Persuasiveness, importance and novelty of arguments about Carbon Capture and Storage

Abstract Carbon Capture and Storage (CCS) is a promising technology for reducing carbon emissions, but the public is often reluctant to support it. To understand why public support is lacking, it is crucial to establish what citizens think about the arguments that are used by proponents and opponents of CCS. We determined the persuasiveness, importance and novelty of 32 arguments for and against CCS using a discrete choice experiment in which respondents made consecutive choices between pairs of pro or con arguments. We used latent class models to identify population segments with different preferences. The results show that citizens find arguments about climate protection, which is the primary goal of CCS, less persuasive than other arguments, such as normative arguments (for example ‘a waste product such as CO 2 should be disposed of properly’) or arguments about benefits of CCS for energy production and economic growth. This discrepancy complicates communication that aims to convince citizens of the benefits of CCS for climate protection.


Introduction
Climate change mitigation requires substantial modifications to energy production and consumption patterns. Yet, the technologies needed to change these patterns often lack public acceptance (Wustenhagen et al., 2007). Carbon Capture and Storage (CCS) is such a technology. CCS involves capturing CO 2 at a large emission source (e.g. a power plant or factory), transporting the CO 2 to a storage location (e.g. a natural gas field) and injecting the CO 2 into a rock formation for permanent storage (see Reiner, 2016 for an overview of recent CCS developments). CCS is a critical component of climate change mitigation strategies as fossil fuel consumption is increasing and carbon-intensive industries remain prominent (IPCC, 2014). If CCS is to become a viable option policy makers and industry must encourage its development (IEA, 2013;Scott et al., 2012). However, the public is reluctant to support this technology (De Best-Waldhober et al., 2012;L'Orange Seigo et al., 2014b;Upham and Roberts, 2011). This discourages stakeholders, such as energy or industrial firms, policy makers and NGOs, from moving toward large-scale implementation .
Stakeholders need to communicate with citizens to build support for CCS (Ashworth et al., 2010).
Existing studies offer comprehensive guidelines for effective communication processes (see Brunsting et al., 2011;L'Orange Seigo et al., 2014a) for a review of CCS communication studies). Yet, citizens' reactions to the content of stakeholder's messages are partially understood. This hampers communications efforts (Reiner, 2008). Studies into message content focus primarily on neutral, descriptive information. Examples are studies into monitoring information (L'Orange Seigo et al., 2011), storage terminology (Ha-duong et al., 2009), figures (L'Orange Seigo et al., 2013), labels (Van Rijnsoever et al., 2015), natural analogues to CO 2 storage (Tokushige et al., 2007a), entities responsible for managing risk (Sharp et al., 2009), basic properties of CO 2 and CCS (Dowd et al., 2014;Tokushige et al., 2007b;Wallquist et al., 2011) or different sets of CO 2 capture and storage technologies (De Best-Waldhober et al., 2012Wallquist et al., 2012). Such information is unlikely to foster substantial support for the stakeholder's opinion, unless it is reinforced with arguments that resonate with the values of citizens (Kahan et al., 2012). Recent studies tackled this issue by also showing which positive or negative characteristics of CCS significantly affect citizen's attitude toward CCS (De Best-Waldhober et al., 2012Kraeusel and Möst, 2012;Oltra et al., 2012;Tokushige et al., 2007b;Wallquist et al., 2011). Despite this progress, three issues remain largely unaddressed.
First, positive or negative characteristics comprise only a subset of the arguments communicated by stakeholders (see Boyd and Paveglio, 2014;Buhr and Hansson, 2011;van Egmond and Hekkert, 2012) for an overview). Stakeholders also use counterarguments (e.g. CCS is not necessary for climate change mitigation), analogies (e.g. CCS is safe, just as natural gas storage is safe; see Tokushige et al., 2007a), or arguments that appeal to norms (e.g. a waste product such as CO2 should be disposed of properly; see Cialdini, 2003). None of the existing studies investigated this broader range of CCS arguments.
Second, existing studies often ignore heterogeneity among citizens by only presenting average opinions (see Allenby and Rossi, 1999) for an overview of the concept). Citizens have diverse reactions to communication about energy technologies (Van Rijnsoever et al., 2015). Arguments that most citizens find irrelevant might be important to a particular population segment. Understanding heterogeneity facilitates the design of segmented communication materials.
Third, existing studies rarely examine message effectiveness beyond persuasiveness or attitude change. Yet, attitude change can be unstable and short-lived or stable and long lasting. Dual processing models suggest that stable attitudes require elaborate or systematic processing (see Chen and Chaiken, 1999;Petty and Wegener, 1999 for an overview). Citizens will process information in depth if they are motivated and knowledgeable about the topic in question. They will therefore likely not scrutinize unimportant or new arguments, but will resort to cognitive shortcuts instead, leading to less stable opinions. A communicator attempting to encourage the audience to adopt a specific, stable opinion should select arguments that the audience perceives as persuasive, important and are not completely novel to them. It is therefore important to include importance and novelty in studies into message effects.
We address these shortcomings by eliciting the perceived persuasiveness, importance and novelty of 16 pro and 16 con CCS arguments for different population segments. To this end, we asked citizens to make eight consecutive choices between two arguments in a discrete choice experiment (DCE). By exploring the persuasiveness, importance and novelty of arguments we advance understanding of citizens' reactions to the content of stakeholder's messages. Our results help to improve communication strategies for CCS. They are also insightful for energy technologies with similar public acceptance issues.

Methods
We elicit the perceived persuasiveness, importance and novelty of arguments by asking a sample of citizens to make eight consecutive choices between two arguments in a discrete choice experiment (DCE) (see Amaya-Amaya et al., 2008) for an overview of DCEs) that was included in an online survey. Other CCS studies used DCEs to identify the importance of technological or economic characteristics of CCS, such as price and amount of CO 2 -emission reductions (Kraeusel and Möst, 2012;Sharp et al., 2009;Wallquist et al., 2012). To the best of our knowledge, DCEs have not yet been used to study arguments. The development of technology for CO 2 storage contributes to employment and economic growth Economic benefits P9 CO 2 storage is cheaper than solar or wind energy in the medium to long term Relatively cheap P10 The Netherlands has a good starting position because of its experience with natural gas Natural gas experience P11 Other countries have used technologies for CO 2 storage safely for many years Used in other countries P12 CO 2 storage is already being used to recover more oil from oilfields Enhanced Oil Recovery P13 CO 2 storage is safe. CO 2 is stored in natural gas fields where natural gas was stored for millions of years Safety of natural gas fields P14 CO 2 storage uses less space than solar panels or wind turbines Space requirements P15 Gas or coal plants with CO 2 storage are a stable supplement to the inconsistent supply of solar and wind energy Stable energy supply P16 A waste product such as CO 2 should be disposed of properly Dispose of CO 2 garbage No.

C1
The climate problem can be tackled without CO 2 storage Unnecessary for climate problem C2 CO 2 storage promotes the use of new coal-fired power plants Promotes coal C3 CO 2 storage is more expensive than solar or wind energy in the long term Relatively expensive C4 It is not certain that there will be a return on large investments in CO 2 storage Investment uncertainty C5 Storage sites for CO 2 have to be monitored indefinitely Indefinite monitoring C6 Real estate prices near CO 2 storage facilities may fall Falling real estate prices C7 CO 2 storage detracts from the development of renewable energy Detracts from renewables C8 Electricity bills will be higher because of CO 2 storage Higher electricity bills C9 CO 2 storage is new and has never been applied on a large scale, so the risks are not fully understood Risks not fully understood C10 It is better to avoid generating CO 2 than to store the CO 2 Avoid generating CO 2 C11 If a lot of CO 2 leaks on a windless day, a suffocating cloud of CO 2 could be created Suffocation C12 Groundwater might become acidified if CO 2 were to leak out of an underground pipeline Groundwater acidification C13 CO 2 storage can cause small earthquakes, comparable to those caused by natural gas extraction Earthquakes C14 Hazardous chemicals are used in the capture of CO 2 . Hazardous chemicals C15 Power plants with CO 2 storage require 10-40% more energy Energy requirements C16 There is little public support for CO 2 storage Lack of public support Note: The arguments refer to 'CO 2 storage', because the Dutch media use this term instead of 'Carbon Capture and Storage'.

Argument selection
We selected the arguments from the pool of arguments used in public debate on CCS in the Netherlands, identified in a previous study (Van Egmond and Hekkert, 2012). We only included arguments that are perceived as valid by experts (i.e. common misconceptions were excluded), straightforward enough to be written down clearly in one or two sentences, and refer to the use of CCS in general, rather than specific policies (e.g. mandatory CCS at power plants). We consulted a panel of CCS experts from academia, knowledge institutes and industry, as well as communication experts, to construct a set of the most prominent 16 pro and 16 con arguments (see Table 1; labels are included for ease of reference). The consultation consisted of a workshop about the goals and design of the study and feedback on the concept survey.

Sample and data collection
We collected data by using a Dutch, national, online marketing panel (n = 920). The sampling procedure used quotas for age, gender, education level and state of residence to ensure that the sample was representative of the adult (i.e. at least 18 years) Dutch population. Respondents in the sample are slightly older (M = 51.66 years; SD = 13.41) and slightly more likely to be female (53.4%), highly educated (37,5%) and to live in the south of the Netherlands (31.9%). To control for these differences, we included a weight factor in the analysis based on these characteristics. Panel members received compensation for their participation, they were assured of the anonymity of the results and they were debriefed at the end of the survey.

Discrete choice experiment
The respondents were introduced to the goal of the DCE in the beginning of the survey. They read the following description: "The use of CO 2 -storage is being considered in the Netherlands, as well as abroad. To make a decision, a trade-off is being made between arguments pro and con CO 2 -storage. We therefore find your opinion about these arguments very important".
Half of the respondents in the experiment then chose between eight consecutive pairs of pro arguments (n = 465), while the other half chose between con arguments (n = 455). We investigated pro and con arguments independently to control for differences in their persuasiveness, importance and novelty. Since citizens generally give greater weight to negative objects or events (Rozin and Royzman, 2001), con arguments are often more salient and persuasive than pro arguments (Cobb and Kuklinski, 1997;Sen and Lerman, 2007;Skowronski and Carlston, 1987).
The experimental design included every combination of two pro or two con arguments. We divided the 240 choice sets into 30 blocks to reduce the number of choice sets per respondent to eight. Respondents were randomly assigned to a survey version. The experimental design was generated using the software package Ngene. Fig. 1 displays an example choice set.

Pre-and post-test of CCS attitude
Before and after the DCE respondents indicated their agreement with three statements using five-point Likert items ('totally disagree' (1) to 'totally agree' (5)): 'I am positive about CO 2 storage', 'CO 2 storage is dangerous', and 'CO 2 storage is useful'. We averaged the scores for the three items to construct an indicator for attitude toward CCS. The reliability of the attitude scale is adequate (Cronbach's a; before = 0.76, after = 0.81).

Data analysis
We first tested whether the range of arguments to which an individual was exposed had any effect on attitude towards CCS. To this end, we compared the means of the pre-and post-test of CCS attitude using paired sample t-tests. We then estimated conditional logit models and latent class models (Vermunt and Magidson, 2002) using the software package Latent Gold 5.0. We estimated separate models for persuasiveness, importance and novelty using choice as a binary dependent variable and the arguments as independent, nominal variables. The conditional logit models are regression models for the probability that respondent i selects alternative m at replication t, given the values of the attributes of the alternatives (z att it ). The conditional logit model therefore has the following form: where y it denotes the value of the binary dependent variable and M denotes the number of alternatives. In our models h mjz it is a linear function of the attribute effects (b att p ) and an alternative specific where the p index refers to a particular attribute. Each of our models included a nominal attribute with sixteen levels that represents the arguments. This attribute was effects coded in the model, which means that the parameters of the levels sum to zero. The alternative specific constant controls for whether the alternative was on the right or left of the choice set. A latent class model extends this model by assigning respondents that make similar choices to the same segment. A categorical latent variable captures the segment membership ðxÞ of each respondent. The model includes separate parameters for each latent segment. The latent class model therefore has the following form: The linear function h mjx;z it is As we included sampling weights based on the sociodemographic characteristics of the respondents, Pseudo ML (PM) estimation is used to estimate the parameters in the model (see Vermunt and Magidson, 2005 for additional information on the exact estimation methods). We explored models consisting of one to four latent segments. We used respondents' choices, socio-demographics and attitudes before the experiment to identify segments. In line with best practice for LCA (Nylund et al., 2007), we selected the models with the lowest Bayesian Information Criterion (BIC).
Attitudes changed significantly after exposure to pro arguments (M = 3.18, SD = 0.76, t = 6.92, p < 0.001) and con arguments (M = 2.71, SD = 0.75, t = 8.78, p < 0.001). This implies that making consecutive choices between pairs of pro and con CCS arguments had a small, but significant effect on the attitude of respondents towards CCS. Con arguments had a slightly stronger effect on attitude (D = 0.21) than pro arguments (D = 0.15). The difference in size is significant at the 0.10 level (t = 1.89). As mentioned in Section 2.3, this difference is to be expected due to the higher salience of negative information in con arguments.

Conditional logit models
The conditional logit parameters reveal the perceived persuasiveness, importance and novelty of CCS arguments. In the interest of brevity, we will refer to persuasiveness rather than perceived persuasiveness in the results section. The parameters are based on effects coding, which means that the models compare the persuasiveness, importance and novelty of an argument to the average persuasiveness, importance or novelty of all arguments (see Table 2). The explanatory value of the models is adequate (McFadden R 2 = 0.05-0.11) (Louviere et al., 2000). The models predict between 59.6% and 64.9% of all respondents' choices correctly. Although respondents are indifferent between some pairs of arguments, the best and worst arguments have substantial predicted probabilities (lowest = 29% (P3), highest = 79% (C10)).
There is a strong positive correlation between the parameters of persuasiveness and importance (pro: r = 0.92 and con: r = 0.93). Important arguments are therefore likely to be persuasive. There is a moderate negative correlation between the parameters of persuasiveness and novelty (pro: r = À0.34 and con: r = À0.41) and importance and novelty (pro: r = À0.46 and con: r = À0.29). New arguments are therefore likely to be unpersuasive and unimportant. The following section discusses the results per dependent variable.
Persuasiveness of pro arguments: Respondents find six pro arguments persuasive. The most persuasive argument by far is P6 (industrial applications), followed by P16 (dispose of CO 2 garbage) and P13 (safety of natural gas fields) after a substantial drop in persuasiveness. Other persuasive arguments are P8 (economic benefits), P10 (natural gas experience) and P2 (international climate agreements). Respondents find five pro arguments unpersuasive. The least persuasive argument by far is P3 (lifestyle changes), followed by P4 (set an example) and P14 (space requirements) after a substantial jump in persuasiveness. Other unpersuasive arguments are P7 (cheap coal) and P12 (enhanced oil recovery). Importance of pro arguments: Five pro arguments are important to respondents. The most important arguments are P16 (dispose of CO 2 garbage) and P6 (industrial applications), followed at a considerable distance by P8 (economic benefits), P13 (safety of natural gas fields) and P1 (climate problem). Respondents find seven arguments unimportant. The least important argument is P3 (lifestyle changes), followed at a considerable distance by P12 (enhanced oil recovery) and P14 (space requirements). Other unimportant arguments are P4 (set an example), P11 (use in other countries), P7 (cheap coal) and P9 (relatively cheap). Novelty of pro arguments: Four pro arguments are new to respondents. In order of decreasing novelty, these arguments are P12 (enhanced oil recovery), P8 (economic benefits), P9 (relatively cheap) and P11 (use in other countries). Except for P8, all of these arguments are also unimportant and/or unpersuasive. Respondents find five arguments not new. In order of increasing novelty, these arguments are P2 (international climate agreements), P1 (climate problem), P16 (dispose of CO 2 garbage), P10 (natural gas experience) and P4 (set an example). Except for P4, all of these arguments are persuasive and/or important. Persuasiveness of con arguments: Respondents find three con arguments persuasive. The most persuasive argument by far is C10 (avoid generating CO 2 ), followed by C9 (risks not fully understood) and C1 (unnecessary for climate problem). The distance in score between these three arguments is substantial. Respondents find five con arguments unpersuasive. The least persuasive argument by far is C2 (promotes coal), followed by C15 (energy requirements) after a substantial jump in persuasiveness. Other unpersuasive arguments are C14 (hazardous chemicals), C8 (rising electricity bills) and C7 (detracts from renewables). Importance of con arguments: Three con arguments are important to respondents. The most important argument by far is C10 (avoid generating CO 2 ), followed at a considerable distance by C9 (risks not fully understood) and C1 (unnecessary for climate problem). Respondents find five con arguments significantly unimportant. The least important argument by far is C2 (promotes coal), followed at a considerable distance by C8 (rising electricity bills), C15 (energy requirements) and C6 (falling real estate prices). Another unimportant argument is C4 (investment uncertainty). Novelty of con arguments: Three con arguments are new to respondents. The newest argument is C14 (hazardous chemicals), followed at a considerable distance by C15 (energy requirements) and C11 (suffocation). Respondents find four con arguments not new. The least new argument is C10 (avoid generating CO 2 ), followed by C16 (lack of public support), C1 (unnecessary for climate problem) and C6 (falling real estate prices).
The conditional logit models show several patterns. First, citizens find arguments about climate change (P1, P2, C1) less persuasive and/or important than other arguments, even though climate change is the primary goal of CCS. Second, the most important pro and con arguments (P16, C10) use injunctive norms; they prescribe desirable actions by using the verb 'should'. Third, arguments about specific risks or the role of CCS in the energy mix are likely to be new, unpersuasive and unimportant. We return to these patterns in the discussion, but will first show how population segments differ from the average.

Latent class models
The latent class models illuminate differences between segments with respect to argument persuasiveness, importance and novelty (see Tables 3 and 4). The performance of the models is good (McFadden R 2 = 0.18-0.33) (Louviere et al., 2000). The models predict between 66.1% and 74.0% of all respondents' choices correctly. The substantial improvement in precision demonstrates the value of uncovering observed heterogeneity. We characterize population segments by identifying common themes, words or Note: displays parameters and significance level of z-test (*p < 0.05, **p < 0.01, ***p < 0.001). The significance test indicates whether the parameter is significantly different from the average of all parameters. concepts in the arguments that score significantly more or less than the average of arguments. We also tested whether there are significant differences between segments in socio-demographics or CCS attitude before the experiment (see Table 5). Only a few differences were significant, which could be caused by the small size of most of the segments.

Pro arguments
Persuasiveness of pro arguments: Segment 1 (85.8%) resembles the average opinion closely (i.e. the conditional logit model). Segment 2 (7.2%) instead focuses on the interconnection between CCS and other energy technologies. The arguments they find persuasive often refer (in) directly to nuclear, solar, wind or fossil fuel based energy, such as P3 (lifestyle changes), P5 (reduces need for nuclear), P7 (cheap coal), P9 (relatively cheap), P12 (enhanced oil recovery) and P15 (stable energy supply). Unlike segment 1, they find arguments about climate change mitigation (P1, P2), economic benefits (P8) and opportunities (P10), and safety (P13) unpersuasive. Segment 3 (7.1%) focuses on the affordability and security of energy supply, rather than an encompassing perspective on the energy mix. This is evidenced by the persuasiveness of P7 (cheap coal), P8 (economic opportunities), P9 (relatively cheap), and P15 (stable energy supply). Segment 3 finds the normative argument P16 (dispose of CO 2 garbage) unpersuasive. Importance of pro arguments: Segment 4 (93.1%) resembles the average opinion closely. Segment 5 (6.9%) focuses on the international position of the Netherlands concerning CCS and climate change, evidenced by the importance of P2 (international climate agreements), P4 (set an example) and P10 (natural gas experience). In contrast to segment 4, they find P1 (climate problem) and P5 (reduces need for nuclear) unimportant. Novelty of pro arguments: Segment 6 (84.9%) resembles the average opinion closely. Segment 7 (5.7%) instead focuses on past or desirable actions of CCS stakeholders, evidenced by the novelty of P2 (international climate agreements), P4 (set an example), P11 (use in other countries) and P16 (dispose of CO 2 garbage). They also find arguments P5 (reducing need for nuclear) and P6 (industrial applications) new. Segment 7 (M = 36.0 years) is significantly younger (F = 5.148, df = 472, p = 0.006) than segments 6 (M = 46.5 years) and 8 (M = 49.6 years). Segment 8 (9.5%) focuses on the safety and security effects of CCS in arguments about enhanced oil recovery (P12), storage in natural gas fields (P13) and beneficial effects on energy security (P15). Segments 7 and 8 are both relatively familiar with argument P1 (climate problem) and P3 (lifestyle changes).

Con arguments
Persuasiveness of con arguments: Segment 9 (90.3%) resembles the average opinion closely. Segment 10 (9.7%) instead focuses on risks that citizens living near storage locations are exposed to, evidenced by the persuasiveness of C6 (falling real estate prices), C11 (suffocation) and C13 (earthquakes). They also find C7 (detracts from renewables), C12 (groundwater acidification), C14 (hazardous chemicals), C15 (energy requirements) and C16 (lack of public support) unpersuasive. Importance of con arguments: In contrast to most other models, the majority is split between two segments. Segment 11 (52.5%) only finds C9 (risks not fully understood) important. They find arguments about the role of CCS in the energy mix unimportant, such as C2 (promotes coal), C3 (relatively expensive), C7 (detracts from renewables) and C15 (energy requirements). Segment 12 (40.1%) instead finds arguments C1 (unnecessary for climate problem) and C10 (avoid CO 2 emissions) important. In contrast to segment 11, they find arguments about costs to citizens and firms unimportant, such as C4 (investment uncertainty), C6 (falling real estate prices) and C8 (electricity prices). Segment 13 (7.4%) focuses on the costs of CCS relative to solar and wind energy (C3) and risks of groundwater acidification (C12) and earthquakes (C13). Like segment 12, they find argument C10 (avoid CO 2 emissions) important. Novelty of con arguments: The solutions for persuasiveness and importance (x 2 = 0.647; df = 2; p = 0.724), persuasiveness and novelty (x 2 = 5.653; df = 3; p = 0.130) and importance and novelty (x 2 = 8.333; df = 6; p = 0.215) are not significantly related.
The latent class models show that the majority of respondents (between 62.8% and 93.1%) perceives CCS arguments similarly. The majority is split in just one latent class model (importance of con arguments). A few small segments (between 5.7% and 23.8%) have distinct views on CCS arguments. Segments focus on the role of CCS in the energy mix (2), affordability and security of energy (3), the international position of the Netherlands concerning CCS and climate change (5), actions of CCS stakeholders (7), safety and security (8), local risks (10), CCS as an uncertain, new technology (11, 15), climate and norms (12), risks and costs (13,16), and risk and public support (17).

Discussion
This study presents the persuasiveness, importance and novelty of pro and con CCS arguments for segments of citizens. We address three gaps in the understanding of citizens' reactions to the content of stakeholders' communication. First, we expand the range of investigated arguments. We show that most citizens find arguments about climate change less persuasive and/or important than other arguments, even though climate change mitigation is the primary goal of CCS. They instead prefer arguments about particular norms, industrial applications of CCS, economic benefits or safety. Second, we uncover heterogeneity by showing how population segments differ from the majority. In contrast to the majority, segments focus on arguments about the role of CCS in the energy mix, the affordability and security of energy supply, specific risks to citizens living near storage locations or the international position of the Netherlands with regard to CCS. Third, we examine message effectiveness beyond persuasiveness and attitude change. We show that important arguments that are not new to citizens are more likely to be persuasive.
The results imply that stakeholders will struggle to convey the importance of CCS for climate change mitigation, unless they discuss additional benefits of various mitigation technologies, such as industrial applications, economic benefits or energy security and affordability. We would not suggest that stakeholders misrepresent the value of CCS by focusing exclusively on additional benefits. Rather, we want to highlight the discrepancy between the primary goal of CCS and preferences of citizens for other arguments and issues. As citizens also prefer normative arguments, stakeholders should incorporate norms into arguments about climate change and into their overall engagement strategies.
Uniform communication can be somewhat effective as the majority of citizens has similar opinions about CCS. Yet, a segmented communication approach can use distinct CCS storylines to engage population segments, which can be an important factor in establishing public acceptance for a technology. Our results provide a foundation for the construction of such storylines. Further research should account for heterogeneity, as substantial deviations from the average opinion will otherwise be overlooked. Although some arguments are CCS specific, most arguments are also applicable to other energy technologies. Our results therefore offer tentative conclusions about citizens' reactions to arguments about energy technologies with a similar public image, such as wind energy or shale gas.
Two limitations to this study raise issues for further research. First, we examined arguments in isolation; our analysis did not attempt to account for interactions between arguments and the source of the message (Eagly et al., 1978), for different frames (Bickerstaff et al., 2008;Meyerowitz and Chaiken, 1987) or for the influence of other content. Future studies should extend the analysis to encompass full messages, with different sources and frames. Second, this study focused on a single country, which limits the generalizability of the results. The importance of norms or values and the perceived relevance of the climate change issue vary across countries and cultures. Future research should include between-country comparisons of the influence of arguments on attitude towards CCS.