.	Median	Mean	SD	Skewness	Kurtosis
Heuristics (N = 1427)
Restriction heuristic	0.89	5.14	3.40	3.39	0.93	−0.41	−0.16
Optimization heuristic	0.89	5.16	3.40	3.37	0.85	−0.49	0.22
Biospheric values (N = 1427)	−1	7	5.33	5.03	1.59	−0.78	0.43
Personal norm (N = 1427)	0.90	5.12	3.67	3.48	1.09	−0.57	−0.18
Stern’s behavioral roles
Curtailment (n = 1423)	1	6	4	3.98	1.08	−0.41	−0.11
Efficiency (n = 1416)	1	6	2.75	2.78	1.08	0.26	−0.41
New alternatives (n = 1419)	1	6	3.33	3.51	1.03	0.12	−0.04
Political consumption (n = 1411)	1	6	3	2.96	1.08	0.21	−0.10
Activism (n = 1409)	1	6	2.67	2.65	1.15	0.44	−0.24
Support for different measures (N = 1427)
Support for restrictive policy measures	1	5.94	3.20	3.19	0.95	−0.17	−0.44
Support for economic push measures	0.45	5.19	2.50	2.55	1.23	0.29	−0.96
Support for economic pull measures	1	5.20	3.83	3.72	0.83	−0.57	0.07

Note. Values below or above the scale range are due to the values being imputed. Allowing values outside the scale can counteract a potential bias.

in Appendix B provides an overview of the descriptive statistics of the manifest factor scores estimated by computing the mean. None of the manifest factor scores were normally distributed (it should be noted that these tests are sensitive to large samples).

The item-level descriptive statistics can be found in

Table A19. Descriptive statistics for items (imputed data set).

		Min.	Max.	Median	Mean	SD	Skewness	Kurtosis
Restriction heuristic (N = 1427)
	RH1	1	6.04	4	3.70	1.13	−0.64	−0.25
	RH2	0.32	5.29	3	3.04	1.19	−0.13	−0.68
	RH3	0.68	6.43	3	3.20	1.22	−0.19	−0.76
	RH4	0.46	6.60	4	3.91	1.13	−0.91	0.21
	RH5	−0.07	5.42	3	3.10	1.20	−0.17	−0.71
Optimization heuristic (N = 1427)
	OH1	0.73	6.06	4	3.70	1.03	−0.67	0.22
	OH2	0.12	6.01	3	2.77	1.25	0.06	−0.92
	OH3	0.52	5.72	3	3.34	1.10	−0.31	−0.34
	OH4	−0.19	5.84	3.43	3.50	0.99	−0.33	−0.02
	OH5	0.04	5.81	4	3.51	1.13	−0.60	−0.17
Biospheric values (N = 1427)
	BV1	−1	7	5	5.14	1.75	−0.90	0.58
	BV2	−1	7	5	4.59	1.92	−0.58	−0.14
	BV3	−1	7	6	5.38	1.68	−1.04	0.76
Personal norm (N = 1427)
	PN1	1	5	4	3.49	1.18	−0.58	−0.40
	PN2	1	5.18	4	3.54	1.14	−0.59	−0.25
	PN3	0.71	5.69	3.50	3.41	1.17	−0.45	−0.48
Curtailment
	CB1 (n = 1395)	1	6	4	4.34	1.53	−0.57	−0.60
	CB2 (n = 1310)	1	6	4	3.66	1.71	−0.08	−1.23
	CB3 (n = 1390)	1	6	4	4.12	1.68	−0.41	−0.99
	CB4 (n = 1400)	1	6	4	3.59	1.25	−0.12	−0.19
	CB5 (n = 1408)	1	6	4	4.13	1.52	−0.35	−0.80
Efficiency
	EB1 (n = 1168)	1	6	2	2.64	1.51	0.70	−0.40
	EB2 (n = 996)	1	6	3	3.12	1.62	0.42	−0.89
	EB3 (n = 1034)	1	6	3	2.91	1.49	0.60	−0.42
	EB4 (n = 1021)	1	6	3	3.17	1.67	0.42	−0.98
	EB5 (n = 1370)	1	6	2	2.65	1.51	0.56	−0.64
	EB6 (n = 1363)	1	6	2	2.46	1.47	0.75	−0.33
New alternatives
	NB1 (n = 1408)	1	6	4	4.51	1.21	−0.57	0.09
	NB2 (n = 1317)	1	6	2	2.60	1.39	0.65	−0.26
	NB3 (n = 1303)	1	6	3	3.24	1.44	0.16	−0.69
Political consumption
	PB1 (n = 1304)	1	6	2	2.56	1.28	0.59	−0.10
	PB2 (n = 1278)	1	6	2	2.37	1.20	0.87	0.67
	PB3 (n = 1376)	1	6	3	2.77	1.44	0.60	−0.27
	PB4 (n = 1334)	0.34	6	4	4.02	1.63	−0.24	−1.06
Activism
	AB1 (n = 1377)	1	6	2	2.17	1.25	1.07	0.79
	AB2 (n = 1376)	1	6	3	3.15	1.47	0.26	−0.65
	AB3 (n = 1345)	1	6	2	2.61	1.45	0.68	−0.34
Support for restrictive policy measures (N = 1427)
	SR1	1	5.85	4	3.72	1.19	−0.63	−0.44
	SR2	−0.06	7.91	3	3.20	1.39	−0.20	−0.99
	SR3	0.79	6.22	4	3.71	1.19	−0.68	−0.31
	SR4	−0.97	6.24	2	2.25	1.35	0.68	−0.71
	SR5	−0.09	6.79	3	3.06	1.37	−0.07	−1.09
Support for economic push measures (N = 1427)
	ST1	−0.09	5.58	2	2.37	1.31	0.50	−0.85
	ST2	−0.15	6.63	3	2.72	1.42	0.22	−1.19
Support for economic pull measures (N = 1427)
	SS1	−0.15	6.15	3	3.07	1.43	−0.11	−1.21
	SS2	1	6.17	4	4.02	1.07	−0.89	0.12
	SS3	0.81	6.25	4	3.75	1.15	−0.66	−0.26
	SS4	0.66	6.22	4	3.73	1.19	−0.66	−0.35
	SS5	1	6.67	4	3.84	1.17	−0.74	−0.13
	SS6	0.22	6.19	4	3.92	1.08	−0.85	0.23

Note. Values below or above the scale range are due to the values being imputed. Allowing values outside the scale can counteract a potential bias.

in Appendix B. An inspection of the items for univariate normality using the Shapiro–Wilk and Shapiro–Francia tests showed that all items deviated from normality [64,65]. These findings were corroborated by the results of the Henze–Zirkler and Mardia tests for multivariate normality, as the tests failed for the groups of items for each construct [66,67]; see Table A15, Table A16, Table A17, Table A18 and Table A19 for the statistics for univariate and multivariate normality.

3.2. Relations between the Two Heuristics and Sociodemographic Variables

For categorical sociodemographic variables, we dichotomized them and computed unpaired t-tests. Numeric variables were correlated with the two heuristics.

Participants who were at least high school graduates showed a significantly higher level of the restriction heuristic, (M = 3.33 vs. M = 3.50), t(1425) = −3.413, p = 0.001, as well as a stronger optimization heuristic, (M = 3.29 vs. M = 3.51), t(1425) = −4.793, p ≤ 0.001, compared with participants without a high school degree. We also found that both heuristics were significantly correlated with political orientation (r_restriction = 0.29, r_optimization = 0.26; p-values for the correlations will only be presented when they are not <0.001). Hence, participants with a more left political orientation tended to display a stronger restriction as well as optimization heuristic. The correlations of both heuristics with the political orientation were not significantly different, z = 1.620, p = 0.105 [68,69]. When comparing participants with a higher household net income (above 3000 EUR/month) with participants with a lower income, the groups did not show significant differences in their levels of either heuristic, t_restriction(1321) = 0.796, p = 0.426; t_optimization(1321) = −1.619, p = 0.106. Furthermore, neither the restriction heuristic nor the optimization heuristic were significantly correlated with age.

3.3. Relations between the Two Heuristics, Biospheric Values, and Personal Norm

The correlation between the two heuristics was r = 0.74 (latent factor correlation: r = 0.88; a hierarchical CFA with both factors loading on a second-order factor also converged). Furthermore, the optimization heuristic was more strongly positively correlated with biospheric value orientation (r_restriction = 0.41, r_optimization = 0.47) and the personal norm for climate-change mitigation (r_restriction = 0.69, r_optimization = 0.80) than the restriction heuristic [68,69].

3.4. Relations between the Two Heuristics and Climate-Change-Mitigation Intentions

3.4.1. Bivariate Analyses

We conducted further analyses to investigate our hypotheses. First, we correlated the manifest scale scores of the two heuristics and the intentions and tested whether one orientation was more strongly correlated with an intention than the other [68,69].

Both heuristics were positively correlated with curtailment (r_restriction = 0.38, r_optimization = 0.44) and efficiency (r_restriction = 0.34, r_optimization = 0.44), new alternatives (r_restriction = 0.31, r_optimization = 0.39), political consumption and divestment (r_restriction = 0.45, r_optimization = 0.58), activism (r_restriction = 0.36, r_optimization = 0.45), the support for restrictive policy measures (r_restriction = 0.58, r_optimization = 0.60), and the support for economic push (r_restriction = 0.57, r_optimization = 0.60) and pull measures (r_restriction = 0.46, r_optimization = 0.54). Hence, a stronger restriction and optimization heuristic were positively associated with the intention for sustainable behavior. Each climate-change-mitigation intention was more strongly correlated with the optimization heuristic than with the restriction heuristic. The exception was the support for restrictive policy measures for which the correlations did not significantly differ.

All correlation coefficients and the significance tests of the difference between the respective coefficients can be found in

Table 3. Pearson correlations for the heuristic scales and dominance of restriction score with biospheric values, personal norm, and climate-change-mitigation intentions.

	Restriction	Optimization	Pearson & Filon’s z	Dominance of Restriction
Optimization (N = 1427)	0.74 ***	1
Biospheric values (N = 1427)	0.41 ***	0.47 ***	z = −3.624, p < 0.001	−0.02
Personal norm (N = 1427)	0.69 ***	0.80 ***	z = −9.310, p < 0.001	−0.05
Curtailment (n = 1423)	0.38 ***	0.44 ***	z = −3.151, p = 0.002	−0.02
Efficiency (n = 1416)	0.34 ***	0.44 ***	z = −5.912, p < 0.001	−0.09 **
New alternatives (n = 1419)	0.31 ***	0.39 ***	z = −4.526, p < 0.001	−0.06 *
Political consumption (n = 1411)	0.45 ***	0.58 ***	z = −7.926, p < 0.001	−0.11 ***
Activism (n = 1409)	0.36 ***	0.45 ***	z = −5.435, p < 0.001	−0.07 **
Support for restrictive measures (N = 1427)	0.58 ***	0.60 ***	z = −1.280, p = 0.201	0.05 *
Support for push measures (N = 1427)	0.57 ***	0.60 ***	z = −2.002, p = 0.045	0.04
Support for pull measures (N = 1427)	0.46 ***	0.54 ***	z = −5.144, p < 0.001	−0.05

Note. Restriction (heuristic), Optimization (heuristic), (intention for) Curtailment (behavior in the private sphere), (intention for) Efficiency (behavior in the private sphere), New alternatives (intention to try out new products and practices in the private sphere), (intention for) Political consumption (and divestment), Support for restrictive (policy) measures, Support for (economic) push measures, Support for (economic) pull measures, Biospheric values (biospheric value orientation), Personal norm (for climate-change mitigation). Pearson correlation coefficients were computed. * p < 0.05. ** p < 0.01. *** p < 0.001.

.

3.4.2. Regression Analyses

Building on these bivariate results, we computed hierarchical regression models, in which we successively added the two heuristics as predictors (see

Table A25. Linear regression models predicting climate-change-mitigation behaviors by the two heuristics.

Criterion: Curtailment behavior (n = 1110)
	Step 1	Step 2a	Step 2b	Step 3
Predictors	β (95% CI of B)	β (95% CI of B)	β (95% CI of B)	β (95% CI of B)
Age	0.166 *** (0.008, 0.016)	0.164 *** (0.008, 0.016)	0.174 *** (0.009, 0.016)	0.172 *** (0.008, 0.016)
Formal education	0.057 (−0.011, 0.266)	0.026 (−0.071, 0.188)	0.015 (−0.091, 0.16)	0.014 (−0.094, 0.156)
Household income	−0.034 (−0.209, 0.057)	−0.028 (−0.186, 0.062)	−0.059 * (−0.253, −0.013)	−0.053 (−0.239, 0.001)
Gender	−0.186 *** (−0.536, −0.282)	−0.188 *** (−0.532, −0.295)	−0.171 *** (−0.491, −0.262)	−0.174 *** (−0.497, −0.269)
Residence	−0.037 (−0.221, 0.047)	−0.045 (−0.23, 0.02)	−0.044 (−0.224, 0.017)	−0.045 (−0.226, 0.015)
Political orientation	0.165 *** (0.067, 0.139)	0.06 * (0.003, 0.072)	0.054 * (0.001, 0.067)	0.043 (−0.006, 0.061)
Restriction Heuristic		0.366 *** (0.361, 0.488)		0.103 ** (0.03, 0.208)
Optimization Heuristic			0.436 *** (0.492, 0.628)	0.362 *** (0.366, 0.564)
Adjusted R2	0.087	0.209	0.261	0.265
Criterion: Efficiency behavior (n = 1106)
	Step 1	Step 2a	Step 2b	Step 3
Age	−0.18 *** (−0.017, −0.009)	−0.181 *** (−0.017, −0.009)	−0.172 *** (−0.016, −0.008)	−0.172 *** (−0.016, −0.008)
Formal education	0.147 *** (0.193, 0.471)	0.123 *** (0.145, 0.41)	0.109 *** (0.119, 0.371)	0.109 *** (0.119, 0.371)
Household income	0.083 ** (0.054, 0.32)	0.087 ** (0.069, 0.322)	0.057 * (0.008, 0.249)	0.057 * (0.007, 0.249)
Gender	0.032 (−0.056, 0.197)	0.031 (−0.053, 0.188)	0.047 (−0.01, 0.219)	0.048 (−0.01, 0.22)
Residence	−0.057 (−0.268, 0.001)	−0.065 * (−0.281, −0.024)	−0.065 * (−0.275, −0.032)	−0.065 * (−0.275, −0.032)
Political orientation	0.102 *** (0.028, 0.099)	0.014 (−0.027, 0.044)	−0.006 (−0.037, 0.029)	−0.006 (−0.037, 0.03)
Restriction Heuristic		0.306 *** (0.29, 0.421)		−0.006 (−0.096, 0.083)
Optimization Heuristic			0.425 *** (0.479, 0.615)	0.429 *** (0.453, 0.651)
Adjusted R2	0.093	0.178	0.259	0.258
Criterion: Trying out new alternatives and practices (n = 1108)
	Step 1	Step 2a	Step 2b	Step 3
Age	*−0.08 ** (−0.009, −0.001)	−0.081 ** (−0.009, −0.002)	−0.074 ** (−0.009, −0.001)	−0.074 ** (−0.009, −0.001)
Formal education	0.144 *** (0.175, 0.437)	0.119 *** (0.127, 0.379)	0.107 *** (0.105, 0.349)	0.107 *** (0.104, 0.348)
Household income	−0.067 * (−0.269, −0.017)	−0.062 * (−0.253, −0.012)	−0.089 ** (−0.305, −0.071)	−0.087 ** (−0.302, −0.067)
Gender	−0.135 *** (−0.4, −0.16)	−0.137 *** (−0.399, −0.169)	−0.122 *** (−0.364, −0.141)	−0.123 *** (−0.367, −0.143)
Residence	0.042 (−0.034, 0.22)	0.036 (−0.041, 0.202)	0.036 (−0.037, 0.198)	0.036 (−0.038, 0.198)
Political orientation	0.157 *** (0.059, 0.126)	0.075 * (0.011, 0.078)	0.063 * (0.005, 0.069)	0.059 * (0.002, 0.068)
Restriction Heuristic		0.287 *** (0.251, 0.375)		0.032 (−0.053, 0.122)
Optimization Heuristic			0.373 *** (0.386, 0.519)	0.350 *** (0.328, 0.521)
Adjusted R2	0.085	0.159	0.212	0.212
Criterion: Political consumption and divestment (n = 1104)
	Step 1	Step 2a	Step 2b	Step 3
Age	−0.059 (−0.008, 0)	−0.062 * (−0.008, −0.001)	−0.051 * (−0.007, 0)	−0.051 * (−0.007, 0)
Formal education	0.153 *** (0.202, 0.476)	0.118 *** (0.137, 0.387)	0.101 *** (0.109, 0.337)	0.1 *** (0.109, 0.337)
Household income	0.07 * (0.024, 0.287)	0.077 ** (0.052, 0.291)	0.039 (−0.023, 0.195)	0.04 (−0.022, 0.197)
Gender	−0.033 (−0.196, 0.055)	−0.034 (−0.187, 0.042)	−0.011 (−0.127, 0.081)	−0.011 (−0.129, 0.08)
Residence	−0.015 (−0.167, 0.098)	−0.023 (−0.175, 0.067)	−0.024 (−0.166, 0.054)	−0.024 (−0.166, 0.054)
Political orientation	0.215 *** (0.097, 0.167)	0.096 ** (0.025, 0.092)	0.072 ** (0.014, 0.074)	0.07 ** (0.012, 0.074)
Restriction Heuristic		0.414 *** (0.411, 0.534)		0.017 (−0.062, 0.101)
Optimization Heuristic			0.558 *** (0.645, 0.769)	0.545 *** (0.601, 0.781)
Adjusted R2	0.089	0.244	0.374	0.374
Criterion: Activism (n = 1102)
	Step 1	Step 2a	Step 2b	Step 3
Age	−0.178 *** (−0.018, −0.009)	−0.179 *** (−0.018, −0.009)	−0.170 *** (−0.017, −0.009)	−0.170 *** (−0.017, −0.009)
Formal education	0.089 ** (0.064, 0.358)	0.065 * (0.013, 0.295)	0.051 (−0.014, 0.255)	0.051 (−0.014, 0.256)
Household income	−0.042 (−0.24, 0.042)	−0.037 (−0.222, 0.047)	−0.066 * (−0.285, −0.028)	−0.066 * (−0.286, −0.028)
Gender	−0.035 (−0.216, 0.053)	−0.036 (−0.212, 0.045)	−0.02 (−0.169, 0.076)	−0.02 (−0.168, 0.077)
Residence	0.003 (−0.135, 0.15)	−0.004 (−0.146, 0.127)	−0.004 (−0.139, 0.12)	−0.004 (−0.139, 0.121)
Political orientation	0.205 *** (0.098, 0.173)	0.120 *** (0.042, 0.117)	0.100 *** (0.03, 0.101)	0.100 *** (0.03, 0.102)
Restriction Heuristic		0.297 *** (0.294, 0.433)		−0.005 (−0.102, 0.09)
Optimization Heuristic			0.411 *** (0.486, 0.633)	0.415 *** (0.458, 0.67)
Adjusted R2	0.093	0.173	0.248	0.247
Criterion: Support for restrictive policy measures (n = 1114)
	Step 1	Step 2a	Step 2b	Step 3
Age	0.053 (0, 0.007)	0.05 (0, 0.006)	0.063 * (0.001, 0.007)	0.059 * (0.001, 0.007)
Formal education	0.122 *** (0.118, 0.358)	0.078 ** (0.049, 0.256)	0.067 * (0.031, 0.231)	0.064 * (0.027, 0.223)
Household income	−0.053 (−0.22, 0.011)	−0.044 (−0.184, 0.014)	−0.085 ** (−0.261, −0.07)	−0.072 ** (−0.234, −0.046)
Gender	0.004 (−0.103, 0.117)	0.001 (−0.092, 0.097)	0.023 (−0.047, 0.135)	0.017 (−0.058, 0.121)
Residence	0.074 * (0.036, 0.268)	0.063 * (0.029, 0.228)	0.065 ** (0.037, 0.229)	0.063 ** (0.034, 0.223)
Political orientation	0.265 *** (0.113, 0.175)	0.119 *** (0.037, 0.092)	0.124 *** (0.041, 0.094)	0.101 *** (0.029, 0.081)
Restriction Heuristic		0.513 *** (0.466, 0.568)		0.221 *** (0.153, 0.293)
Optimization Heuristic			0.559 *** (0.571, 0.679)	0.400 *** (0.37, 0.524)
Adjusted R2	0.097	0.335	0.384	0.404
Criterion: Support for economic push measures (n = 1114)
	Step 1	Step 2a	Step 2b	Step 3
Age	−0.095 ** (−0.012, −0.003)	−0.098 *** (−0.012, −0.004)	−0.085 ** (−0.011, −0.003)	−0.090 *** (−0.011, −0.003)
Formal education	0.181 *** (0.306, 0.614)	0.137 *** (0.217, 0.481)	0.127 *** (0.195, 0.451)	0.124 *** (0.188, 0.439)
Household income	0.009 (−0.125, 0.171)	0.019 (−0.079, 0.174)	−0.022 (−0.178, 0.067)	−0.008 (−0.141, 0.1)
Gender	−0.012 (−0.17, 0.112)	−0.014 (−0.155, 0.086)	0.008 (−0.098, 0.135)	0.001 (−0.113, 0.116)
Residence	0.081 ** (0.067, 0.365)	0.07 ** (0.058, 0.312)	0.072 ** (0.068, 0.315)	0.070 ** (0.064, 0.306)
Political orientation	0.221 *** (0.117, 0.196)	0.076 ** (0.018, 0.089)	0.083 ** (0.025, 0.092)	0.059 * (0.008, 0.075)
Restriction Heuristic		0.511 *** (0.604, 0.734)		0.231 *** (0.213, 0.392)
Optimization Heuristic			0.551 *** (0.73, 0.869)	0.384 *** (0.458, 0.656)
Adjusted R2	0.118	0.355	0.396	0.419
Criterion: Support for economic pull measures (n = 1114)
	Step 1	Step 2a	Step 2b	Step 3
Age	0.015 (−0.002, 0.004)	0.013 (−0.002, 0.004)	0.024 (−0.002, 0.004)	0.022 (−0.002, 0.004)
Formal education	0.120 *** (0.098, 0.311)	0.084 ** (0.046, 0.241)	0.070 * (0.027, 0.212)	0.068 * (0.025, 0.209)
Household income	−0.026 (−0.146, 0.058)	−0.018 (−0.124, 0.063)	−0.054 * (−0.181, −0.004)	−0.048 (−0.171, 0.006)
Gender	−0.045 (−0.173, 0.022)	−0.047 (−0.167, 0.011)	−0.027 (−0.13, 0.039)	−0.03 (−0.135, 0.034)
Residence	0.007 (−0.091, 0.115)	−0.003 (−0.099, 0.089)	−0.002 (−0.092, 0.086)	−0.003 (−0.094, 0.084)
Political orientation	0.222 *** (0.078, 0.133)	0.103 *** (0.023, 0.075)	0.094 *** (0.02, 0.069)	0.084 ** (0.015, 0.064)
Restriction Heuristic		0.417 *** (0.319, 0.415)		0.099 ** (0.021, 0.153)
Optimization Heuristic			0.507 *** (0.445, 0.545)	0.436 *** (0.353, 0.498)
Adjusted R2	0.066	0.223	0.302	0.306

Note. β = standardized regression coefficient, 95% CI of B = 95% confidence interval of the unstandardized regression coefficient. Coding of the sociodemographic variables: Formal education (lower = 0; higher = 1), household income (lower = 0; higher = 1), gender (female = 0; male = 1), residence (rural = 0; urban = 1), higher numbers on political orientation reflect a more left political orientation. * p < 0.05. ** p < 0.01. *** p < 0.001.

in Appendix B for an overview). In a first step, we only added sociodemographic variables. In a second step, we added either the restriction (step 2a) or the optimization heuristic (step 2b), and in a third step we included both heuristics.

Despite their high intercorrelation, the conventional indicators of multicollinearity—variance inflation factor (VIF) and tolerance—did not indicate a high level of multicollinearity, 2.22 < VIF < 2.23; 0.44 < tolerance < 0.46; [70]. Both heuristics explained a significant amount of variance in all criteria in addition to the sociodemographic variables ($0.074 \le {\mathsf{\Delta }}_{R_{step 2a}^2-R_{step 1}^2}\left(restriction\right) \le 0.238$; $0.155 \le {\mathsf{\Delta }}_{R_{step 2b}^2-R_{step 1}^2}\left(optimization\right) \le 0.287$). When the restriction heuristic was added first and the optimization heuristic second, the adjusted R² in the third step increased for all criteria compared to the second step ($0.053 \le {\mathsf{\Delta }}_{R_{step 3}^2-R_{step 2a}^2}\left(restriction\right) \le 0.130$; 0$.212 \le R_{step 3}^2\le 0.419$). This was not the case when the heuristics were added to the model in the reverse order ($-0.001 \le {\mathsf{\Delta }}_{R_{step 3}^2-R_{step 2b}^2}\left(optimization\right) \le 0.023$). Furthermore, the optimization heuristic was the more important predictor of all climate-change-mitigation intentions ($-0.006 \le {\beta }_{step 3}\left(restriction\right)\le 0.231$; $0.350 \le {\beta }_{step 3}\left(optimization\right)\le 0.545$; see also 95% confidence intervals in Table A25 in Appendix B). For the regression models on efficiency, new alternatives, political consumption, and activism, the restriction heuristic was not a significant predictor in the third step.

3.5. Intraindividual Dominance of the Restriction Heuristic and Climate-Change-Mitigation Intentions

Furthermore, we computed an ipsative scale score by subtracting the manifest scale score of the optimization heuristic from the restriction heuristic, thus reflecting the intrapersonal difference in the expression of both heuristics. A positive score reflects that the restriction heuristic is more dominant than the optimization heuristic within a participant. This dominance of restriction was also correlated with sociodemographic variables, biospheric values, personal norm, and intentions (see Table 3).

The following pattern emerged from the analyses with the dominance of restriction. The means of the two formal educational groups were not significantly different, t(1425) = 1.309, p = 0.191, and the correlation with age was not significant. Participants with a higher income had on average a comparatively stronger optimization than restriction heuristic (M = −0.06) than participants with a lower income, who tended to have a stronger restriction than optimization heuristic, (M = 0.06), t(1321) = 3.308, p = 0.001. Political orientation was positively correlated with the dominance of restriction (r = 0.09, p = 0.002); thus, being more left was associated with having a higher restriction than optimization heuristic.

The dominance of restriction was not significantly correlated with biospheric values, personal norm, curtailment, and the support for push as well as pull measures. In contrast, the dominance of restriction was negatively correlated with efficiency (r = −0.09, p = 0.001), new alternatives (r = −0.06, p = 0.026), political consumption (r = −0.11), and activism (r = −0.07, p = 0.004). Hence, the stronger the restriction heuristic compared to the optimization heuristic within the participants (an increasing intrapersonal difference between the two heuristics), the lower the intention for efficiency behavior, usage of new alternatives, political consumption, and activism on average. For the support of restrictive policy measures, the association was reversed (r = 0.05, p = 0.049). An increasing intrapersonal tendency toward a restriction heuristic rather than an optimization heuristic was positively correlated with the support for restrictive policy measures.

Table 4. Hypotheses on the relationships and actual correlations between the dominance of the restriction heuristic (ipsative measure) and the climate-change-mitigation intentions.

Type of Behavior (Intention)	Hypotheses about the Relevance of Restriction (R) and Optimization (O) Heuristic for the Intentions	Correlation with Dominance of Restriction	Evidence for Hypothesis
Curtailment	H1.3: R > O	−0.02	not significant
Efficiency	H2.3: O > R	−0.09 **	As expected
New alternatives	H3.3: O > R	−0.06 *	As expected
Political consumption	H4.3: O > R	−0.11 ***	As expected
Activism	No hypothesis	−0.07 **
Support for restrictive measures	H5.3: R > O	0.05 *	As expected
Support for push measures	H6.3: O > R	−0.05	not significant
Support for pull measures	H7.3: O > R	0.04	not significant

* p < 0.05. ** p < 0.01. *** p < 0.001.

provides an overview of the results of the analyses of the dominance of restriction with respect to its relationships with the climate-change-mitigation intentions and the respective hypotheses.

4. Discussion

The aim of the present study was to introduce a new heuristic that might be supportive for the more impactful climate-change-mitigation behaviors and goes beyond the common heuristic of mere restriction. We assumed that different climate-change-mitigation behaviors (measured through the intention to engage in them) might be rooted in different heuristics. Against the background of the ongoing discourse on how to reach the 1.5 °C degree goal, we developed a measure that captures two distinct yet strongly related heuristics for climate-change-mitigation behaviors. One heuristic covers the more conventional restriction heuristic, which is rooted in a concern about limited resources and overconsumption. The other heuristic focuses on optimization, representing an active seeking of solutions and an openness to various strategies.

We conducted an online survey study with a sample of N = 1427 participants representative of the German population with regard to age, gender, and education. The results of the study supported our postulated two heuristics being significantly associated with climate-change-mitigation motivation. As expected, the two heuristics turned out to be highly correlated but distinct constructs. We thus were successful in identifying a new heuristic.

On a descriptive level, results indicate that both heuristics were about equally represented in our sample. It thus has to be questioned whether the restriction heuristic should be considered as the dominant heuristic in climate-change-mitigation discourse. Furthermore, both heuristics were highly correlated with biospheric value orientation. However, opposed to what we expected, the correlation with biospheric value orientation was stronger for the optimization heuristic than for the restriction heuristic.

As expected, the eight types of climate-change-mitigation intentions assessed were of different prevalences in our sample. Based on self-report, curtailment was the behavior type associated with the highest behavioral intention, while political activism and efficiency behavior were the least intended behaviors. Thus, it was confirmed that curtailment behaviors form the most dominant type, whereas more impact relevant behaviors might be performed less often (at least the intention to show them is lower). We furthermore found that—with the exception of the support for restrictive policy measures—the optimization heuristic had significantly higher correlations with all types of climate-change-mitigation intentions. We thus assume that the optimization heuristic supports not only an openness to impactful and new climate-change-mitigation behaviors, but seems to be the more relevant heuristic for climate-change mitigation in general. This assumption was supported by two findings: (a) the optimization heuristic had a significantly higher correlation with the personal norm for climate-change mitigation; and (b) the optimization heuristic was the dominant predictor of all types of climate-change-mitigation intentions (except support of regulatory measures). To investigate the assumed possible negative effect of the restriction heuristic on the different intentions, we went a step further and directly investigated the dominance of the restriction heuristic on an individual, ipsative level.

4.1. Role of Dominance of the Restriction Heuristic

To test our claim that a mere restriction heuristic may not be sufficient on the individual level with respect to the newer and more impactful climate-change-mitigation behaviors, we directly investigated the associations of a dominant restriction heuristic. We computed an ipsative measure by subtracting the manifest scale score for the optimization heuristic from the restriction heuristic and analyzed whether this dominance interferes with the intention to engage in certain new climate-change-mitigation behaviors, particularly with regard to efficiency behavior and the support for policy measures (see Section 1.3. Present Study).

Five of the eight types of climate-change-mitigation intention were significantly related to the dominance of the restriction heuristic, all in line with the predictions. A dominant restriction heuristic was negatively correlated with the intention for efficiency behavior (e.g., investing in insulating one’s house), trying out new behaviors (e.g., sharing), and political consumption (e.g., switching to a social-ecological bank). It however was positively associated with the support of regulatory policy instruments (e.g., a ban on gas and oil heating in new buildings). Contrary to our hypotheses, a dominant restriction heuristic was neither significantly correlated with the intention for curtailment behaviors nor with support of economic policy measures (neither push nor pull). The dominant restriction heuristic was also not correlated with the biospheric value orientation nor the personal norm for climate-change mitigation. Thus, it seems that a dominant restriction heuristic is an independent factor that does not interfere with the known most relevant factors supporting climate-change-mitigation behaviors. However, the dominant restriction heuristic emerged as a relevant negative correlate for the intention of engaging in newer and more impactful private-sphere behaviors.

Although we did not find significant differences in the relationships between the two heuristics and sociodemographic variables (age, political orientation), we found such relationships for the dominant restriction heuristic: on average, participants with a lower income and those who located themselves more on the left side of the political spectrum reported a significantly higher restriction than optimization heuristic.

Our results thus add to the current research tradition on the motivation of various climate-change-mitigation behaviors. The restriction heuristic, if stronger than the optimization heuristic, could become a barrier to newer and more impactful climate-change-mitigation behaviors. Our findings thus give hints that common heuristics from the domain of environmental protection can be detrimental in the field of climate-change mitigation.

We are aware that the presented findings and interpretations cannot be generalized beyond the German population and need to be interpreted in the context of the ongoing climate change debates. Since March 2019, there have been calls for and intense discussions about the need for stricter policies to achieve the required carbon reduction goals in Germany. Economic instruments, such as a carbon taxes and subsidies for efficient heating or individual mobility, have received broad support, yet adverse distributional effects have repeatedly been discussed in the media. Subsidies are more available to high-income groups and homeowners, particularly with regard to high upfront costs for e-cars or for private renewable energy systems. Simultaneously, a carbon tax increases energy prices and thereby burdens lower income households to a greater extent since they have to spend a larger share of their income on energy. On the other hand, restrictive policy measures, such as a ban on fossil heating systems, primarily affect and threaten homeowners. It is, therefore, not surprising that a dominant restriction heuristic and support for restrictive policy measures were associated with a left political orientation and a lower income.

However, studies from other countries have observed an association between pro-environmental attitudes, particularly the biospheric value orientation, and a more liberal and less conservative orientation. We therefore assume that participants with a dominant restriction heuristic might be a possibly widespread type of the more “traditional” environmentalists who tend to be more liberal and left-oriented and who probably have been actively mitigating climate change for a long time. According to our data, they tend to support policy measures in general, and restrictive measures in particular. They seem to show mitigation behavior on the individual household level; however, the more dominant the restriction heuristic, the more they oppose to the more impactful private-sphere behaviors.

4.2. Limitations

The difficulty of integrating the somewhat contradictory findings for policy support illustrates an important limitation of our study: the national focus. We studied a sample that was representative of and thus restricted to the German population. However, several studies have shown that, in many cases, the psychological structure behind pro-environmental behaviors can been transferred to other European countries and the U.S. context, see, e.g., [71]. Still, our findings would be more convincing if we had applied a cross-national perspective.

5. Conclusions

To successfully mitigate climate change, we need to grow beyond the heuristic of restriction, in particular when it comes to the domain of individual climate-change-mitigation behavior. In contrast to a mere restriction heuristic, an optimization heuristic might more strongly support the whole width of individual behaviors. Although the restriction and optimization heuristic were highly correlated, we were able to show that a dominant restriction heuristic—measured on an ipsative level—counteracts the intention to show certain relevant climate-change-mitigation behaviors, such as efficiency behaviors, new political behaviors, or trying out new alternatives. According to the latest IPCC report, WG 3 [12]; these are, however, the behaviors that are most powerful for supporting climate-change mitigation on the level of household consumption. It thus seems worthwhile to develop, e.g., neighborhood programs, or TV formats that support the active reflection on the range of impact of the various private-sphere behaviors. Knowledge about the impact of new climate-change-mitigation behaviors or household investment could transform a dominant restriction heuristic into one that is open to impact-relevant climate mitigation strategies.

Considering these findings, it seems worthwhile to reflect on the restriction-oriented narratives that seem to dominate the current debates. The complexity and urgency of a transformation toward sustainable and climate-neutral societies require more of citizens than the simple heuristic of restriction and renunciation. While this heuristic seems to be of value on a macro level, and thus for global political behavior, it seems to be insufficient for the more impactful and newer private-sphere behaviors like investing in efficient technologies or divestment behaviors. Our findings thus point out the need for education and new formats of information that can empower individuals who nurse a dominant restriction heuristic to take on an informed transformation perspective.