Attitudes to climate change risk: classification of and transitions in the UK population between 2012 and 2020.
Free full text in Europe PMC
Abstract
Free full text
Attitudes to climate change risk: classification of and transitions in the UK population between 2012 and 2020
Abstract
Strategies for achieving carbon emissions goals presuppose changes in individual behaviour, which can be indirectly nudged by interventions or tailored information but ultimately depend upon individual attitudes. Specifically, the perception that climate change is low risk has been identified as a barrier to participation in climate change adaptation and mitigation efforts. Therefore, understanding public attitudes towards climate change risk is an important element of reducing emissions. We applied k-means cluster analysis to explore attitudes to climate change risk in the UK population using data from the UK Household Longitudinal Study, a national survey running from 2009 to present. We identified three distinct attitude clusters: “Sceptical”, “Concerned”, and “Paradoxical” in both waves 4 (from 2012 to 2014) and 10 (from 2018 to 2020) of this survey. The Sceptical cluster tended to deny the seriousness of climate change and the urgency or even the necessity of dealing with it. The Concerned cluster displayed anxiety about climate change risks and supported action to reduce them. The Paradoxical cluster acknowledged the reality of climate change impacts but did not support actions to mitigate them. We further observed statistical associations between cluster membership and the social characteristics of the participants, including sex, age, income, education, and political affiliation. We also found a temporal stability of cluster structure between the two waves. However, the transition matrices indicated a general transition away from the Sceptical and Paradoxical clusters, and toward the Concerned cluster between wave 4 to wave 10. The findings suggest that more tailored public information campaigns regarding climate change risk may be necessary.
Introduction
In order to play its part in meeting the 1.5 degree Celsius global warming target—set in the Paris Climate Agreement—the UK has a long-term goal to achieve net-zero emissions by 2050 (Committe on Climate Change, 2019; United Nations, 2015). As well as requiring institutional changes in policy, individual behaviour changes will be needed to achieve this goal. This, in turn, may be dependent on the public’s beliefs about the seriousness of climate change and the urgency of action to mitigate it.
Analysis of UK national survey data has presented a complex picture of public perceptions of the risk of climate change. On the one hand, many British people believe climate change is serious. For instance, respondents’ support for the statement that “the seriousness of climate change is exaggerated” was 40% in 2010 to 30% in 2012 and 34% in 2013 (Poortinga et al., 2014). Similarly, a 2015 Pew Research Center (2015) poll found that 77% of people in the UK recognise global climate change as a very or somewhat serious problem. On the other hand, some people are in denial about the existence of climate change. Poortinga et al. (2011) found that 28% of people were uncertain about the existence of climate change. However, a more recent study conducted in 2018 indicated that complete scepticism about climate change is now unusual; with only 6% of the respondents being classed as deniers and many of those are not overly confident in their beliefs (Crawley et al., 2020).
The foregoing raises the questions of how attitudes toward and beliefs about climate change risk are organised and how they vary. As many authors have pointed out, the perception that climate change is low risk constitutes a barrier to participation in climate change adaptation and mitigation efforts (see, for example, Spence et al., 2011; Weber, 2011). However, merely describing public attitudes to climate change based on single-item measuring people’s agreement on seriousness of climate change might constrain our understanding of how people view its risk and how risk attitudes in different population segments might vary in nuanced ways. Previous work has provided snapshots of the heterogeneity of general assessments of climate change and/or wider environmental issues; they have focused on understanding people’s concern about climate change using a broad set of items relating to values, beliefs, willingness to support mitigation policy or behaviour. Consequently, they have not specifically focused on how people view the risks of climate or environmental change and understanding that specific element is critical in thinking constructively about behaviour change. In addition, most environmental and climate change segmentation studies in the UK have employed cross-sectional data and leaving open questions about change over time. As public attitudes to climate change have changed, it is, as Detenber and Rosenthal (2020) observe, valuable to monitor the composition and trajectory of population clusters over time. The current study explores the clustering of attitudes to climate change risk (ACCRs) that existed in the UK during the progress toward the GHG emissions reduction targets, and how the population segmentation of ACCRs both varies by socio-demographics and has changed over time. The findings will assist practitioners in the construction of communication frames; informing the tailoring and targeting of messages for members of each cluster, to more effectively maintain and/or enhance climate change engagement in each cluster.
The remainder of this paper is organised as follows. Section “Background” reviews relevant literature and develops research questions. Sections “The current study” and “Methods” outlines the methods employed, section “Results” presents the results and section “Discussion” discusses implications of the research, its limitations and possibilities for future research.
Background
Attitudes to climate change risk
Attitudes to Climate Change Risk (ACCRs) are generally considered to be consonant with the companion concept of climate change risk perception; and the two terms are often used interchangeably in the climate change research literature. Risk perception is a subjective construct often conceptualised as the perception of potential damage or harm (Slovic, 2000). To measure climate change risk perception, some researchers have utilised a variety of intermediate constructs including: perceived seriousness (McCright et al., 2014b), general concern about climate change threats (Milfont, 2012), likelihood measures of the various impact of global and/or local climate change will have in the future (Akerlof et al., 2013; Brody et al., 2008), the extent and timing of harms (Ding et al., 2011), while others adopt a combination of all or some of these measures (e.g., Leiserowitz, 2006; van der Linden, 2015).
In this study, we follow van der Linden (2017) in using ACCRs to refer to the combination of the perception of the probability that climate change will happen and the affective response to the perceived severity of negative climate-related consequences. This definition reflects a bipartite theoretical position. On the one hand, the cognitive process underlying ACCRs is based on knowledge about climate change, which includes the existence, cause, impacts and solutions of climate change (Tobler et al., 2012; van der Linden, 2015). On the other hand, the affective processing of climate change involves apprehension and worry about the potential negative impacts of climate change (Arbuckle et al., 2013).
In considering climate risk, people may distinguish between personal and societal implications (Bord et al., 1998), or they may perceive this as a risk for future generations, or as something out of their personal control (Poortinga and Pidgeon, 2003). The temporal perception of climate change is a critical component of the cognitive element of ACCRs and concerns when its effects are felt and who will feel them (e.g., Poortinga et al., 2011). Spatial perception is also important and may produce biases based on self-other comparisons (i.e., if the individual perceives that they are less likely to experience the effects of climate change than others are; Pahl et al., 2005) and geographic distance (i.e., local conditions are better than global conditions, Gifford et al., 2009). Hence, people do not always view climate change as personally threatening or relevant to them. van der Linden (2017) developed a “hierarchy of concern” model, which hypothesised climate change risk perception might be represented by a transitive relationship: “An individual may think that climate change (and associated impacts) are likely to occur, but that doesn’t mean that someone also perceives climate change to be a serious issue. In turn, an individual can perceive climate change to be a serious issue, but that doesn’t necessarily imply that they are concerned about it. Finally, although the public may express generalised concern about climate change, this often does not mean that people also personally worry about the issue or think it is a high priority.” (p. 24).
Population segmentation of climate change attitudes
Climate change is slow-moving, cumulative, and unsituated (Van Vugt et al., 2014). Consequently, people do not experience the change directly or even detect it on a sensory level. This is accentuated by the global scale and abstract statistical nature of our climate; covering as it does long-term changes of the whole of the world, with varying effects in particular localities (Weber, 2010). Public perceptions are therefore heterogeneous with risk attitudes being driven by extraneous social and psychological characteristics (Sciullo et al., 2019; van der Linden, 2017).
Several recent studies have sought to organise people’s climate change beliefs and attitudes based on homogenous item response patterns (see, for example, Maibach et al., 2011; Myers et al., 2012; Rhead et al., 2018; Sibley and Kurz, 2013). Using a survey of 2164 respondents in the US, with 36 ordinal and nominal variables related to beliefs in global warming, degree of involvement in the issue, behaviours, and policy preference, Maibach et al. (2011) used Latent Class Analysis (LCA) to classify six audience segments: Alarmed, Concerned, Cautious, Disengaged, Doubtful, and Dismissive, which they refer to as the Six Americas. Subsequently, Myers et al. (2012) tested Maibach et al.’s segments of US residents (N=
1127) and compared the effectiveness of climate change message framings among the segments. Conversely, Sibley and Kurz (2013) used data from New Zealand and produced a four-class model consisting of Climate Believers, Undecided/Neutral, Climate Sceptics, and Anthropogenic Climate Sceptics. Narrower classes also have been found in the UK nationally representative survey; Rhead et al. (2018) produced four classes of concern for the environment (including climate change): Pro-environment, Neutral Majority, Disengaged, and Paradoxical. Most recently, Crawley et al. (2020) used LCA to segment five climate change opinion groups (Highly Engaged, Moderately Engaged, Action-wary, Uncertain, and Deniers) in 2018 in the UK (N
=
787).
Other studies have used cluster analysis to segment the population into broader environmental attitude groups. For example, DEFRA (2008) used cluster analysis and contextualised interviews to produce a seven-category UK-wide environmental typology ranging from Positive Greens to Honestly Disengaged. Thornton et al. (2011) identified nine clusters of attitudes to transport choices based on the interviews conducted with English residents. Their segments combined attitudes and behaviour patterns with socioeconomic groups, with divisions ranging from Older, less mobile car owners to Young urbanites without cars. Waitt et al. (2012) clustered climate change responses into three segments: Strong Engagement, Modest Engagement, and Limited Engagement using Ward’s and k-means cluster analysis in a single Australian city (Wollongong). In Australia, Ashworth et al. (2011) identified a four-category segmentation ranging from Engaged to Doubtful using a two-stage cluster analysis (unspecified hierarchical and k-means). Similar segmentations also produced in a multi-national sample, a recent work conducted by Kácha et al. (2022) classified four groups of climate change beliefs and attitudes in 23 countries in Europe: Engaged, Pessimistic, Indifferent, and Doubtful.
Beyond the cross-sectional analysis of population classifications, some researchers have attempted to track changes of climate change values, beliefs, or behaviours clusters over time. For example, Leiserowitz et al. (2011) explored the changes in segment size of the so-called “Six Americas” from 2008 through 2011. However, Leiserowitz and colleagues assumed a fixed and constant cluster solution across time and specifically that the original Six Americas cluster structure identified in the 2008 data remain stable over time, which ignores the possibility that segment structure may change across time or that previous clusters might combine or split (Hine et al., 2014). Conversely, Heberlein (2012) provides examples where environmental attitude clusters do appear to disappear or merge into other clusters as people age and/or experience period effects (such as major environmental disasters or extreme weathers). Nevertheless, such changes in attitudes take time and do not necessarily lead to changes in overall attitudinal structures.
It follows from the above, that it is reasonable to track population shifts between segments in ways that are both sensitive to the emergence of new cluster structures and driven by the actual data structure, rather than constrained by the solutions from previous studies. To date, there are still gaps in our knowledge of the trajectory of attitudes toward climate change risks across time. Therefore, there is a need to draw a picture of attitudes transition adopting approaches that may produce more precise and valid change measures.
A review of past work on the climate change attitude population segmentation suggests that LCA and cluster analysis are two popular approaches. Given its ability to analyse both categorical and interval-scaled variables, include cases with missing data, and access model fit (McLachlan and Peel, 2000), LCA has been the preferred strategy for generating such segmentation (Hine et al., 2014). However, for segmenting large national survey datasets (especially for the sample sizes over ten thousand), its efficiency might be limited by the potentially large numbers of clusters. For instance, small and idiosyncratic response patterns can contribute to the model fit when LCA is coupled with very large sample size, resulting in a large number of classes that are hard to interpret substantively and a risk that mispecified models are absorbed into spurious latent classes (Bauer and Curran, 2004). LCA is also more likely to encounter convergence problems, while cluster analysis methods (such as K-mean clustering) are more likely to converge with larger samples and in general will be computationally robust as the sample size increases (see Vermunt, 2011).
So, following the above reasoning, this study will generate ACCRs group using k-means cluster analysis in a national survey dataset (with a sample size over ten thousand) and will use as the basis for an analysis of cross-time cluster transitions (at the individual and population levels).
Socio-demographic characteristics and climate change risk attitudes
An important area of study has been the exploration of how socio-demographic variables (e.g., education, age, gender, income, and political affiliation) predict climate change and environmental attitudes. Evidence on the influence of such socio-demographic factors is mixed with results varying from sample to sample and study to study.
(Binary) Sex
Some studies have found that males are more likely to be deniers of climate change than females (Clements, 2012; McCright, 2010). Milfont and Sibley (2016) theorised that women tend to be socialised to have higher levels of empathy than men, leading them to hold stronger pro-environmental attitudes and heightened perception of the risk of environmental impoverishment. However, other studies have found that females were less environmentally concerned than males. McEvoy (1972) attributed this to men having higher levels of education (at the time) and greater involvement in the community and political issues. Yet others found no significant effect of gender on environmental concern (e.g., Melis et al., 2014).
Age
Results are equally inconsistent for age, with some studies revealing a small negative correlation between age and climate change perception, (e.g., see Heath and Gifford, 2006; Hornsey et al., 2016), while others find no significant effect (e.g., Akerlof et al., 2013; O’Connor et al., 1999) or a positive correlation (Slimak and Dietz, 2006). Most of the studies suggested that age was inversely related to more general environmental attitudes, with younger cohorts tending to be more environmentally concerned than older cohorts (e.g., see Gökşen et al., 2002; Rhead et al., 2018).
Income
On average, people with higher income report stronger beliefs of climate change—yet lower risk perception—compared to those with less income (Xiao and McCright, 2012). One of the explanations for this is that income is a proxy measure of social class/social grade, partially because it shapes the social contexts where people are exposed to, and share information (Kraus and Stephens, 2012) and partly because people of higher social class and economic position are more likely to be risk-averse given that environmental challenges represent a risk to the hegemony of industrial capitalism and with it their social position (Ballew et al., 2020; McCright, 2011). Similarly, Rhead et al. (2018) found that Britain’s highest social grade (managerial/professional) had the highest proportion of pro-environmental attitudes and behaviours. For those who might have lower income, Rhead et al. (2018) found that non-workers, pensioners, or job-seekers are inclined to moderately regard climate change as a low priority and hold apathetic attitudes towards environmental issues despite acknowledging the risks.
Political affiliation
Partisanship is envisaged as the largest demographic correlate of attitudes to climate change (Hornsey et al., 2016; Davidson and Haan, 2012). People who vote for left-leaning parties are more likely to view climate change as risky than those who support mainstream-right political parties (Zahran et al., 2006; Davidson and Haan, 2012). Within the US, the tendency for Republicans (right-wing) to express more scepticism than Democrats (left-wing) has long been observed and is credited with contributing to a growing ideological gulf between sceptics and non-sceptics (McCright and Dunlap, 2011b; McCright et al., 2013, 2014a). In the majority of European countries, Gregersen et al. (2020) found that self-positioning further to the right of the political spectrum is correlated with lower levels of worry about climate change. Melis et al. (2014) found that voting for the UK Conservative Party (right-wing) is associated with lower levels of environmental concern than any other voting choice (including non-voting). Similarly, research by Phillips et al. (2018) found that support for climate change mitigation policies is slightly higher for Labour voters (Mainstream-left) than for Conservative (Mainstream-right) voters. However, recent research conducted post-COVID-19 in the UK has shown that even Conservative voters prefer the environment to be at the heart of economic reconstruction (Kenward and Brick, 2021).
Education
Again, the results for level of education are mixed. On one hand, having attended higher education predicts stronger perception of the risk of climate change (e.g., Hornsey et al., 2016; Lee et al., 2015; van der Linden, 2015) and less denial about climate change (Boag et al., 2016). On the other hand, years of education has also been found to have an inverse relationship with concern about climate change (e.g., Malka et al., 2009; Slimak and Dietz, 2006). Other studies, for instance, Melis et al. (2014) and Rhead et al. (2018) found that education level is not a significant predictor for environmental concern in the UK population once other factors are controlled for. This may partly be because the effects of education on concern about climate change tend to interact with those of political affiliation (Hamilton, 2011). People tend to acknowledge and interpret information through the filter off their existing value and belief systems (Taber and Lodge, 2006) and are more inclined to develop and reinforce attitude systems based on the cultural worldviews and social groups to which they belong (Kahan et al., 2007). For example, groups that have a general tendency to dismiss climate change (e.g., white-male right-wingers) may, when exposed to politicised issues like climate change through say higher education, interpret information in a way that bolsters value and attitudes systems associated with groups of which they are members (Drummond and Fischhoff, 2017). Hence, the probability of perceiving climate change as risky increases with education among those with left-wing political affiliation but decreases with education among right-wingers.
In a nutshell, despite the inconsistent results, there is some evidence for a socio-demographic risk profile where typically younger, female, higher educated, those with higher income and a left-leaning political affiliation express more concern about climate change. Hence, here we will examine how population segments with different ACCRs vary by age, gender, income, education, and political affiliation.
The current study
Although the majority of previous research has focused on the latent variable analysis of general concern about the environment or climate change, such as four environmental concern clusters partitioned in the British population by Rhead et al. (2018) and five climate change opinion groups in the UK by Crawley et al. (2020), to date little research has been focused on cluster analysis of ACCRs and cluster trajectory across time. The questions of how people perceive and are concerned about the timescale, trend, consequences, and crisis of climate change has been largely unstudied. Shifting from the focus on general concern to the focus on how people view risks of climate change specifically would enable communications and interventions to be tailored towards specific groups (with varying response patterns to climate change risk statements). The present study seeks to explore ACCRs using k-means cluster analysis in order to identify statistically distinct response clusters across time.
How environmental concern cluster membership varies by socio-demographics has been tested in previous studies, but the findings have been mixed. Additionally, the demographics of cluster membership may only be consistent with those found previously if our ACCRs clusters are themselves isomorphic with the environmental concern clusters extracted in previous studies. Therefore, we will examine whether the probability of belonging to any of the ACCRs clusters that we identify is influenced by individuals’ sex, age, income, education, and political affiliation without any pre-determined hypothesis.
The study uses the climate change module with the UK Household Longitudinal Study (UKHLS) from waves 4 and 10 (which were collected between 2012 and 2014, and 2018 and 2020, respectively).
The goal of the present study was therefore to answer three key research questions:
What groups of ACCRs exist in the UK?
What is the relationship between socio-demographic variables and ACCRs group membership?
How do ACCRs group structure and membership change over time?
Methods
Data and sample
Our analysis uses the UK Household Longitudinal Study (UKHLS)1; an annual panel survey of UK households. The first wave was conducted in 2009 included approximately 40,000 households and 100,000 individuals. These households were initially selected using a multistage random sampling (a random sample of addresses drawn from a random sample of UK post codes) and used multiple sample components to facilitate research into different smaller sub-groups: a General Population Sample, an Ethnic Minority Boost Sample, an Immigrant and Ethnic Minority Boost Sample, and the British Household Panel Survey Sample2. The main adult questionnaire3 for waves 4 and 10 was an interviewer-administered CAI questionnaire (either face-to-face using the CASI module or by telephone). The questionnaire was administered to all eligible household members aged sixteen or above who responded to a wide range of questions covering: family and households, income, education, politics, education, health and wellbeing, and attitudes.
The UKHLS in waves 1, 4, and 10 contained questions regarding attitudes and beliefs toward the seriousness, urgency, and preventability of climate change. We used data from waves 4 and 10 because they were more recent and provided consistent and more fine-grained measurements (5-point Likert scales compared to dichotomous responses in wave 1). Wave 4 (covering 2012 to 2014) and wave 10 (covering 2018 to 2020) contained responses from 38037 individuals (valid rate=
80.81%) and 31498 individuals (valid rate
=
91.78%) respectively, after listwise deletion of cases with missing values in the attitudes to risks of climate change variables. Missing data analysis—including univariate missingness and multivariate missing data pattern analysis—showed that missingness of climate change risk attitudes variables in both waves 4 and 10 was plausibly Missing At Random (MAR; for more details of missing data diagnosis, see Supplementary Appendix A), which allowed the estimation of models to use maximum likelihood.
Measures of attitudes to climate change risk
We selected five statements4 regarding opinions on climate change, similar to those used by Rhead et al. (2018). Participants indicated their agreement with each statement using either a 5-point Likert scale (1=
Strongly disagree to 5
=
Strongly agree) or a dichotomous response (0
=
No and 1
=
Yes). These items captured attitudes toward the risk of climate change and its impacts on the UK in the future (see Table Table11).
Table 1
Measures of ACCRs.
| Variable name | Statement |
|---|---|
| Beyond control | Climate change is beyond control, it’s too late to do anything about it. |
| Too far in future | The effects of climate change are too far in the future to really worry me. |
| Affected within 30 years | People in the UK will be affected by climate change in the next 30 years. |
| Major disaster | If things continue on their current course, we will soon experience a major environmental disaster. |
| Crisis exaggerated | The so-called ‘environmental crisis’ facing humanity has been greatly exaggerated. |
Statistical analyses
We carried out k-means cluster analysis to partition people’s attitudes into homogeneous clusters. We used the three criteria proposed by Makles (2012) to decide on the optimal number of clusters: (i) The number of clusters associated with the first non-linear inflection point when plotting the within-cluster sums-of-squares (WSS) in order of size, and (ii) The number of clusters with the largest ratio of WSS to total sums-of-squares (TSS), referred to as the η2 coefficient and (iii) the proportional reduction of error (PRE) coefficient (Schwarz, 2008), which measures the proportional reduction of WSS for a solution with k-clusters compared with one with K-1 clusters. Further cluster comparison and validation was conducted by a one-way-analysis-of-variance (ANOVA) of the differences of items’ means across clusters.
We assigned each participant to membership of the cluster with the nearest cluster centroid in the optimal solution, and then used multinomial logistic regression analysis to predict cluster membership based upon characteristics of the respondent. The characteristics we used were age, sex, education, income, and political affiliation. Age and income were mean-centred.
Results
Descriptive statistics
Descriptive statistics of ACCRs and social characteristics (age, income, sex, education, and political affiliation) are shown in Table Table2.2. Sex was treated as dummy variable (1=
male, 0
=
female). Age was coded as a continuous variable, ranging from 16 to 100 in wave 4 and 16 to 103 in wave 10. Income was captured by people’s gross monthly personal income in thousands of UK pounds, ranging from −8.85 to 27.47 in wave 4 and −3.33 to 26.63 in wave 105. Education was an ordinal variable with five categories, ranging from 1
=
no qualification to 5
=
degree6. Political orientation7 was coded according to the party the respondent feels closest to. In wave 4 data, we coded political affiliation as, 0
=
Right-wing (Conservative, Ulster unionist, Democratic unionist), which is used as the reference category, 1
=
Left-wing (Labour, Liberal democrat, Scottish National Party, Plaid Cymru, Green Party, Social Democratic and Labour Party, Alliance Party, Sinn Fein), 2
=
Other Parties (None of the above). In wave 10 data, several parties have been removed in the questionnaire (i.e., Democratic unionist and Social Democratic and Labour Party) while some have been added (i.e., UK Independence Party, The Brexit Party, and Change the UK). Accordingly, we coded political affiliation as, 0
=
Right-wing (Conservatives, Ulster unionist, UK Independence Party, The Brexit Party), which is used as the reference category, 1
=
Left-wing (Labour, Liberal democrat, Scottish National Party, Plaid Cymru, Green Party, Alliance Party, Sinn Fein), 2
=
Other Parties (None of the above).
Table 2
Descriptive statistics of ACCRs and demographic variables in waves 4 and 10.
| ACCRs | Mean | SD | N | Min | Max | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | |
| Major disaster | 3.31 | 3.65 | 1.00 | 0.96 | 38,037 | 31,498 | 1 | 1 | 5 | 5 |
| Crisis exaggerated | 3.02 | 2.58 | 1.03 | 1.05 | 38,037 | 31,498 | 1 | 1 | 5 | 5 |
| Beyond control | 2.65 | 2.48 | 0.98 | 0.95 | 38,037 | 31,498 | 1 | 1 | 5 | 5 |
| Too far in future | 2.64 | 2.36 | 1.05 | 1.04 | 38,037 | 31,498 | 1 | 1 | 5 | 5 |
| Affected within 30 years | 0.78 | 0.84 | 0.41 | 0.36 | 38,037 | 31,498 | 0 | 0 | 1 | 1 |
| Demographic variables | Mean | SD | N | Min | Max | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | |
| Age | 47.13 | 49.65 | 18.04 | 18.57 | 37,878 | 30,981 | 16 | 16 | 100 | 103 |
| Income (£1000) | 1.67 | 1.91 | 1.53 | 1.70 | 37,878 | 30,981 | −8.85 | −3.33 | 27.47 | 26.63 |
| Demographic variables | Categories | N | Percent | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| W4 | W10 | W4 | W10 | |||||||
| Sex | Female | 21,076 | 17,206 | 55.64 | 55.54 | |||||
| Male | 16,802 | 13,775 | 44.36 | 44.46 | ||||||
| Education | No qualification | 4405 | 2720 | 11.63 | 8.77 | |||||
| GCSE etc | 11,608 | 8711 | 30.65 | 28.12 | ||||||
| A-level etc | 8252 | 6625 | 21.79 | 21.38 | ||||||
| Other higher education | 4516 | 3889 | 11.92 | 12.55 | ||||||
| Degree | 9097 | 9036 | 24.02 | 29.17 | ||||||
| Political affiliation | Right-wing | 6729 | 1493 | 36.21 | 38.43 | |||||
| Left-wing | 11,145 | 2374 | 59.98 | 61.11 | ||||||
| Other party | 707 | 18 | 3.80 | 0.46 | ||||||
k-means cluster analysis
k-means results showed that clustering with k=
3 is the optimal solution in both waves 4 and 10. At k
=
3, a kink or cut-off point occurred in both the WSS and log (WSS) figures (see Fig. Fig.1),1), where the difference in the within-cluster dissimilarity is not substantial. Compared to the k
=
2 solution,
>
3. Sensitivity analyses showed that k
=
3 was robust. Repeating the analysis 50 times with randomly selected cluster centroid starting points resulted in k
=
3 as the optimal solution 49 times, with k
=
4 once in wave 4 data; k
=
3 as the optimal solution 47 times, with k
=
2 twice and k
=
4 once in wave 10 data (see Fig. Fig.2).2). Results of the ANOVAs with two waves data both showed that there were significant differences between the three clusters to all attitude questions (Major Disasterwave4 (F(2, 38034)
=
5056.80, p
<
0.001), Crisis Exaggeratedwave4 (F(2, 38034)
=
8254.75, p
<
0.001), Beyond Controlwave4 (F(2, 38034)
=
7379.56, p
<
0.001), Too far in Futurewave4 (F(2, 38034)
=
14656.25, p
<
0.001), Affected within 30 yearswave4 (F(2, 38034)
=
553022.80, p
<
0.001; Major Disasterwave10 (F(2, 31497)
=
8782.13, p
<
0.001), Crisis Exaggeratedwave10 (F(2, 31497)
=
13973.79, p
<
0.001), Beyond Controlwavw10 (F(2, 31497)
=
5151.37, p
<
0.001), Too far in Futurewave10 (F(2, 31497)
=
12776.86, p
<
0.001), Affected within 30 yearswave10 (F(2, 31497)
=
970.00, p
<
0.001).
a–d displayed wave 4 data. a WSS (within sum of squares), b log (WSS), c η2, and d PRE (proportional reduction of error) for 20 cluster solutions. e–h displayed wave 10 data. e WSS (within sum of squares), f log (WSS), g η2, and h PRE (proportional reduction of error) for 20 cluster solutions. k stands for the number of cluster.
a–d displayed wave 4 data. a
WSS (within sum of squares), b log (WSS), c
η2, and d PRE (proportional reduction of error) for 20 cluster solutions (k=
20). e–h displayed wave 10 data. e WSS (within sum of squares), f log (WSS), g
η2, and h PRE (proportional reduction of error) for 20 cluster solutions (k
=
20). Lines in 50 different colours indicate different cluster solutions by repeating the clustering 50 times with different starting points.
A key finding is that the three-cluster pattern is almost identical in the two waves (see Fig. Fig.3).3). Members of cluster 1 disagreed most strongly that people will be affected by climate change in the next thirty years compared to the other two clusters (see Fig. Fig.3).3). They had the strongest propensity to embrace a negative statement of climate change risk. For example, when considered against people’s outlook towards the trend of climate change, members would be less likely to worry about the effect of climate change in the future and more likely to negate the action to cope with climate change. In this sense, members of cluster 1 are most optimistic about the risk of climate change and therefore make light of potential major environmental disasters on various grounds, such as that climate change is exaggerated and distant. Therefore, it is likely that cluster 1 is the group that has sceptical thinking. This corresponds to the sceptical cluster identified by other studies (see e.g., Dunlap, 2013; Maibach et al., 2011; Sibley and Kurz, 2013). In effect, the defining feature of sceptical thinking is to downplay the risk. Hence, we named cluster 1 “Sceptical”.
W4, wave 4. W10, wave 10.
Members of cluster 2 have an overall lowest probability of agreeing with the three negative statements and the highest probability of agreeing with the two positive statements. People in this cluster tend to agree that problems are coming for both the environment (Major disaster) and climate (Affected within thirty years) and tend to agree that the environmental crisis truly exists and has not been exaggerated. Moreover, members of cluster 2 indicate concern about climate change risk. For instance, this cluster concurred that climate change is not too far in the future to worry about nor beyond control, implying their potential approval of urgent efforts to tackle with climate change. Given this, cluster 2 is the group that are concerned about climate change risk and the need to deal with it urgently. We named cluster 2 “Concerned”.
Members of cluster 3 seem undecided as to what position they stand for. This cluster finds the effects of climate change are too far in the future to worry them but also likely to come within thirty years. From the point of attitudes pattern, they neither agreed with the positive nor disagreed with negative statements, ranging from −0.10 to 0.53 in wave 4 and −0.34 to 0.56 in wave 10. Item probabilities for cluster 3 suggest that members, on the one hand, possess a similar attitude pattern with cluster “Sceptical” regarding the negative cognitive evaluation of climate change, such as crisis facing humanity has been exaggerated. On the other hand, cluster 3 is far less likely to disagree that the problems of climate change are coming compared to cluster “Sceptical”. For instance, they displayed the highest probability to agree that people will be affected by climate change in the near future. However, this cluster showed the strongest inclination to agree that climate change is out of control and it is too late to do anything about it, implying they do feel powerless and passive as the risk is going to happen to an extent but they do not believe there is much can be done. Note that members of cluster 3 also modestly disagree that major environmental disasters will occur though it is weak in strength, indicating they feel climate change is probably overblown. To sum up, this cluster appears to be expressing ambivalent risk attitudes. Given this, we labelled it “Paradoxical”. This group is consistent with Rhead et al. (2018) environmental concern cluster with the same name and similar pattern.
Multinomial logistic regression analysis
We fitted two multinomial logistic regression models in each wave, one treating education as continuous and one as dummy-coded categories. We chose the latter model as it had a lower Akaike’s Information Criterion (AIC; 77395.93 with 20 parameters for categorical and 77,437.46 with 14 parameters in wave 4, 60,062.28 with 20 parameters for categorical and 60120.85 with 14 parameters for continuous in wave 10). Table Table33 shows the parameter estimates for the multinomial logit model8.
Table 3
Odds ratios for variations in sex, age, education, income, and political affiliation by ACCRs clusters in waves 4 and 10.
| Sceptical | Paradoxical | Concerned | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Odds-ratio | p | Odds-ratio | p | Odds-ratio | p | |||||||
| W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | W4 | W10 | |
| Male | 1.428 | 1.524 | 0.000 | 0.000 | 1.110 | 1.168 | 0.000 | 0.000 | – | – | – | – |
| Age | 1.001 | 0.995 | 0.437 | 0.000 | 1.005 | 1.000 | 0.000 | 0.888 | – | – | – | – |
| Education | ||||||||||||
| No qualification | – | – | – | – | – | – | – | – | ||||
| GCSE etc | 0.553 | 0.526 | 0.000 | 0.000 | 0.598 | 0.595 | 0.000 | 0.000 | – | – | – | – |
| A-level etc | 0.382 | 0.351 | 0.000 | 0.000 | 0.453 | 0.436 | 0.000 | 0.000 | – | – | – | – |
| Other higher education | 0.356 | 0.270 | 0.000 | 0.000 | 0.409 | 0.410 | 0.000 | 0.000 | – | – | – | – |
| Degree | 0.193 | 0.143 | 0.000 | 0.000 | 0.251 | 0.205 | 0.000 | 0.000 | – | – | – | – |
| Income | 0.935 | 0.920 | 0.000 | 0.000 | 0.967 | 0.953 | 0.000 | 0.000 | – | – | – | – |
| Political affiliation | ||||||||||||
| Right-wing | – | – | – | – | – | – | – | – | – | – | – | – |
| Left-wing | 0.428 | 0.269 | 0.000 | 0.000 | 0.633 | 0.413 | 0.000 | 0.000 | – | – | – | – |
| Other party | 1.048 | 0.377 | 0.657 | 0.225 | 0.943 | 0.562 | 0.546 | 0.270 | – | – | – | – |
| Constant | 1.811 | 1.022 | 0.000 | 0.830 | 2.907 | 2.482 | 0.000 | 0.000 | – | – | – | – |
Note: W4, wave 4. W10, wave 10.
The two waves of data manifested consistent predicted probabilities of ACCRs cluster membership by socio-demographic characteristics except for age. In wave 4 data, age differences were not a good predictor of membership of the Sceptical versus the Concerned cluster, but higher age was associated with belonging to the Paradoxical versus the Concerned cluster: about 5% more likely per decade of age difference. Conversely, wave 10 data suggested that lower age was associated with belonging to the Sceptical versus the Concerned cluster—about 5% more likely per decade of age difference—but age was no longer a significant predictor of membership of the Paradoxical versus the Concerned cluster.
Males were around 50% percent more likely than females to belong to the Sceptical cluster and over 10% more likely to belong to the Paradoxical cluster (11% in wave 4 and 16.8% in wave 10) compared to the Concerned cluster. Higher income was associated with belonging to the Concerned cluster versus either of the others, by 6.5% (wave 4) and 8% (wave 10) more likely per £1000 of income compared to the Sceptical cluster and 3.3% (wave 4) and 4.7% (wave 10) more likely compared to the Paradoxical cluster.
The strongest predictors of cluster membership were education and political affiliation. The higher the level of qualifications, the more likely participants were to belong to the Concerned cluster and less likely they were to belong to the other clusters in two waves. The biggest difference between adjacent education levels was between those with no qualifications versus those with age-16, school-level qualifications (GCSE etc.). The latter were at least 40% (in wave 4; 40.5% in wave 10) more likely than those with no-qualifications to belong to the Concerned cluster compared to either of the others. Across the full range of qualifications, those with a degree were at least 70% (in wave 4, 79.5% in wave 10) more likely than those with no qualifications to belong to the Concerned cluster compared to the other clusters. Regarding political affiliation, those aligned with left-wing parties were at least 36.7% in wave 4 and 58.7% in wave 10 more likely than right-wingers to belong to the Concerned cluster compared to the other clusters. Supporters of other parties were largely aligned with right-wingers in wave 4 and left-wingers in wave 10 with regard to cluster membership.
Figure Figure44 shows the marginal probabilities of cluster membership for the different levels of education and political affiliation. There was a clear distinction between supporters of left-wing parties vs. right-wing parties with regard to membership of the Sceptical (panel a) and the Concerned (panel b) cluster in wave 4 data, as well as membership of three clusters (panels d, e, f) in wave 10 data. In each case, cluster membership was strongly associated with education, negatively for the Sceptical and Paradoxical cluster. In each wave, education was substantially associated positively with membership of the Concerned cluster, the differences associated with political affiliation were much bigger for this cluster than the others9.
a–c displayed wave 4 data. a Sceptical, b Concerned, and c Paradoxical. d–f displayed wave 10 data. d Sceptical, e Concerned, and f Paradoxical. 1=
No qualification, 2
=
GSCE etc., 3
=
A-Level, 4
=
Other higher education, 5
=
Degree.
Transitions in ACCR clusters between wave 4 and wave 10
Although cluster structure was very similar in the two waves, the overall proportions of each cluster have changed. The Paradoxical (43.2%, Concerned=
35.4%, Sceptical
=
21.3%) was the biggest cluster in wave 4, while the Concerned (43.8%; Paradoxical
=
40.7%, Sceptical
=
15.6%) has become the biggest in wave 10. Table Table44 gives the transition matrix of group members who shift attitudes to climate change risk from wave 4 to wave 10. Results indicated that 71.6% of the population in group Concerned in wave 4 continue to be members of the Concerned cluster risks in wave 10. By contrast, over half of the population of the Paradoxical in wave 4 shifted attitudes to the Concerned in wave 10, around one-fifth people in group Sceptical moved to the Concerned in wave 10. It is also noteworthy that nearly half of the Sceptical cluster shifted to the Paradoxical cluster (43.2%) in wave 10, and over one-fifth of the Concerned cluster also moved to the Paradoxical cluster (23.8%).
Table 4
Transition matrices by three clusters across waves 4 and 10.
| Wave 10 (2018–2020) | |||||||
|---|---|---|---|---|---|---|---|
| Cluster | Sceptical | Concerned | Paradoxical | Missing | Total | ||
| Wave 4 (2012–2014) | Sceptical | 1595 (37.1%) | 843 (19.6%) | 1858 (43.2%) | 300 | 4596 | |
| Concerned | 394 (4.7%) | 6036 (71.6%) | 2003 (23.8%) | 260 | 8693 | ||
| Paradoxical | 1190 (13.2%) | 3008 (33.3%) | 4848 (53.6%) | 590 | 9636 | ||
| Missing | 518 | 1002 | 1230 | 628 | 3378 | ||
| Total | 3697 | 10,889 | 9939 | 1778 | 26,303 | ||
To summarise, we observe a considerable movement between the clusters with a general trend towards higher membership of the Concerned cluster, but the overall cluster structure appears to be robust over time.
Discussion
General discussion
This study segmented respondents based on their appraisal of statements of climate change risk. Three discrete clusters were suggested by the application of k-means cluster analysis using the within-sum-of-squares statistic: Sceptical, Concerned, and Paradoxical. The proportions, item probability patterns of the three segments, and transition by clusters over time express Britain’s dynamic pattern of ACCRs. Our multinomial logistic regression analysis of the socio-demographic profile of cluster members has found a variation by sex, age, education, income, and political affiliation between the three clusters.
Members of the Sceptical cluster deny that climate change is going to be problematic. This is consistent with the “Sceptical”, “Doubtful” and “Denial” groups identified in previous research (e.g., Dunlap, 2013; Kácha et al., 2022; Maibach et al., 2011; Sibley and Kurz, 2013; Whitmarsh, 2011), this group minimise the projected effects of climate change and therefore is less likely to support actions to mitigate those effects. Different from outright scepticism (thinking that denies the existence of climate change), their doubts are focused on the seriousness, necessity, effectivity, and urgency of tackling climate change. This aligns with previous observations that absolute denial about the existence of climate change is relatively rare in the UK (e.g., Poortinga et al., 2011; Taylor et al., 2014; Whitmarsh, 2011). On the one hand, the Sceptical shares some similarity with apathetic attitudes in the Disengaged class found in Rhead et al. (2018) study; both do hold optimistic views about—and dismissive attitudes toward—climate threats. For the Sceptical cluster, climate change seems distant, happening mostly in the future and to other people. This is potentially indicative of optimistic temporal and spatial biases in their perception of climate change threats and therefore appear to dampen enthusiasm for tackling climate change (see Gifford et al., 2009 for a discussion of optimistic temporal and spatial biases). Given that pessimistic affective messages in climate change appeals could increase risk perception and facilitate engagement with climate issues more than optimistic affective messages (Morris et al., 2020), future studies examining the effect of possible optimism bias on sceptical thinking about climate change risk may prove valuable.
The Concerned cluster is the most convinced that climate change is happening and has not been exaggerated, which is strongly indicative of potential approval of urgent climate change mitigation efforts and being ready to engage with climate change (e.g., Maibach et al., 2011; Roser-Renouf et al., 2014). A large proportion of the Concerned cluster (35.44% of all respondents) do not agree that climate change is too far in the future and not worth being concerned about, making them similar to the “Alarmed” and “Concerned” groups identified in previous studies (see also Maibach et al., 2011; Metag et al., 2017; Myers et al., 2012; Poudyal et al., 2020). A similar climate attitude group has also been found in the US (Goldberg et al., 2020). We note that Leiserowitz et al. (2009) found that people who are concerned about the issue use media channels to capture the information about climate change. This view was supported by Wonneberger et al. (2019), who found that those concerned about climate change paid more attention to media coverage of the COP21 meeting than did those who were doubtful about climate change. However, Rosenthal (2022) found that individuals’ worldviews reflected their personal experience of climate change, which can be influenced by information sources such as mainstream and social media. These associations could be explored as a way to persuade the Sceptical segment.
The Paradoxical cluster displayed a mixed ACCR profile; they had the highest probability of agreeing that people will be affected by climate change in the next thirty years and also agreed with the statement that it was too far in the future to worry about. Consistent with Rhead et al. (2018) study, the Paradoxical class recognised environmental problems but then seem to dismiss or trivialise them. The Paradoxical cluster appear confused and perhaps less capable of dealing with threats that are happening in the future with drawn out and uncertain impacts with complex causality (Marshall, 2005). The Paradoxical group in our study was the largest (43.22% of the respondents) in wave 4 and just second largest by a narrow margin in wave 10, showing that many in Britain still feel ambivalent and/or confused about climate change risk. This corresponds to Kácha et al.’s (2022) findings (in their study of 23 Europe countries)—where is the Indifferent group—which is the segment similar to the Paradoxical cluster who hold “neither-nor” attitudes to negative and positive statements of climate change. Rhead et al. (2018) explored the association between attitudes and pro-environmental behaviours and found that the odds of “Paradoxical” class members engaging in pro-environmental behaviours were lower than the “Pro-environment” and the “Neutral Majority” classes.
As well as differing in ACCRs, the clusters also differ in socio-demographics. Compared to females, males had a significantly higher probability of being the Sceptical and Paradoxical clusters, and a lower probability of being the Concerned cluster. There are many potential explanations for this tendency, for example, females are found to have more nurturing and maternal natures and more altruistic values; they also tend to have a longer time horizon (see e.g., Coelho et al., 2017), thinking about how the environment will be for the next generation rather than merely how it is at present (e.g., Milfont et al., 2015; Smith et al., 2017). Another explanation focuses on the origins of racial and gender variance in risk perception, as being rooted in cultural identity-protective cognition. To generalise, the theory claims that women are more worried about risks, including the risks associated with environmental pollution and the climate crisis, than men because the statements of risk challenges the cultural identities of hierarchical and individualistic males—especially white males (Kahan et al., 2007). This is consonant with the findings that conservative white males contribute significantly to climate change denial (McCright and Dunlap, 2011a).
Regarding age, younger respondents were more likely to belong to the Concerned versus the Paradoxical cluster in wave 4. Wave 10 data suggested that younger respondents were more likely to be in the Concerned versus the Sceptical cluster. This implies that younger people hold more concerned attitudes to climate change risks than the older ones. In line with Marks et al. (2021), our speculation is that young people may be more concerned because of worries about what harms they will be left to cope with while currently having little power to limit those harms. Older participants were more likely to belong to the Paradoxical cluster in wave 4 and the Sceptical cluster in wave 10. For the older participants, perhaps future climate change is a smaller worry for them relative to—for example—current personal health issues. Another possible element of this complex picture is that worrying about climate change and being willing to tackle it is a relative threat to older people who stand to lose the most because their social status is most entrenched in traditional worldviews (Goto Gray et al., 2019; Liere and Dunlap 1980). We also found that the higher one’s income, the lower probability of being the Sceptical and Paradoxical cluster over the Concerned cluster. One well-known explanation rests on the assumption that concern about environmental quality has some of the properties of a luxury good, which can be indulged only after more basic material needs such as adequate food, shelter, and economic security are met (see e.g., Liere and Dunlap, 1980; Maslow, 1970).
Traditional societal fault lines of sex, age and income seem to have some relationships with ACCRs, but these orthodox predictors were overshadowed in predictive power by education and political affiliation; the odds ratios of education and political affiliation were consistently the highest across the two waves data compared to other social characteristics. Regarding education, we saw that higher educational attainment increases the likelihood of belonging to the Concerned cluster versus the other two clusters compared to no qualification. We might theorise that the more educated a person is, the more able that person is to process complex information and rationally assimilate scientific evidence about climate change (Hoffmann and Muttarak, 2017). Another speculation is that people who have higher education may be more directly exposed to learning about climate change, which develops their climate change awareness. It seems reasonable to assume that exposure to climate science typically increases with education (Bohr, 2014), and higher education also brings raised exposure to political ingroup norms (Ehret et al., 2017); Although we note that Kahan (2017) has observed that science comprehension may be not necessarily correlated with beliefs about climate change. Hence, right-wingers are more inclined to be motivated to defend ideological positions associated with their conservative worldviews when confronted with information about climate change and climate-related crises. Similarly, individuals with left-wing political affiliations are more likely to be concerned; this confirms the finding of Milfont et al. (2015). Other researchers have also found that an interaction of political affiliation with education to be a good predictor of citizen concern about climate (e.g., Hamilton, 2011; McCright, 2011). However, in this study, we found no such interaction and that the higher level of education that a person has, the less likely they are to belong to the Sceptical and Paradoxical clusters regardless of their political orientation. Similarly, the effect of political orientation is independent of education.
While these co-variates do seem to be critical in distinguishing people’s ACCRs, sex, age, income, education, and political affiliation are not meant to provide an explanation nor are the list of included predictors meant to be exhaustive. Other potential socio-demographic predictors should also be considered in the future, including social class or grade. At the same time, the potential structural effect of age on education, income, and other variables directly related to the age variation is worth noting10. For example, the youngest respondents in our sample were aged 16, but the minimum age for the top category of education is 21. Therefore, there might be a small structural age effect of education. Attitudes are not fixed and might change throughout a person’s lifetime, perhaps driven by their country’s climate policy or mitigation targets, and indeed we observed a lot of movement in the respondents’ attitudes across two waves. Given that this is a fixed set of respondents who are aging together we cannot completely distinguish between age and period effects with these data. However, given that within the data older people tend to be more sceptical, we speculate that the nett shift evidenced here is probably a period effect—one affecting the whole population. However, this overall trend belies a great deal of turbulence with individuals shifting between all three clusters. Therefore, it would be meaningful to examine the characteristics of people who changed in one direction rather than the other in future work. Beyond exploring socio-demographic predictors, the question of how values, worldviews, ideology, and identity (shaped by income, social class, education, and political affiliation) further influence climate change attitudes warrants an in-depth investigation. Previous literature has indicated that climate scepticism is rooted in people’s core values and worldviews, mirroring the findings that identity-protective, cognition-driven political affiliation and education are the strongest correlates of denial and/or dismissive attitudes to climate change (Poortinga et al., 2011). Additionally, future research could usefully examine other explanatory concepts including the effects of historical events and individual experiences of extreme weather, which might also predict people who moves between clusters.
Implications
In this paper, we make a conceptual contribution through introducing a typology of ACCRs using k-means cluster analysis on a national probability sample and explore the cross-time transitions of cluster membership, thereby diversifying the evidence base beyond cross-sectional factor and latent class analysis. We extend existing literature on population classification by tracing the UK public ACCRs cluster trajectories between 2012 and 2020. Leiserowitz et al. (2011) assumed constant segmentation structure, based upon six Americas segmentation measured at time one, and explored how segmentation size changed over time without evaluating if the segmentation structure was applicable to the actual data at each time point. In contrast, we have gone a step further by evaluating the applicability to real-world data by data-driven k-means clustering, where the cluster solution is constructed independently at each wave. Cluster patterns in our study were consistent in 2012 and 2020, even though at the individual level there was significant movement between clusters. This combination—stable patterns but mobile individuals—suggests a lot more fluidity of opinion than perhaps media debates might suggest and this opens up the possibility of targeting policy interventions and creating information campaigns that are tailored towards the needs of specific groups with different risk attitude patterns.
Our findings suggest that targeting the Paradoxical populations should be prioritised when developing communication strategies. The Paradoxical, the largest cluster during the period of wave 4 (2012 to 2014) still included over 40% of population at wave 10 (2018 to 2020). However, membership of the Paradoxical cluster is least stable and the significant movement from the Paradoxical to the Concerned cluster between waves 4 and 10 suggest that messages targeted at those with paradoxical beliefs may be most effective at causing significant shifts in overall ACCRs. It is noteworthy then that some degree of ambivalence, mixed feelings, or non-straightforward sceptical thinking appear to be widespread in the UK, which might be indicative of people needing more or perhaps differently framed information to form clear opinions. Since the Paradoxical tend to be worried about climate change but feel powerless to cope with, it could be that the climate change affective messages for the Paradoxical should be motivation-oriented with pessimistic ending because it could trigger higher engagement with the issue than optimistic ending (see Morris et al. (2020) for a recent study of the issue of the relationship between affect, efficacy and climate change risk perception).
The Sceptical often questioned the uncertainty of climate change and may need to see risk quantifications to move them to middle ground. Since climate information or consultants directly or indirectly influence how people judge and assess risk, climate scientists may consider quantifying the risk and visualising the mitigation efforts in order to communicate with both ends of that spectrum the very real damage that climate change may inflict (Wendel, 2016) and translate scientific and local knowledge in climate change risk reduction to realities on the ground (Trogrlić et al., 2021). Scientific messages for the Sceptical cluster may also consider providing news of climate disasters and informative mitigation action policy as previous research has found that dire risks and climate apocalypse can enhance public engagement with climate change issues (see e.g., Mayer and Smith, 2019). Yet, it would be tiresome to hear every effect of climate change put on the worst-case scenario of death and destruction; messages with strong emotional content may erode intention to act among those who were not already highly concerned about climate change (Kerr, 2007; McNeish, 2017). Tailored risk communication messages may focus not only on how to best represent key information and messages to evince behaviour change in general but also on how to tailor such information to the segments and then disseminate it through appropriate channels in order to reach each segment (Andor and Fels, 2018; Rosenthal and Linder, 2021). There is now a significant body of research in communication studies that suggests that information channels, platforms, news outlets and media types can vary widely by demographic profiles and population segments (see, for example, Althaus et al., 2009; Diehl et al., 2019; Dutta-Bergman, 2004; Shehata and Strömbäck, 2021; Vara-Miguel, 2020). Hence, the relationship between population classes of ACCRs attitudes, patterns of media use, trust in different sources, and information seeking behaviours warrant exploration in future research. In sum, tailoring climate change information and communication for different clusters within the UK public may help narrow the divide on attitudes to climate change risks and foster greater consensus towards pro-environmental views.
Limitations
Despite these important insights, our study has several limitations.
The use of secondary data does create ontological constraints. The data were designed to be used in the context of a broad national survey in which climate change questionnaire is only one module of the whole interview. Consequently, the survey is not a perfect fit to our requirements. Some nuance is lost about details, which are not included in the survey for example regarding a respondent’s willingness to cope with climate change risk.
A second issue is that a relatively small number of directly matching survey items also drove the modest compromise of including two items that did not explicitly mention climate change: the statements about the “environmental disaster” and “environmental crisis”. It is possible that respondents may view environmental hazards as referring to other factors (e.g., plastic pollution, or roadside particulates) rather than specifically climate change issues. However, we would argue (i) that the appendage of the terms “disaster” and “crisis” will inevitably bring to mind the most talked about environmental challenge and (ii) in the survey itself, climate change is the individual issue that warranted the most individual questions. It is reasonable to posit that the impacts of climate change are the most salient aspects of “environmental disaster” and the “environmental crisis” for the respondents when they are responding to these questions. However, we acknowledge that this is an assumption, which we have no way of explicitly testing. These types of compromises are of course not unique to this study; all secondary analysis faces this issue. Future studies constructed with the specific purpose of audience segmentation of risk perception of climate change would be a natural remedy for this.
A third limitation is the exclusive reliance on risk attitudes variables solely, rather than a set of measurement such as knowledge, policy preferences, and behavioural responses. Although this avoided confounding attitudes towards risks of climate change with other climate literacy issues and producing results relating to the behavioural dimension by introducing additional variance into the analysis, this also oversimplifies attitudes that are multi-faceted and complex. For example, an intervention based around increasing people’s knowledge of the scientific evidence is only relevant if they do not have that knowledge. Undoubtedly, such variables do have a relationship with climate change attitude cluster membership and introducing them into the analysis may increase the predictive power of behaviour change models (Metag and Schäfer, 2018). Future research may consider using a broader array of constructs to draw a whole picture of climate change risk attitudes typology. Although cluster analysis is an effective method for identifying homogenous groups (with the underlying assumption that distinct groups exist), the approach is not entirely objective. When performing the statistical procedures, some subjective decisions were nevertheless required. For example, the selection of which variables to use as input, the algorithms, and the centroids for the k-clusters (see Hartigan and Wong, 1979).
It is also worth mentioning that method biases arise from having a common rater, a common measurement context, and the characteristics of the climate change risk attitudes items themselves are likely to be a powerful in influence people’s responses. For example, face-to-face interviews—as employed by the UKHLS—may induce consistency motif issues, and lower accuracy than self-completed questionnaires. Interviewer characteristics, expectations, and verbal idiosyncrasies may also cause response biases (Podsakoff et al., 2003). Lastly, the results are specific to the UK population and may not generalise to other countries. Specifically, environmental concern within developing countries may have very different structures. Future comparative studies may pay attention to variation in the typologies of ACCRs between developed and developing countries.
Conclusion
This study produced a classification of ACCRs as they were during a period when the UK slashed carbon emissions. We used survey data from the UKHLS to place people into three discrete segments using k-means cluster analysis: Sceptical, Concerned, and Paradoxical. The Sceptical group, see climate change threats as overblown and agree with the statement that the crisis has been exaggerated though still recognising the existence of climate change. The Concerned group displayed an overwhelmingly urgent worry about climate change risks. The Paradoxical group did not indicate any active attitudinal response toward planning for climate threats though they admit that major environmental disasters might occur if things continue as they are. The overall cluster structure was maintained over the period through 2012 to 2020, however, the members of Paradoxical and Sceptical tended to move toward the Concerned cluster between the two waves of data that we analysed. It is also noteworthy that despite the fact of transition toward concerned attitudes, the Paradoxical group still makes up nearly half of the British population making it worth exploring further with respect of policy interventions, environmental communication and messaging. Corresponding with the three clusters’ pattern differences in ACCRs, respondent segments also differ in age, sex, income, education, and political affiliation. Further research is needed to shed light on the social characteristics and context of people who transitioned in one direction compared to the other.
Acknowledgements
This project was supported by the Economic and Social Research Council [Grant reference number: ES/P000665/1].
Author contributions
All authors contributed to the study conception, design, material preparation, and data analysis. The first draft of the manuscript was written by TL. NS and ME commented, revised, and edited on previous versions of the manuscript. All authors read and approved the final manuscript.
Data availability
The data used in the study are available from the Understanding Society (https://www.understandingsociety.ac.uk/) by registering with and accepting the end user licence at the UK Data Service (https://ukdataservice.ac.uk/).
Ethical approval
Ethical approval was not needed for the present study because this study involves a routinely available, functionally anonymised, secondary dataset (with data subjects having consented to secondary research use).
Informed content
This article does not contain any studies with human participants performed by any of the authors.
Footnotes
1This dataset can be downloaded, ordered or analysed online by registering with and accepting the end user licence at the UK Data Service (https://ukdataservice.ac.uk/).
2The General Population sample is based on two separate samples of residential addresses in England, Scotland and Wales and in Northern Ireland. The Ethnic Minority Boost sample was designed to provide at least 1000 adults from each of five groups: Indian, Pakistani, Bangladeshi, Caribbean, and African. The British Household Panel Survey sample (BHPS) created at Wave 2 consisted of all members from the BHPS (a predecessor survey) who were still active at Wave 18 of the BHPS and who had not refused to take part in UKHLS. The Immigrant and Ethnic Minority Boost Sample was added at Wave 6 and includes people who were born outside the United Kingdom (“immigrants”) and members of five ethnic minority groups: Indian, Pakistani, Bangladeshi, Caribbean, and African. For details about each sample, see the Study Design section of the Main-survey user-guide: https://www.understandingsociety.ac.uk/documentation/mainstage/user-guides/main-survey-user-guide/.
3There is also a household questionnaire and a youth questionnaire, which collects information about households and participants aged 10–15 years old, respectively. However, these are not our target population, so these data were not used. For details about the questionnaires, see Interview Process in the Main-survey user-guide: https://www.understandingsociety.ac.uk/documentation/mainstage/user-guides/main-survey-user-guide/.
4We follow van der Linden’s (2017) definition of ACCRs and select five ACCRs variables because they evaluated how people perceive and are concerned about the timescale, trend, consequences, and crisis of climate change. Variables were not included in our analysis if they only captured attitudes towards climate change (as our aim is to capture how people perceive (and worry about) the risk of climate change, not actions impacting climate change and the relationship between climate change attitudes and behaviour Furthermore, statements that explicitly concerned attitudes to the general environment without referencing climate change were also excluded. Responses to such statements are indicative of participants’ opinions on their lifestyle or habits. Including such variables in our analysis would likely introduce additional variance into the analysis, conflating attitudes towards risks of climate change from other environmental issues and producing results relating to a behavioural dimension. Undoubtedly, such variables do have a relationship with climate change attitudes, but they are likely to be confounded. We did however include two items, which refer to the “environmental crisis”—our rationale here was that because of the extensive and ongoing media engagement with the topic climate change is likely to be at the forefront of respondents’ minds when considering forthcoming crises.
We also considered but declined to use a sixth item, capturing people’s perception of climate change risk on a distant timescale: ‘People in the UK will be affected by climate change in the next 200 years’. We decided not to use it because it asked about timescales far beyond the predictive capacity of even our best scientific climate models, and it had a relatively low item-rest correlation (rwave4=
0.39, rwave10
=
0.38).
5According to the user-guide for UKHLS dataset of Understanding Society, income variables can be negative due to self-employment reported losses or those dedicated for missing values. A value of 0 means their income was 0 and negative values except for missing values are actual negative values reported. In our study, only 0.10% of the respondents in wave 4 and 0.05% of the respondents in wave 10 reported negative values for income. For more details on negative value and missing values of variables please see https://www.understandingsociety.ac.uk/documentation/mainstage/user-guides/main-survey-user-guide/missing-values.
65=
Degree, including first and higher degrees (e.g., MSc, Ph.D), also foundation degrees, graduate membership of a professional Institute, PGCE. 4
=
Other higher education, including diplomas, teaching qualifications (excluding PGCE), and nursing or other medical qualifications. 3
=
A/AS Levels, Welsh and international baccalaureates, highers/advanced highers (Scotland), and certificate of sixth year studies. 2
=
GCSE/O level, STANDARD/Ordinary (O) grade/Lower (Scotland), CSE, and other school qualifications. 1
=
No qualification mentioned.
7When classifying parties based on ideology in waves 4 and 10 data, we relied on the variable “LRGEN” in the 2014, 2017, and 2019 Chapel Hill Expert Survey dataset, which asked experts to rate the positions of each party in terms of its overall ideological stance, with 0 meaning extreme left, 10 meaning extreme right. We define left parties as those that score below 5 on this scale and right parties as those above 5, with a score of 5 being classed as centre. For more details on the recoding of other parties please see Supplementary Appendix B. Wave 4 data was collected from 2012 to 2014. Although this is a sort of period in the UK politics where the liberal democrat party went into a government with conservative party, the average placement of liberal democrat party scored at 4.86 at 2014. For detailed information on the political spectrum see Supplementary Appendix B and https://www.chesdata.eu/2014-chapel-hill-expert-survey, https://yougov.co.uk/topics/politics/articles-reports/2014/07/23/britains-changing-political-spectrum, https://www.chesdata.eu/2019-chapel-hill-expert-survey, https://www.chesdata.eu/2017-chapel-hill-expert-survey, https://www.belfastlive.co.uk/news/uup-needs-reclaim-ground-centre-20574823, “The Independent, 22 April 2019”. Archived from the original on 23 April 2019, Jarrett (2016).
8We also used the Hausman test to examine the Independence of Irrelevant Alternatives (IIA) for the multinomial logistic regression model, excluding the Sceptical cluster and Paradoxical cluster, respectively. The Hausman test was significant at the 0.05 level when excluding the larger paradoxical cluster but was not significant when excluding the smaller sceptical cluster in two waves. In both cases the differences in estimated coefficients were very small and did not affect our conclusions.
9We also carried out a sensitivity analysis of k-means clustering and multinomial logistic regression between raw data and weighted sample, which applied survey weight constructed by combining design weights and non-response weights. Results showed that three-cluster means patterns of ACCRs and odds ratios of age, sex, education, income, and political affiliation on cluster membership have not been changed. Given that waves 4 and 10 data has no over-represented socio-demographics compared to population proportions, we chose to present the results using the unweighted sample in the paper.
10We tested for multicollinearity among predictors (sex, age, income, education, political affiliation) of cluster membership using multiple regression analysis; tolerance was 0.870 in wave 4 and 0.901 in wave 10 data, indicating no collinearity problem.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
The online version contains supplementary material available at 10.1057/s41599-022-01287-1.
References
- Akerlof K, Maibach EW, Fitzgerald D, Cedeno AY, Neuman A. Do people “personally experience” global warming, and if so how, and does it matter? Glob Environ Chang. 2013;23(1):81–91. 10.1016/j.gloenvcha.2012.07.006. [CrossRef] [Google Scholar]
- Althaus SL, Cizmar AM, Gimpel JG. Media supply, audience demand, and the geography of news consumption in the United States. Polit Commun. 2009;26(3):249–277. 10.1080/10584600903053361. [CrossRef] [Google Scholar]
- Andor M, Fels K. Behavioral economics and energy conservation—a systematic review of non-price interventions and their causal effects. Ecol Econ. 2018;148:178–210. 10.1016/j.ecolecon.2018.01.018. [CrossRef] [Google Scholar]
- Arbuckle JG, Morton LW, Hobbs J. Farmer beliefs and concerns about climate change and attitudes toward adaptation and mitigation: evidence from Iowa. Clim Change. 2013;118(3–4):551–563. 10.1007/s10584-013-0700-0. [CrossRef] [Google Scholar]
- Ashworth P, Jeanneret T, Gardner J, Shaw H. Communication and climate change: what the Australian public thinks. Pullenvale: CSIRO; 2011. [Google Scholar]
- Ballew MT, Pearson AR, Goldberg MH, Rosenthal SA, Leiserowitz A. Does socioeconomic status moderate the political divide on climate change? The roles of education, income, and individualism. Glob Environ Chang. 2020;60:102024. 10.1016/j.gloenvcha.2019.102024. [CrossRef] [Google Scholar]
- Bauer DJ, Curran PJ. The integration of continuous and discrete latent variable models: potential problems and promising opportunities. Psychol Methods. 2004;9(1):3–29. 10.1037/1082-989X.9.1.3. [Abstract] [CrossRef] [Google Scholar]
- Boag AE, Hamilton LC, Hartter J, Stevens FR, Palace MW, Ducey MJ. Shifting environmental concern in rural eastern Oregon: the role of demographic and place-based factors. Popul Environ. 2016;38(2):207–216. 10.1007/s11111-016-0261-z. [CrossRef] [Google Scholar]
- Bohr J. Public views on the dangers and importance of climate change: predicting climate change beliefs in the United States through income moderated by party identification. Clim Change. 2014;126:217–227. 10.1007/s10584-014-1198-9. [CrossRef] [Google Scholar]
- Bord RJ, Fisher A, O’Connor RE. Public perceptions of global warming: United States and international perspectives. Clim Res. 1998;11(1):75–84. 10.3354/cr011075. [CrossRef] [Google Scholar]
- Brody SD, Zahran S, Vedlitz A, Grover H. Examining the relationship between physical vulnerability and public perceptions of global climate change in the United States. Environ Behav. 2008;40(1):72–95. 10.1177/0013916506298800. [CrossRef] [Google Scholar]
- Clements B. Exploring public opinion on the issue of climate change in Britain. Br Polit. 2012;7:183–202. 10.1057/bp.2012.1. [CrossRef] [Google Scholar]
- Coelho F, Pereira MC, Cruz L, Simões P, Barata E. Affect and the adoption of pro-environmental behaviour: a structural model. J Environ Psychol. 2017;54:127–138. 10.1016/j.jenvp.2017.10.008. [CrossRef] [Google Scholar]
- Committe on Climate Change (2019) Net Zero: The UK’s contribution to stopping global warming. Committe on Climate Change
- Crawley S, Coffé H, Chapman R. Public opinion on climate change: Belief and concern, issue salience and support for government action. Br J Polit Int Relations. 2020;22(1):102–121. 10.1177/1369148119888827. [CrossRef] [Google Scholar]
- Davidson DJ, Haan M. Gender, political ideology, and climate change beliefs in an extractive industry community. Popul Environ. 2012;34(2):217–234. 10.1007/s11111-011-0156-y. [CrossRef] [Google Scholar]
- DEFRA (2008) A framework for pro-enviromental behaviours report. www.defra.gov.uktel:02072386000Website:www.defra.gov.uk or http://www.opsi.gov.uk/click-use/value-added-licence-information/index.htm
- Detenber BH, Rosenthal S (2020) Climate change audience segmentation: an international review. In: Research handbook on communicating climate change. Edward Elgar Publishing
- Diehl T, Barnidge M, Gil de Zuniga H. Multi-platform news use and political participation across age groups: toward a valid metric of platform diversity and its effects. Journal Mass Commun Q. 2019;96(2):428–451. 10.1177/1077699018783960. [CrossRef] [Google Scholar]
- Ding D, Maibach EW, Zhao X, Roser-Renouf C, Leiserowitz A. Support for climate policy and societal action are linked to perceptions about scientific agreement. Nat Clim Chang. 2011;1(9):462–466. 10.1038/nclimate1295. [CrossRef] [Google Scholar]
- Drummond C, Fischhoff B. Individuals with greater science literacy and education have more polarized beliefs on controversial science topics. Proc Natl Acad Sci USA. 2017;114(36):9587–9592. 10.1073/pnas.1704882114. [Europe PMC free article] [Abstract] [CrossRef] [Google Scholar]
- Dunlap RE. Climate change skepticism and denial: an introduction. Am Behav Sci. 2013;57(6):691–698. 10.1177/0002764213477097. [Europe PMC free article] [Abstract] [CrossRef] [Google Scholar]
- Dutta-Bergman MJ. Complementarity in consumption of news types across traditional and new media. J Broadcast Electron Media. 2004;48(1):41–60. 10.1207/s15506878jobem4801_3. [CrossRef] [Google Scholar]
- Ehret PJ, Sparks AC, Sherman DK. Support for environmental protection: an integration of ideological-consistency and information-deficit models. Env Polit. 2017;26:253–277. 10.1080/09644016.2016.1256960. [CrossRef] [Google Scholar]
- Gifford R, Scannell L, Kormos C, et al. Temporal pessimism and spatial optimism in environmental assessments: an 18-nation study. J Environ Psychol. 2009;29(1):1–12. 10.1016/j.jenvp.2008.06.001. [CrossRef] [Google Scholar]
- Gökşen F, Adaman F, Ünal Zenginobuz E. On environmental concern, willingness to pay, and postmaterialist values: evidence from Istanbul. Environ Behav. 2002;34(5):616–633. 10.1177/0013916502034005003. [CrossRef] [Google Scholar]
- Goldberg M, Gustafson A, Rosenthal S, Kotcher J, Maibach E, Leiserowitz A. For the first time, the Alarmed are now the largest of Global Warming’s Six Americas. New Haven, CT: Yale Program on Climate Change Communication, Yale University and George Mason University; 2020. [Google Scholar]
- Goto Gray S, Raimi KT, Wilson R, Árvai J, Ross SM. Will Millennials save the world? The effect of age and generational differences on environmental concern. J Environ Manage. 2019;242:394–402. 10.1016/j.jenvman.2019.04.071. [Abstract] [CrossRef] [Google Scholar]
- Gregersen T, Doran R, Böhm G, Tvinnereim E, Poortinga W (2020) Political Orientation Moderates the Relationship Between Climate Change Beliefs and Worry About Climate Change. Front Psychol 11. 10.3389/fpsyg.2020.01573 [Europe PMC free article] [Abstract]
- Hamilton LC. Education, politics and opinions about climate change evidence for interaction effects. Clim Change. 2011;104(2):231–242. 10.1007/s10584-010-9957-8. [CrossRef] [Google Scholar]
- Hartigan A, Wong MA. A K-means clustering algorithm. J R Stat Soc. 1979;28(1):100–108. [Google Scholar]
- Heath Y, Gifford R. Free-market ideology and environmental degradation: The case of belief in global climate change. Environ Behav. 2006;38(1):48–71. 10.1177/0013916505277998. [CrossRef] [Google Scholar]
- Heberlein TA. Navigating Environmental Attitudes. Oxford: Oxford University Press; 2012. [Abstract] [Google Scholar]
- Hine DW, Reser JP, Morrison M, Phillips WJ, Nunn P, Cooksey R. Audience segmentation and climate change communication: conceptual and methodological considerations. Wiley Interdiscip Rev Clim Chang. 2014;5(4):441–459. 10.1002/wcc.279. [CrossRef] [Google Scholar]
- Hoffmann R, Muttarak R. Learn from the past, prepare for the future: impacts of education and experience on disaster preparedness in the Philippines and Thailand. World Dev. 2017;96:32–51. 10.1016/j.worlddev.2017.02.016. [CrossRef] [Google Scholar]
- Hornsey MJ, Harris EA, Bain PG, Fielding KS. Meta-analyses of the determinants and outcomes of belief in climate change. Nat Clim Chang. 2016;6:622–626. 10.1038/nclimate2943. [CrossRef] [Google Scholar]
- Jarrett H. The Single Transferable Vote and the Alliance Party of Northern Ireland. Representation. 2016;52(4):311–323. 10.1080/00344893.2017.1301987. [CrossRef] [Google Scholar]
- Kácha O, Jáchym V, Cameron B. Four Europes: climate change beliefs and attitudes predict behavior and policy preferences using a latent class analysis on 23 countries. J Environ Psychol. 2022;81:10181. 10.31234/osf.io/rfz98. [CrossRef] [Google Scholar]
- Kahan DM, Braman D, Gastil J, Slovic P, Mertz CK. Culture and identity‐protective cognition: explaining the white‐male effect in risk perception. J Empir Leg Stud. 2007;4(3):465–505. 10.1111/j.1740-1461.2007.00097.x. [CrossRef] [Google Scholar]
- Kahan DM (2017) Misconceptions, misinformation and the logic of identity-protective cognition, Yale Law & Economics Research Paper. No. 575
- Kenward B, Brick C. Even Conservative voters prefer the environment to be at the heart of post-COVID-19 economic reconstruction in the UK. J Soc Political Psychol. 2021;9(1):321–333. 10.5964/jspp.6917. [CrossRef] [Google Scholar]
- Kerr RA. Pushing the scary side of global warming. Science. 2007;316:1412–1415. 10.1126/science.316.5830.1412. [Abstract] [CrossRef] [Google Scholar]
- Kraus MW, Stephens NM. A road map for an emerging psychology of social class. Soc Personal Psychol Compass. 2012;6:642–656. 10.1111/j.1751-9004.2012.00453.x. [CrossRef] [Google Scholar]
- Lee TM, Markowitz EM, Howe PD, Ko CY, Leiserowitz AA. Predictors of public climate change awareness and risk perception around the world. Nat Clim Chang. 2015;5:1014–1020. 10.1038/nclimate2728. [CrossRef] [Google Scholar]
- Leiserowitz A. Climate change risk perception and policy preferences: The role of affect, imagery, and values. Clim Change. 2006;77(1–2):45–72. 10.1007/s10584-006-9059-9. [CrossRef] [Google Scholar]
- Leiserowitz A, Maibach E, Roser-Renouf C (2009) Global warming’s six Americas 2009: An audience segmentation analysis. Yale University and George Mason University. Available at: http://climatecom-munication.yale.edu/publications/global-warmings-six-americas-2009/ (accessed 1 Jun 2021)
- Leiserowitz A, Maibach E, Roser-Renouf C, Smith N (2011) Global warming’s six Americas. Yale University and George Mason University
- Liere KDV, Dunlap RE. The social bases of environmental concern: a review of hypotheses, explanations and empirical evidence. Public Opin Q. 1980;44(2):181–197. 10.1086/268583. [CrossRef] [Google Scholar]
- Maibach EW, Leiserowitz A, Roser-Renouf C, Mertz CK. Identifying like-minded audiences for global warming public engagement campaigns: an audience segmentation analysis and tool development. PLoS ONE. 2011;6(3):e17571. 10.1371/journal.pone.0017571. [Europe PMC free article] [Abstract] [CrossRef] [Google Scholar]
- Makles A. Stata tip 110: how to get the optimal k-means cluster solution. Stata J. 2012;12(2):347–351. 10.1177/1536867x1201200213. [CrossRef] [Google Scholar]
- Malka A, Krosnick JA, Langer G. The association of knowledge with concern about global warming: trusted information sources shape public thinking. Risk Anal. 2009;29(5):633–647. 10.1111/j.1539-6924.2009.01220.x. [Abstract] [CrossRef] [Google Scholar]
- Marks E, Hickman C, Pihkala P, Clayton S, Lewandowski ER, Mayall EE, Wray B, Mellor C, van Susteren L (2021) Young people’s voices on climate anxiety, government betrayal and moral injury: a global phenomenon. SSRN Electron J. 10.2139/SSRN.3918955
- Marshall G (2005) Sleepwalking into disaster—Are we in a state of denial about climate change? Available at: http://www.networkforclimateaction.org.uk/toolkit/positive_alternatives/people_and_society/COIN_Are_we_in_a_State_of_Denial_about_Climate_Change.pdf (Accessed 15 August 2022).
- Maslow A. Motivation and personality. 2nd edn. New York, NY: Viking Press; 1970. [Google Scholar]
- Mayer A, Smith EK. Unstoppable climate change? The influence of fatalistic beliefs about climate change on behavioural change and willingness to pay cross-nationally. Clim Policy. 2019;19(4):511–523. 10.1080/14693062.2018.1532872. [CrossRef] [Google Scholar]
- McCright AM. Political orientation moderates Americans’ beliefs and concern about climate change. Clim Change. 2011;104:243–253. 10.1007/s10584-010-9946-y. [CrossRef] [Google Scholar]
- McCright AM. The effects of gender on climate change knowledge and concern in the American public. Popul Environ. 2010;32(1):66–87. 10.1007/s11111-010-0113-1. [CrossRef] [Google Scholar]
- McCright AM, Dunlap RE. Cool dudes: the denial of climate change among conservative white males in the United States. Glob Environ Chang. 2011;21(4):1163–1172. 10.1016/j.gloenvcha.2011.06.003. [CrossRef] [Google Scholar]
- McCright AM, Dunlap RE (2011b) The politicization of climate change and polarization in the American public’s views of global warming, 2001–2010. Sociol Q 10.1111/j.1533-8525.2011.01198.x
- McCright AM, Dunlap RE, Xiao C (2013) Perceived scientific agreement and support for government action on climate change in the USA. Clim Change. 10.1007/s10584-013-0704-9
- McCright AM, Dunlap RE, Xiao C. Increasing influence of party identification on perceived scientific agreement and support for government action on climate change in the united states, 2006-12. Weather Clim Soc. 2014;52(2):155–194. 10.1175/WCAS-D-13-00058.1. [CrossRef] [Google Scholar]
- McCright AM, Dunlap RE, Xiao C. The impacts of temperature anomalies and political orientation on perceived winter warming. Nat Clim Chang. 2014;4(12):1077–1081. 10.1038/nclimate2443. [CrossRef] [Google Scholar]
- McEvoy J. The American concern with the environment. In: Burch WR, Cheek NH, Taylor L, editors. Social behavior, natural resources, and the environment. New York, NY: Harper and Row; 1972. pp. 214–236. [Google Scholar]
- McLachlan GJ, Peel D. Finite mixture models. New York, NY: Wiley; 2000. [Google Scholar]
- McNeish W. From revelation to revolution: apocalypticism in green politics. Env Polit. 2017;26:1035–1054. 10.1080/09644016.2017.1343766. [CrossRef] [Google Scholar]
- Melis G, Elliot M, Shryane N. Environmental Concern Over Time: Evidence from the Longitudinal Analysis of a British Cohort Study from 1991 to 2008. Soc Sci Q. 2014;95(4):905–919. 10.1111/ssqu.12107. [CrossRef] [Google Scholar]
- Metag J, Füchslin T, Schäfer MS. Global warming’s five Germanys: A typology of Germans’ views on climate change and patterns of media use and information. Public Underst Sci. 2017;26(4):434–451. 10.1177/0963662515592558. [Abstract] [CrossRef] [Google Scholar]
- Metag J, Schäfer MS. Audience segments in environmental and science communication: recent findings and future perspectives. Environ Commun. 2018;12(8):995–1004. 10.1080/17524032.2018.1521542. [CrossRef] [Google Scholar]
- Milfont TL. The interplay between knowledge, perceived efficacy, and concern about global warming and climate change: a one-year longitudinal study. Risk Anal. 2012;32(6):1003–1020. 10.1111/j.1539-6924.2012.01800.x. [Abstract] [CrossRef] [Google Scholar]
- Milfont TL, Milojev P, Greaves LM, Sibley CG. Socio-structural and psychological foundations of climate change beliefs. NZ J Psychol. 2015;44(1):17–30. [Google Scholar]
- Milfont TL, Sibley CG. Empathic and social dominance orientations help explain gender differences in environmentalism: A one-year Bayesian mediation analysis. Pers Individ Dif. 2016;90:85–88. 10.1016/j.paid.2015.10.044. [CrossRef] [Google Scholar]
- Morris BS, Chrysochou P, Karg ST et al (2020) Optimistic vs. pessimistic endings in climate change appeals. Humanit Soc Sci Commun 8(82). 10.1057/s41599-020-00574-z
- Myers TA, Nisbet MC, Maibach EW, Leiserowitz AA. A public health frame arouses hopeful emotions about climate change: a Letter. Clim Change. 2012;113:1105–1112. 10.1007/s10584-012-0513-6. [CrossRef] [Google Scholar]
- O’Connor RE, Bord RJ, Fisher A. Risk perceptions, general environmental beliefs, and willingness to address climate change. Risk Anal. 1999;19:461–471. 10.1023/A:1007004813446. [CrossRef] [Google Scholar]
- Pahl S, Harris PR, Todd HA, Rutter DR. Comparative optimism for environmental risks. J Environ Psychol. 2005;25(1):1–11. 10.1016/j.jenvp.2004.12.004. [CrossRef] [Google Scholar]
- Pew Research Center (2015) Global concern about climate change, broad support for limiting emissions partisan divides in key countries (Issue November). www.pewresearch.org
- Phillips D, Curtice J, Phillips M, Perry J (2018) British social attitudes: climate change. The 35th Report. https://www.bsa.natcen.ac.uk/media/39251/bsa35_climate_change.pdf
- Podsakoff PM, MacKenzie SB, Lee JY, Podsakoff NP. Common method biases in behavioral research: a critical review of the literature and recommended remedies. J Appl Psychol. 2003;88(5):879–903. 10.1037/0021-9010.88.5.879. [Abstract] [CrossRef] [Google Scholar]
- Poortinga W, Pidgeon N (2003) Public perceptions of risk, science and governance: main findings of a British Survey of Five risk Cases (technical report). Centre for Environmental Risk, University of East Anglia, Norwich
- Poortinga W, Pidgeon N, Capstick S, Aoyagi M (2014) Public attitudes to nuclear power and climate change in Britain two years after the Fukushima accident synthesis Report. www.ukerc.ac.uk/support/TheMeetingPlace
- Poortinga W, Spence A, Whitmarsh L, Capstick S, Pidgeon NF. Uncertain climate: an investigation into public scepticism about anthropogenic climate change. Glob Environ Chang. 2011;21(3):1015–1024. 10.1016/j.gloenvcha.2011.03.001. [CrossRef] [Google Scholar]
- Poudyal NC, Joshi O, Hodges DG, Bhandari H, Bhattarai P. Climate change, risk perception, and protection motivation among high-altitude residents of the Mt. Everest region in Nepal. Ambio. 2020;50:505–518. 10.1007/s13280-020-01369-x. [Europe PMC free article] [Abstract] [CrossRef] [Google Scholar]
- Rhead R, Elliot M, Upham P. Using latent class analysis to produce a typology of environmental concern in the UK. Soc Sci Res. 2018;74:210–222. 10.1016/j.ssresearch.2018.06.001. [Abstract] [CrossRef] [Google Scholar]
- Rosenthal S, Linder N. Effects of bin proximity and informational prompts on recycling and contamination. Resour Conserv Recycl. 2021;168:105430. 10.1016/j.resconrec.2021.105430. [CrossRef] [Google Scholar]
- Rosenthal S. Information sources, perceived personal experience, and climate change beliefs. J Environ Psychol. 2022;81:272–4944. 10.1016/j.jenvp.2022.101796. [CrossRef] [Google Scholar]
- Roser-Renouf C, Stenhouse N, Rolfe-Redding J, Maibach EW, Leiserowitz A (2014) Engaging diverse audiences with climate change: message strategies for global warming’s six Americas. Available at SSRN: 10.2139/ssrn.2410650
- Schwarz A. Lokale scoring-modelle. Lohmar: Eul Verlag; 2008. [Google Scholar]
- Sciullo A, Arrobbio O, Padovan D. Exploring the heterogeneous facets of individual environmental attitudes. Polis (Italy) 2019;2:267–294. 10.1424/94248. [CrossRef] [Google Scholar]
- Shehata A, Strömbäck J. Learning political news from social media: Network media logic and current affairs news learning in a high-choice media environment. Communic Res. 2021;48(1):125–147. 10.1177/0093650217749354. [CrossRef] [Google Scholar]
- Sibley CG, Kurz T. A model of climate belief profiles: How much does it matter if people question human causation? Anal Soc Issues Public Policy. 2013;13(1):245–261. 10.1111/asap.12008. [CrossRef] [Google Scholar]
- Slimak MW, Dietz T. Personal values, beliefs, and ecological risk perception. Risk Anal. 2006;26(6):1689–1705. 10.1111/j.1539-6924.2006.00832.x. [Abstract] [CrossRef] [Google Scholar]
- Slovic P (2000) The Perception of Risk Author. In Earthscane, Routledge
- Smith TW, Kim J, Son J. Public attitudes toward climate change and other environmental issues across countries. Int J Sociol. 2017;47(1):62–80. 10.1080/00207659.2017.1264837. [CrossRef] [Google Scholar]
- Spence A, Poortinga W, Butler C, Pidgeon NF. Perceptions of climate change and willingness to save energy related to flood experience. Nat Clim Chang. 2011;1:46–49. 10.1038/nclimate1059. [CrossRef] [Google Scholar]
- Taber CS, Lodge M. Motivated skepticism in the evaluation of political beliefs. Am J Pol Sci. 2006;50:755–769. 10.1111/j.1540-5907.2006.00214.x. [CrossRef] [Google Scholar]
- Taylor AL, Dessai S, Bruine de Bruin W. Public perception of climate risk and adaptation in the UK: a review of the literature. Clim Risk Manag. 2014;4-5:1–16. 10.1016/j.crm.2014.09.001. [CrossRef] [Google Scholar]
- Thornton A, Evans L, Bunt K, Simon A, King S, Webster T (2011) Climate change and transport choices segmentation model-A framework for reducing CO2 emissions from personal travel. http://docshare04.docshare.tips/files/12446/124468196.pdf
- Tobler C, Visschers VHM, Siegrist M. Consumers’ knowledge about climate change. Clim Change. 2012;114:189–209. 10.1007/s10584-011-0393-1. [CrossRef] [Google Scholar]
- Trogrlić RŠ, Duncan M, Wright G, van den Homberg M, Adeloye A, Mwale F, McQuistan C. External stakeholders’ attitudes towards and engagement with local knowledge in disaster risk reduction: are we only paying lip service? Int J Disaster Risk Sci. 2021;58:2212–4209. 10.1016/j.ijdrr.2021.102196. [CrossRef] [Google Scholar]
- United Nations (2015) Adoption of the Paris agreement. Conf Parties its twenty-first Sess. https://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf
- van der Linden S. The social-psychological determinants of climate change risk perceptions: towards a comprehensive model. J Environ Psychol. 2015;41:112–124. 10.1016/j.jenvp.2014.11.012. [CrossRef] [Google Scholar]
- van der Linden S. Determinants and measurement of climate change risk perception, worry, and concern. In: Nisbet MC, Schafer M, Markowitz E, Ho S, O’Neill S, Thaker J, editors. The Oxford encyclopedia of climate change communication. Oxford: Oxford University Press; 2017. [Google Scholar]
- Van Vugt M, Griskevicius V, Schultz PW. Naturally green: Harnessing stone age psychological biases to foster environmental behavior. Soc Issues Policy Rev. 2014;8(1):1–32. 10.1111/sipr.12000. [CrossRef] [Google Scholar]
- Vara-Miguel A. Cross-national similarities and differences between legacy and digital-born news media audiences. Media Commun. 2020;8(2):16–27. 10.17645/mac.v8i2.2733. [CrossRef] [Google Scholar]
- Vermunt JK. K-means may perform as well as mixture model clustering but may also be much worse: comment on Steinley and Brusco. Psychol Methods. 2011;16(1):82–88. 10.1037/a0020144. [Abstract] [CrossRef] [Google Scholar]
- Waitt G, Caputi P, Gibson C, Farbotko C, Head L, Gill N, Stanes E. Sustainable household capability: which households are doing the work of environmental sustainability? Aust Geogr. 2012;43(1):51–74. 10.1080/00049182.2012.649519. [CrossRef] [Google Scholar]
- Weber EU. What shapes perceptions of climate change? Wiley Interdiscip Rev Clim Chang. 2010;1(3):332–342. 10.1002/wcc.41. [CrossRef] [Google Scholar]
- Weber EU. Psychology: climate change hits home. Nat Clim Chang. 2011;1:25–26. 10.1038/nclimate1070. [CrossRef] [Google Scholar]
- Wendel J (2016) Climate scientists’ New Hurdle: overcoming climate change apathy. Eos. 10.1029/2016eo057547
- Whitmarsh L. Scepticism and uncertainty about climate change: dimensions, determinants and change over time. Glob Environ Chang. 2011;21(2):690–700. 10.1016/j.gloenvcha.2011.01.016. [CrossRef] [Google Scholar]
- Wonneberger A, Meijers MHC, Schuck ART. Shifting public engagement: how media coverage of climate change conferences affects climate change audience segments. Public Underst Sci. 2019;29(2):176–193. 10.1177/0963662519886474. [Europe PMC free article] [Abstract] [CrossRef] [Google Scholar]
- Xiao C, McCright AM. Explaining gender differences in concern about environmental problems in the United States. Soc Nat Resour. 2012;25(11):1067–1084. 10.1080/08941920.2011.651191. [CrossRef] [Google Scholar]
- Zahran S, Brody SD, Grover H, Vedlitz A. Climate change vulnerability and policy support. Soc Nat Resour. 2006;19(9):771–789. 10.1080/08941920600835528. [CrossRef] [Google Scholar]
Full text links
Read article at publisher's site: https://doi.org/10.1057/s41599-022-01287-1
Read article for free, from open access legal sources, via Unpaywall:
https://www.nature.com/articles/s41599-022-01287-1.pdf
Citations & impact
Impact metrics
Alternative metrics
Discover the attention surrounding your research
https://www.altmetric.com/details/134696412
Article citations
Climate Change Influence on Health Domains, a Public Perception from a District in South India - A Cross-sectional Study.
Indian J Community Med, 49(5):752-758, 03 Jul 2024
Cited by: 0 articles | PMID: 39421520 | PMCID: PMC11482395
The Green Surgery report: a guide to reducing the environmental impact of surgical care, but will it be implemented?
Ann R Coll Surg Engl, 106(6):475-477, 29 Apr 2024
Cited by: 0 articles | PMID: 38683381
Similar Articles
To arrive at the top five similar articles we use a word-weighted algorithm to compare words from the Title and Abstract of each citation.
Climate change-related knowledge and attitudes among a sample of the general population in Egypt.
Front Public Health, 10:1047301, 03 Nov 2022
Cited by: 5 articles | PMID: 36408006 | PMCID: PMC9669343
Development of the Low Emissions Analysis Platform - Integrated Benefits Calculator (LEAP-IBC) tool to assess air quality and climate co-benefits: Application for Bangladesh.
Environ Int, 145:106155, 04 Oct 2020
Cited by: 3 articles | PMID: 33027737
Perception, attitude and behavior in relation to climate change: a survey among CDC health professionals in Shanxi province, China.
Environ Res, 134:301-308, 07 Sep 2014
Cited by: 24 articles | PMID: 25199970
Impact of climate change on the domestic indoor environment and associated health risks in the UK.
Environ Int, 85:299-313, 30 Oct 2015
Cited by: 44 articles | PMID: 26453820
Review