Scholarly article on topic 'Citizens’ preferences for tackling climate change. Quantitative and qualitative analyses of their freely formulated solutions'

Citizens’ preferences for tackling climate change. Quantitative and qualitative analyses of their freely formulated solutions Academic research paper on "Social and economic geography"

Share paper
Academic journal
Global Environmental Change
OECD Field of science
{"climate change" / "public opinion" / "quantitative text analysis" / "qualitative text analysis" / "climate change mitigation" / "climate change adaptation"}

Abstract of research paper on Social and economic geography, author of scientific article — Endre Tvinnereim, Kjersti Fløttum, Øyvind Gjerstad, Mikael Poul Johannesson, Åsta Dyrnes Nordø

Abstract Tackling climate change requires both policy and individual action. Mobilizing such action can be made more optimal with knowledge about how the public views climate change solutions and what they think needs to be done in the face of climate change. Yet most public opinion research to date uses either closed questions about agreement with various pre-determined statements (such as views on science, worry, and support for given policy options) or use open-ended questions eliciting generic associations with climate change. This article uses an open-ended survey question in a probability-based Internet survey panel in Norway, analyzing 4634 textual responses to the question of “what should be done” about climate change. Using structural topic modeling (STM), we induce seven topics: Transportation, energy transition, attribution of climate change, emission reduction, the international dimension, lifestyle/consumption and government measures. We find that Norwegians strongly emphasize mitigation over adaptation, as few responses mention the latter topic. Also, men seem to externalize the solutions to climate change, emphasizing energy policies, the international dimension, and discussions about the causes of climate change, while women to a larger extent understand climate action as an issue involving individual behavior, calling for better public transportation and lifestyle changes. Overall, our results suggest a willingness to accept stronger mitigation action, but also that central and local governments need to facilitate low-carbon choices, bridging policy and individual action to mitigate climate change.

Academic research paper on topic "Citizens’ preferences for tackling climate change. Quantitative and qualitative analyses of their freely formulated solutions"

Contents lists available at ScienceDirect

Global Environmental Change

journal homepage:

Citizens' preferences for tackling climate change. Quantitative and qualitative analyses of their freely formulated solutions


Endre Tvinnereima'*, Kjersti Fl0ttumb Âsta Dyrnes Nord0c

0yvind Gjerstad , Mikael Poul Johannessonc

a Uni Research Rokkan Centre, Bergen, Norway b Department of Foreign Languages, University of Bergen, Norway c Department of Comparative Politics, University of Bergen, Norway


Keywords: climate change public opinion quantitative text analysis qualitative text analysis climate change mitigation climate change adaptation


Tackling climate change requires both policy and individual action. Mobilizing such action can be made more optimal with knowledge about how the public views climate change solutions and what they think needs to be done in the face of climate change. Yet most public opinion research to date uses either closed questions about agreement with various pre-determined statements (such as views on science, worry, and support for given policy options) or use open-ended questions eliciting generic associations with climate change. This article uses an open-ended survey question in a probability-based Internet survey panel in Norway, analyzing 4634 textual responses to the question of "what should be done" about climate change. Using structural topic modeling (STM), we induce seven topics: Transportation, energy transition, attribution of climate change, emission reduction, the international dimension, lifestyle/consumption and government measures. We find that Norwegians strongly emphasize mitigation over adaptation, as few responses mention the latter topic. Also, men seem to externalize the solutions to climate change, emphasizing energy policies, the international dimension, and discussions about the causes of climate change, while women to a larger extent understand climate action as an issue involving individual behavior, calling for better public transportation and lifestyle changes. Overall, our results suggest a willingness to accept stronger mitigation action, but also that central and local governments need to facilitate low-carbon choices, bridging policy and individual action to mitigate climate change.

1. Public opinion and climate change

Due to the complex nature of climate change, it is to be expected that views on it are diverse and potentially unformed (Hulme, 2009). Much opinion research on climate change has mirrored that of the public debate, focusing on agreement and disagreement with climate scientists, concern and support for mitigation, and, to a lesser extent, adaptation policies. A recent literature has taken a more holistic approach, focusing on individual associations and interpretations (Capstick et al., 2016; Morton et al., 2011; Tvinnereim and Fl0ttum, 2015). Given the increasing public acknowledgement that climate change is happening, a natural next step is to look more closely at public opinion on response measures such as adaptation and mitigation. As this issue is complicated technically as well as politically, it is inherently interesting to examine which aspects of the debate citizens internalize and make their own. Also, increased knowledge of what climate policies citizens endorse should be of great interest to policy-

makers to strengthen their chances of successfully developing and implementing comprehensive climate policies and measures.

Effective responses to climate change need to include both government initiatives and individual action. There are myriad potential ways to tackle climate change, crossing all levels of government as well as private and public spheres, and the choice of the portfolio of actions to take is both political and personal. Given varying social norms and trade-offs connected to the various potential courses of action, it is important to know how members of the public think about climate change solutions to choose workable paths forward. Where many paths are possible, navigating not just perceptions of climate change, but also perspectives on possible solutions, is key to identifying the best strategies.

There is a growing body of work on public views on what should be done about climate change. These studies provide a relatively detailed picture of policy preferences, but the information they provide related to high-level trade-offs such as the weight of government vs. individual

* Corresponding author. E-mail address: (E. Tvinnereim).

Received 7 October 2016; Received in revised form 24 May 2017; Accepted 16 June 2017

0959-3780/ © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (

action, or between mitigation and adaptation, is limited.

An initial question is what type of action, if any, should be taken to address climate change. The Intergovernmental Panel on Climate Change (IPCC) divides its work into three main groups: 1) Physical science; 2) impacts, adaptation and vulnerability; and 3) mitigation. This subdivision suggests a bifurcation between adaptation (action to make the effects of climate change less dangerous to humans and ecosystems — such as building flood defenses) and mitigation (action to reduce climate change itself - notably emission reductions and absorption of carbon). Some public opinion studies present choices between mitigation and adaptation (Gellein et al., 2015) but more research is needed to establish how the public would prioritize between the two.

Popular support for various forms of government action has been studied in many countries. One type of study seeks to measure the willingness to pay for reaching certain climate goals (Aldy et al., 2012; Kotchen et al., 2013). Others explain patterns of support for or opposition to various proposed measures (Krosnick and MacInnis, 2013; Tvinnereim and Steinshamn, 2016; Smith and Leiserowitz, 2014; Tvinnereim and Ivarsflaten, 2016). Overall, subsidies for renewable energy are viewed relatively positively by respondents in these studies, whereas taxes are less popular.

Another question regards whether individuals, governments, business, or others should take charge in tackling climate change. Rathzel and Uzzell (2009: 332) find that "[b]oth British and Swedish students invest their hope in their respective governments" although Swedish students are somewhat more likely to see individuals as "the subjects rather than the objects of change." This emphasis on government responsibility is also found in a review of European and US studies (Lorenzoni and Pidgeon, 2006). Additionally, the way in which climate change is framed — say, as a technical or "stewardship" issue — may influence what types of solutions are supported (Wolf and Moser, 2011: 559). Public opinion research has thus provided a number of insights into how people think climate change should be tackled. However, most of the studies reviewed above use set questions defined by researchers, and may thus suffer from bias as they may emphasize different aspects of climate change than respondents would if given a less constraining set of options. Very general questions, such as proposing a choice between adaptation and mitigation, or between government and individual action, could solve this problem but also run the risk of perpetuating false dichotomies.

In this article, we access popular priorities in a more fundamental way by asking open-ended questions. This type of survey questions, while more demanding to answer and analyze, may access more fundamental attitudes and associations than questions with pre-defined response options. Notably for our purposes, textual answers imply a prioritization of relevant themes, as respondents are likely to suggest what is important to them, while leaving out what is less important. This open-ended design represents an exploratory mode of research where categorization and contributions to theory constitute the subsequent step.

Open-ended survey questions to measure attitudes to climate change have become more frequent in recent years (Leiserowitz, 2005; Lorenzoni et al., 2006; Smith and Leiserowitz, 2014; Shwom et al., 2010; Tvinnereim and Fl0ttum, 2015). Whitmarsh (2008) uses both open-ended and closed questions to assess popular associations with climate change and global warming. Relatedly, Capstick et al. (2016) use open-ended survey questions to assess popular knowledge about ocean acidification. An earlier study by Bohm and Pfister (2001) asked respondents to list the remedies they knew for 14 risks including global warming, yielding responses such as traffic reduction, CO2 emission reductions, alternative energy sources, reduction of air pollution, and forest preservation. However, no study has to our knowledge used open-ended survey questions to measure public support for climate change solutions.

We fill this gap in the literature by posing general, open-ended

questions about what should be done about climate change. We then use quantitative text analysis to guide the induction of topics for organizing these open-ended responses. The topic induction is followed by thorough qualitative examination of a large set of possible topic configurations to identify which model is best, in terms of within-topic homogeneity and topic validity. Based on the selected model run, we then examine the relative importance of each topic and correlate topic prevalence with individual-level background variables such as demographics and political attitudes.

2. Data and methods

The data were collected using the Norwegian Citizen Panel, a probability-based Internet panel drawn from all Norwegian residents (Skjervheim and H0gest0l, 2015). The questions, fielded in Waves 4 and 5 in the spring and autumn of 2015, invited citizens to formulate, in their own words, what they think should be done about climate change. The specific wording was: "When it comes to climate change, what do you think should be done?" In the fourth wave (March 9-April 8), 1266 open-ended responses were received; in the fifth wave (Oct 28-Nov 10), a different and larger sub-sample of the Norwegian Citizen Panel was consulted, producing 3368 textual responses. Average response length was 21.5 words, with a median length of 14 words and a mode of two words. The longest response contained 146 words; ninety-eight responses had more than 100 words. The total number of words across all responses was 93,952.

To conduct the analyses, we make use of a semi-automated quantitative text analysis technique named Structural Topic Modelling (STM) (Roberts et al., 2014). Instead of relying solely on human coding of the open-ended answers, we base our analysis on recent developments in machine learning based on analysis of textual data allowing for an inductive search for distinct topics in the text corpus. As such, it "allows the researcher to discover topics from the data, rather than assume them" (Roberts et al., 2014: 1066). Another novelty of STM is that it allows us to incorporate covariates, such as demographics or attribute data, with a topic model that is inferred directly from the written document.

STM generates topics using a clustering algorithm based on the cooccurrence of words across documents. First, the researcher chooses a number of topics, denoted k. The method then assigns to each document a vector with k values, where each value expresses the degree to which the document belongs to topic k. These values sum up to one. For example, a document or survey response may be deemed in a four-topic model to belong 10% to Topic 1, 20% to Topic 2, 30% to Topic 3, and 40% to Topic 4. Multiple membership is thus possible, and each response will typically be estimated by the model to belong to several topics to different degrees.

We define topic prevalence as the degree to which a single response belongs to a given topic, based on the words it contains. Topic prevalence may also be calculated in the aggregate, for each topic, to assess relative topic size, or by other categories such as demographic variables. Analyzing and explaining topic prevalence will constitute the bulk of our analysis and discussion below.

To analyze topic prevalence, we extract the estimated prevalence of each induced topic from the chosen STM model. Together with the variables provided as meta-data in the estimation (age, gender, education, and stated interest in politics) we also merge the estimated topic prevalence with more data about the respondents from the Norwegian Citizen Panel. We use row-wise deletion for respondents with missing values in any of these variables so the initial sample size of 4634 used in the STM estimation is reduced to 4372.

3. Model runs and induced topics

STM produces multi-modal results, meaning that the estimation process may produce several optimal solutions in numerical terms (Roberts et al., 2015). To select a model for further analysis, we chose

Table 1

The selected seven-topic model with most frequent/exclusive words and overall topic prevalence.

Label Topic prevalence Most important words (FREX). See Appendix A for the original Norwegian wording.

(1) Transportation 0.185 Public, better, transport, traffic, airplanes, trains, cheaper

(2) Energy transition 0.164 Energy, renewable, research, oil, alternative, source, develop

(3) Attribution 0.158 Climate change, environment, nature, humankind, alternatively, whole, will

(4) Emission reduction 0.141 Reduce, emissions, consumption, industry, limit, decrease, air traffic

(5) International collaboration 0.131 Countries, help, single/some, responsibility, China, really, USA

(6) Lifestyle/consumption 0.126 Less, food, buy, dispose of, produce, reuse, recycle

(7) Government measures 0.096 Most, economic, create, growth, before, the present, at least

to run and manually evaluate models in the range from three to 12 topics. This range is in line with, and somewhat more conservative than, a preliminary numerical analysis suggesting that semantic coherence (that probable words for a topic should co-occur within a document) drops rapidly with 11 or more topics, whereas topic exclusivity (the conditional probability of seeing a topic given a word) is relatively low with five or fewer topics.

In this range from three to 12, we initially performed 100 STM runs for each topic number, choosing the four best runs automatically from each based on numerical indicators. This produced an aggregate 40 runs for human evaluation. These 40 were subsequently read independently by five researchers who subsequently made blind recommendations and finally determined the best run through consensus. This procedure led to the selection of one STM run with seven topics.

The selected run was assessed to be the best expression of the diversity of topics identified through the manual readings. Table 1 shows the seven topics represented by their most discriminating terms (based on an equal weighting of term frequency and term exclusivity in each topic). The table also displays topic proportion and suggested labels for each topic.

However, it may be useful also to consider a lower number of topics. For example, with three topics we got one run that rendered three comprehensive characteristics of the complete data set: transportation/consumption, attribution and energy. The disadvantage of such a low number of topics, as also indicated by the initial numerical analysis, is that nuances such as the international dimension are lost.

Concerning runs with more than seven topics, they tended to render a division of topics that intuitively belong to one, e.g. transportation divided into two topics. Furthermore, even though our chosen run comes out as the one that matches the data set in the best way, we have noticed potential topics found in other runs that do not come out clearly in our selected set of topics. For example, we identified topics relevant to the issue raised in the question, such as responsibility (who/what is responsible for undertaking the various measures proposed by the respondents) and the need for raising awareness of the challenges among people. These did not come out as clear-cut topics in the models with seven topics.

The seven topics of our chosen model run can be labeled and described as done below. They are ordered according to the cumulative share of the total responses with which they are associated. The three most representative answers are included for each topic.

3.1. Transportation

Entries strongly associated with this topic frequently emphasize public transit and the need to make it more affordable and more accessible. In addition to local transportation such as buses and tramways, there are calls for a better and more affordable railway system. Some respondents also advocate better and more affordable electric vehicles. Responses related to transportation more generally, notably to the emissions caused by road transportation and by aviation, are also found in this topic.

Three most representative answers:

i) Improve public transportation, i.e. lower the prices and build public transit (more tramways, buses, increase frequency of departures, etc.).

ii) Car traffic, transportation of goods by trucks, more cars per household, must be lower. Trains and light rail must be revamped.

iii) Free public transportation in the biggest cities. Facilitate parking in the outskirts and expand public transportation.

3.2. Energy transition

The main narrative among respondents in this category is the need to transition away from fossil fuels to renewables. Among the proposed solutions, we find wind, solar, wave, hydroelectric and nuclear energy. Interestingly, among those who mention the latter there is very little opposition to it.

Three most representative answers:

i.) Stop the extraction of oil; invest in research/development of environmentally friendly technology, transition to renewable energy sources, etc.

ii.) Committing to environmentally friendly energy sources. Committing to the transition to environmentally friendly energy. Forest/rain forest conservation and planting.

iii.) Finish and empty oil fields that have already been started. Use our technology for climate friendly projects. End unprofitable projects such as shale oil. Spend aid money to phase out coal.

3.3. Attribution of climate change

The survey question presupposes the reality of anthropogenic climate change, so it is not surprising that those who choose to focus on the question of attribution tend to do so in order to challenge this presupposition. However, while some deny any human influence on the climate system, others express uncertainty or a belief in concurrent causes, both natural and human.

Three most representative answers:

i.) Climatic changes have happened on earth throughout the ages. Warm times and cold times come and go. Experts preach about one threat after another, but nothing is completely certain. They have been wrong before and are not "all-knowing". [...] We all need to accept periods of change on the globe, whether we want to or not. Man predicts but God/the Universe rules.

ii.) I think we have gone off the rails when it comes to climate change. The climate has changed constantly throughout Earth's history, which is millions of years. It is completely natural that the climate changes, and after a while there will be a new ice age.

iii.) On this point I think I'm outside the accepted opinion. I am not convinced that climate change is human-made. In debates, climate and environment are often confused. Even if they can intersect in certain areas, they are really two different things.

3.4. Emission reduction

The most prominent verb in this category is 'reduce', most frequently associated with 'emissions'. Most respondents also specified the sectors in which these reductions should take place, such as road traffic, industry, air transportation, and, less frequently, shipping and coal. The topic is thus less specific than the other topics, in that it calls for emission reductions in general as the primary response to climate change. To the extent that the topic points out a specific direction, it is the emphasis on mitigation as the main strategy to cope with climate change, rather than adaptation or geo-engineering, and perhaps more academically, the emphasis on reducing emissions rather than absorption by sinks.

Three most representative answers:

i.) Reduction of emissions in the transportation sector. Cleaning up emissions in the air from production. Tax "dangerous" emissions.

ii.) Reduce harmful emissions, both through different/reduced consumption and through technical improvements.

iii.) Reduce harmful emissions of gases. Adopt a preventive attitude when it comes to harmful emissions.

3.5. International collaboration and responsibilities

This topic comprises many sub-topics, such as the responsibilities of rich countries and the need for big polluters to act, but the most prominent one is that of Norway's role in the world, presented as a small country whose mitigation measures have very little global impact.

Three most representative answers:

i.) Norway is so small, the big environmental villains like China and other countries should implement measures at once. Norway can contribute by spreading knowledge and information.

ii.) Information, raising awareness in schools, clubs, organizations, adult education. Present concrete suggestions that everybody can do at home, such as the [...] information magazine on recycling, second hand shops, composting etc. Link the local and the global, and emphasize that every little thing helps, and that each individual has a major responsibility for the future.

iii.) We need an internationally binding agreement, where most countries take part. We need to be pragmatic and tolerate that some countries opt out, as long as they do not represent any significant part of current or future emissions. We can arrive at such an agreement if all necessary parties find it to be fair. Then the long-term goal should be a common limit for CO2 emissions per capita. Those countries that are below the limit must be allowed to increase their emissions. Those who are above must lower theirs. But to ensure that the most developed countries (Western and a few Asian countries) do not feel threatened, emissions need to be lowered proportionally, so that no one can wait for others to meet their commitments before they begin [. ].

3.6. Changes in consumption and lifestyle

Entries in this category tend to enumerate a large array of personal behaviors pertaining to climate change, such as consumption, transportation and travel, waste disposal and energy use. Food is a frequent element in this topic. While most respondents focus on the responsibilities of individuals, some also mention the need for legislation and government incentives to spur such changes.

Three most representative answers:

i.) Less driving fewer long flights vacation grow more organic food.

ii.) There has to be an end to non-recyclable plastic packaging. Less use of pesticides in agriculture and more support for those who want to grow organically.

iii.) Among other things, we should recycle even more. Eat less meat, and more vegetables. Choose local and ecologically produced goods. We should also use more public transportation, and rather invest in hybrid cars, or walk/bike shorter distances.

3.7. Governmental measures/instruments

Among those who focus on the role of government, the most popular instruments are taxes, economic incentives and regulations. The three most representative answers are:

i.) It is not possible to make actors choose climate-friendly alternatives unless it also makes sense economically. That is why it is necessary to implement political tools that make climate-friendly solutions economically friendly solutions. The costs of environmental costs should be placed with those who cause them, i.e. taxes etc.

ii.) The pressure to move from discussions and resolutions to concrete actions must be intensified in all relevant local, national, regional and international forums. At the same time there needs to be a continuous effort to raise awareness regarding what each person can/should/must contribute. A sustainable "green" development must be stimulated with "all" means, not least through economic and motivational incentives.

iii.) Follow up paragraph 112 of the Constitution in practice, i.e. move the burden of proof so that sustainable solutions are mandatory unless there are special reasons not to choose them, not like now, where you choose traditionally (and what you think is the cheapest solution in the short term), instead of robust long-term solutions — which are also required to develop the market and industry so that they are capable of coming up with real and good sustainable solutions. There is also a need for policies that make us not go for "low-hanging fruits" (i.e. fairly simple solutions), which block the way to solutions which might yield energy savings of up to 5-8%. I think old men (at least some of them) must be ejected from the decision-making process [...].

4. Linguistic analysis of responses

STM is a multi-membership approach, which means that it allows for a blend of multiple topics in any one response. However, the model does not take into account the multiple ways in which individual respondents may link these topics to each other or to individual or societal contexts. Thus, many phenomena can be detected only by a qualitative linguistic analysis. Furthermore, in addition to topic variation, answers to open-ended survey questions can be expected to contain various linguistic properties that may reflect variations in respondents' engagement with the issue. Does the respondent accept the premise of the survey question? Does the answer contain complete sentences, and if so, do these sentences form a deliberative text consisting of elements such as arguments, counter-arguments and conclusions? In order to explore these questions, and get a more complete picture of the answers with the diversity of opinions and attitudes that are integrated in them, an in-depth linguistic study is needed, taking into account both syntactic and semantic perspectives. For illustrative purposes, we will give only a few examples here by pointing at some typical forms that these answers can take and propose three main categories:

4.1. Rejection of the question

This category can take the form either of a simple "Nothing" or with a more elaborate statement, such as:

"There has always been climate change on our planet, warm periods and cold periods come and go. Experts preach about one threat after the other, but nothing is certain. [...]".

The second example is in fact not an answer to the question. It is a rejection of the presupposition embedded in the question: when we ask "what should be done", we implicitly claim that there is a problem, taken for granted, and that humans can (should) do something about it (Fl0ttum and Gjerstad, 2013; Kerbrat-Orecchioni, 2002). For respondents who deny the existence of human influence on climate change, the question is invalid. These answers belong to topic 3 (Attribution), which is characterized by complete "stories" including deliberation and personal opinions about the "nonsense" of the climate change issue. But we also find more nuanced and less categorical answers, where skepticism turns into ambivalence and hesitation, as in the following:

"All the years I have lived and remember have been different [...] so I tend to think most things are the way they have always been [...] But of course we must pay attention to what is happening and do all we can in order to avoid climate change."

This is consistent with previous findings of ambivalence in textual survey responses discussing attribution of climate change (Tvinnereim and Fl0ttum, 2015).

4.2. Enumeration of several measures

Many answers consist of sentence fragments (with or without a verb indicating the action needed). This form is frequent in topics 1 (transportation), 2 (energy transition), 4 (emission reduction), and 6 (changes in consumption and lifestyle). The following represents a typical example from topic 4: "Reduce emissions of climate gases, for example reduce use of cars", and from topic 1: "Free public transportation in large cities. Facilitate parking on the outskirts and develop public transportation." We find the same in topics 2 (energy transition) and 6 (changes in consumption and lifestyle), but answers in topic 6 also have a relatively high frequency of deontic constructions, the explicit expression of obligation ('vi b0r' — 'we should'):

"Among other things we should recycle more. Eat less meat, and more fruit and vegetables. Choose products made locally and ecologically. We should also use public transportation more, and rather invest in hybrid cars, or walk/bicycle at shorter distances."

Interestingly, respondents within topic 6 impose such obligations on various actors, e.g. "people", "we" (referring to both the country and to all citizens as individual actors), "everyone in Norway", "authorities". This reflects the relevance of both political measures and personal behavior regarding the topic of consumption. More generally, we find impersonal introductions expressing that something should be done, but not by whom, as in the following: "it is important/necessary to/ that" and "it is wrong to/that". However, in some cases, we find "one should", "all should", "everybody should", expressions indicating an opinion on the necessity of individuals to engage in actions related to climate solutions.

In topic seven (Governmental measures) we find some examples where respondents clearly point to politicians, through the wording "the state must ...". This use of deontic constructions is to a certain extent also found in topic 2 (Energy transition), where responsibility centers on society or the state:

"Stop investments and new projects for extracting fossil energy, i.e. gas, oil, and coal. Stimulate more investments in renewable energy, i.e. solar, wind, tidal, hydroelectric, bio-energy and energy efficiency. Because of the vast energy needs of the world, I think that we cannot avoid nuclear power. Therefore, more emphasis should be given to developing safe and environment-friendly technology in this area. "

This example could also be classified in the third form that we find, the answers constituting complete stories or narratives.

Global Environmental Change 46 (2017) 34-41 4.3. Narrative — full story representation

This form often includes deliberation related to both the complication (typically climate change) and solution stages of the climate change narrative (see Adam, 2008; Fl0ttum and Gjerstad, 2016). It is frequent in topic 3 (Attribution), as already illustrated above, as well as in topics 5 (International collaboration) and 7 (Governmental measures/instruments). In topic 5, several of the respondents tend to excuse 'little Norway' in addition to calling for international collaboration and blaming the big country emitters:

"Agreements with the countries that are the biggest polluters should be concluded, such as China, Russia and the US, in order to lower the limits of emissions. Norway is also a polluter, but to a small extent in a global perspective. Norway has a responsibility concerning the amount of extracted oil in addition to other oil extracting countries."

The respondent deliberates, or argues his/her point, by presenting a counter-argument to the view that Norwegian emission reductions should be a priority. In other words, the respondent enters into a discussion regarding possible solutions. In that sense, such answers are richer than those in the previous category (4.2), providing a more complete picture of what people are thinking than in the pure listing of measures to be undertaken.

Such deliberative structures are also found in topic 7; the following example combines this with a deontic perspective at the end ('Dette ma skje' — 'This must be done'):

"Change governmental policies as quickly as possible so that stronger economic instruments are used in order to make it clearly profitable for private persons and businesses to engage in and choose environment-friendly alternatives. This can mean removing taxes/charges on environment-friendly and sustainable products and services while increasing taxes on polluting/environmentally harmful products and services, in accordance with the polluter pays principle. This must be done simultaneously with a development of clearer environmental labeling."

A small number of answers in categories 6 (changes in consumption and lifestyle) and 2 (energy transition) are also argumentative, or deliberative, in their structure. Here is an example with a high topic 6 proportion:

"Buy less food, because, technically, we don't need as much food as many households are buying today for private use. It will save energy used in food production. In addition, we will waste less food both as private persons and by food stores."

Even though the examples listed above do not provide a representative selection of the data we are discussing in this paper, they demonstrate some of the advantages of using open-ended survey questions on issues as multifaceted as solutions of climate change. By exploring the structure and content of the answers, we see a substantial variation in their syntactic completeness as well as degree of deliberation, reflecting different levels of engagement with the issue. However, in the survey question we ask only for solutions and not for (responsible) actors. Thus, it is not surprising that the large majority of the answers consist of lists of words, naming various measures and instruments as potential solutions to tackle climate change, without making explicit who should do what, even though some actors are explicitly mentioned through deontic constructions (on responsibilities and obligation, see Fl0ttum, 2017).

5. What explains variation in topic emphasis across respondents?

Fig. 1 displays the results of regression analyses of topic proportions over the four main explanatory variables used in our topic induction procedure. One linear regression model was specified for each of the

Fig. 1. Effects of background variables on topic prevalence of individual responses.

The panel shows point estimates and confidence intervals of the effects of background variables on topic prevalence, holding all other variables constant. The two top figures show the effects of the binary variables gender and education (1 = university/college; 0 = no university/college). The two bottom figures show the effects of the ordered categorical variables age group and political interest for selected topics. Confidence intervals show the 95 percent bounds, with uncertainty estimates including both statistical uncertainty and uncertainty from the model induction process. N = 4372.

seven topics (indexed by topic number k):

Topic_Proportionk ~ Age + Gender + Education + Political_Interest + Constantk + errork

where Education was transformed into a binary variable indicating whether the respondent had university/college education or not. Political interest was measured in the survey using the question "In general, how interested are you in politics?" and with a five-point Likert scale ranging from 'Not interested at all' to "Very interested."

As shown in Fig. 1, women are more likely than men to bring up the topics of transportation and consumption, whereas men are more likely to emphasize energy transitions, government measures, and attribution of climate change. The higher propensity of men to bring up discussions of the causes of climate change — thus questioning the implicit need for action — agrees with previous findings of a gender gap in environmental risk perceptions (Kahan et al., 2007) and concern (McCright, 2010). Furthermore, the gender tendency seen in the data, especially as regards questions of technology (energy) and individual responsibility (consumption) align with previously identified qualitative work on masculine and feminine marked discourses (Henwood et al., 2014). We thus find congruence between previously found gender differences in climate change perceptions and our current text data on the climate solutions spontaneously suggested by men compared to women. As regards education, respondents with shorter time spent in the schooling system tend to write more about consumption and lifestyle choices, whereas respondents with higher education emphasize energy transitions and government measures. As climate change is a complex

phenomenon, and the discussed solutions are diverse, the highly educated are likely to have more information about the technical and societal measures available, whereas individuals with lower levels of education may more easily think of measures that are closer to their everyday life. Furthermore, the fact that the prevalence of the attribution topic correlates negatively with education level agrees with earlier findings in the literature that acceptance of mainstream climate science is greater among individuals with higher education.

Variation in political interest has a similar effect on topic prevalence. Specifically, individuals with greater self-reported political interest are more likely to provide responses drawn from the energy transition and government regulation topics, whereas those with less political interest emphasize the topic of consumption and lifestyle more. This may seem like an obvious point, given that political interest is higher among respondents with university education, yet it could have been otherwise as individuals with strong opinions on environmental issues might also be expected to emphasize individual responsibilities. Such an effect is not, however, captured by our overall data analysis.

Finally, as regards age, the clearest variation is seen on the topics of energy transitions, consumption/lifestyle, and the more generic 'emission reduction' topic. Younger respondents write more about energy and consumption than older respondents, a difference that is statistically significant. By contrast, older respondents are much more likely to respond that emission reductions in general are needed, without further detail. This suggests that the greater overall knowledge about climate change generally found among the young (McCright, 2010) also extends to knowledge about climate change solutions.

6. The near-absence of adaptation

As noted above, the second of the three main IPCC working groups is entitled "Impacts, Adaptation, and Vulnerability," besides the working groups on the physical science basis (no. 1) and mitigation (no. 3). Consequently, one might expect that adaptation would be one of the key topics offered by respondents as something that "should be done" about climate change. However, among all the 4614 textual responses across the two waves, the word "adaptation" and its various forms ("tiZpas*") was used in only 28, or by 0.6 percent of the respondents. It is possible that other words express adaptation ideas, and that the share talking about adaptation is somewhat higher, although our readings of the full material does not suggest a high number of such responses.

Typical adaptation statements include: "To reverse climate change will probably not be possible for a long time, so in the mean time we need to adapt — where we build/live/work — where we build roads etc. — as regards extreme weather and landslide risk etc."; "... it affects plants and animals that will have difficulties adapting."; and furthermore "We must adapt. We must accept that climate change is natural. ."

At the same time, 11 of the 28 responses using the word "adapt*" used it in a meaning that does not fit the traditional meaning of climate change adaptation, that is, making changes to buildings, infrastructure, agriculture etc. in the face of physical climate changes. Rather, these respondents wrote about adaptation of society and human behavior to a world where greenhouse gas emissions need to be much lower. Examples are:

"Customers need to adapt to there being fewer flights"; "Adapt taxes

in society such that it costs to pollute ..."; and "... Better adaptation

for electric cars, investigate better batteries."

Thus, this fairly prevalent interpretation of "adaptation" implies adaptation not to physical climate change, but to human responses to climate change. One potential reason for this lack of attention to climate change adaptation could be that many of the climate-related changes to society discussed in Norway concern issues such as the decarbonization of the transportation sector and the substitution of renewable energy for fossil oil and gas. The latter issue furthermore affects large proportions of the labor force, as the oil and gas industry is a major employer in Norway. Furthermore, the lack of interest in adaptation in the sense of reducing vulnerabilities may derive from a sense that climate change is a remote issue that will not affect respondents personally.

7. International responsibility and flexibility mechanisms

Looking at specific countries, among the major emitters China is mentioned by most respondents (94), followed by the US (52). Thereafter follow India and Russia (both mentioned by 23). Only five mentions of Germany and Brazil are seen. The EU as an entity is brought up by nine respondents, but 38 use the term "Europe" in one form or another. The low frequency of "Europe*" is somewhat surprising, as Norway's climate policy is intimately connected to that of the EU, via the Emission Trading Scheme (ETS), renewable energy obligations and numerous other regulations.

Regarding the latter topic, our data furnish new insight into Norwegians' views on large-scale climate policies such as the ETS and international flexible mechanisms, as pioneered in the Kyoto Protocol. Specifically, our survey responses include 49 instances of the word 'carbon credit' or "allowance" ("kvote"), which signifies a unit permitting industrial users to emit one ton of CO2 equivalent. It is also used in popular parlance to mean a certificate representing the reduction of one ton of emissions under frameworks such as the Kyoto Protocol's Clean Development Mechanism (CDM). Previous studies using Norwegian Citizen Panel data have shown that such flexibility produces negative associations (Tvinnereim and Steinshamn, 2016), most likely since

respondents are frequently skeptical of paying others to reduce emissions instead of reducing one's own. At the same time, individual carbon allowances have been proposed as a way to enhance transparency and responsibility for climate change mitigation.

A close reading of the 49 responses that contain the word "carbon credit" shows that 33 of the responses are mostly negative, 15 are positive, and one is unclear. The negative responses associate carbon credits with cheating and hypocrisy, stating that "Norway also has to contribute, not just buy carbon credits" and that flexibility mechanisms imply that "the rich buy from the poor" instead of cleaning up their own act. One respondent also sees carbon credits as "fictitious." Among the positive views, respondents suggest individual permits to fly, buy meat, and consume fuel — bordering on a rationing system. These are often associated with calls for less consumption and more local production, as seen in our consumption/lifestyle topic. Finally, two respondents emphasize the efficiency advantages of a global, standardized system of emission permits. Thus, while respondents appear mostly negative towards the idea of paying others to reduce emissions instead of reducing one's own, the idea of quantitative limits on emissions is viewed in a generally positive light.

8. Discussion and conclusion

A key finding from our research is the fact that mitigation completely dominates the discussion of climate change solutions. Among our seven induced topics, none explicitly covers adaptation; all others at least touch on mitigation. This is in line with earlier studies showing that people generally consider climate change as a remote problem, at least in rich countries. Furthermore, despite numerous spectacular flooding and other potentially climate-related events, Norwegians appear to think of climate change as a problem they need to address by its fundamental causes, rather than finding ways to live with it.

We also find that most responses support mitigation rather than questioning it. The most prominent dissident voices are found among those who doubt that climate change is a serious problem, or even real (mostly found in the Attribution topic), and those who think that Norway is too small to make a difference globally (some of the responses found in the International topic).

Generally, however, most respondents suggest mitigation as the preferred course of action. We see a prevalence of suggestions for government action over individual action, which is in line with previous research (Rathzel and Uzzell, 2009; Lorenzoni and Pidgeon, 2006). These suggestions are mainly found in the Energy, Governmental measures, and to some extent Transportation topics. However, many respondents also underline the importance of individual action, notably in the Consumption/Lifestyle and Transportation topic. We also find a clear strain of suggestions of combined government and individual action, in the numerous suggestions that government facilitate individual action — notably in the Transportation topic.

Our findings imply that policymakers seeking to enhance current mitigation policies will largely encounter a public that is ready to accept that mitigation is "what should be done about climate change," at least in Norway. Furthermore, the prevalence of responses related to government or large-scale mitigation suggest that any campaigns to change individual behavior in isolation from government action will be more challenging.

The open-ended design used in this paper makes two methodological advances to the study of public acceptance of and support for environmental actions. First, we solve the problem of which policies to evaluate. Climate change is a contested, complex problem with no single set of solutions. Rather, there is an ongoing debate also among experts about what actions are the most effective and efficient (Ostrom, 2010; Bertram et al., 2015). Using closed questions with rating scales thus inevitably means limiting policy options according to the researcher's own discretion, leaving out measures that some may think of as essential. The ideal of comprehensiveness in closed questions survey

(Krosnick, 1999) is thus particularly difficult to achieve in the field of climate change. Furthermore, climate actions may be found on numerous levels from the individual to the global, involving many types of actions, framed by various ideologies (from efficiency to no-growth) and with different goals (mitigation, adaptation, geo-engineering). Asking an open-ended question goes some way to solve this problem, as no potential policies, including controversial ones, are left out.

Second, our open-ended design addresses the problems of social desirability bias. In general, real support for environmental policies and measures can be hard to measure, as respondents have a tendency to approve of what the interviewer suggests because they feel that agreement is expected from them (Krosnick, 1999). By not suggesting any policies in particular, we refrain from suggesting any particular course of action, besides implying that something should probably be done about climate change.


The research was funded by grants from the Bergen Research Foundation (grant no. BFS2015DIG) and the Research Council of Norway under the projects Lingclim (grant no. 220654) and JPI European Perceptions of Climate Change (grant no. 244904). We would like to thank Lise Bjânes0y, Ottar Hellevik, and the participants of the Norwegian Citizen Panel Workshops of June 2016 and December 2015 for help and feedback at various stages of the manuscript.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at

Data deposition: Replication data and code are available at the Harvard Dataverse ( The full data set is available at the Norwegian Centre for Research Data.


Adam, J.-M., 2008. La linguistique textuelle. Introduction à l'analyse textuelles des discours. Armand Colin, Paris. Aldy, J.E., Kotchen, M.J., Leiserowitz, A.A., 2012. Willingness to pay and political support

for a US national clean energy standard. Nat. Clim. Change 2, 596-599. Bohm, G., Pfister, H.-R., 2001. Mental representation of global environmental risks. Res.

Soc. Proble. Public Policy 9, 1-30. Bertram, C., Luderer, G., Pietzcker, R.C., et al., 2015. Complementing carbon prices with technology policies to keep climate targets within reach. Nat. Clim. Change 5, 235-239.

Capstick, S.B., Pidgeon, N.F., Corner, A.J., et al., 2016. Public understanding in Great

Britain of ocean acidification. Nat. Clim. Change 6, 763-767. Fl0ttum, K., 2017. Willingness of action. In: Fl0ttum, K. (Ed.), The role of language in the

climate change debate. Routledge, New York/London, pp. 113-129. Fl0ttum, K., Gjerstad, 0., 2013. Arguing for climate policy through the linguistic

construction of Narratives and voices: the case of the South-African green paper. National Climate Change Response Clim. Change 118 (2), 417-430. http://dx.doi. org/10.1007/s10584-012-0654-7.

Fl0ttum, K., Gjerstad, 0., 2016. Narratives in climate change discourse. Wiley Interdiscip. Rev. Clim. Change.

Gellein, M.L., Kipperberg, G., Risa, A.V., 2015. Preferanser for norske utslippsreduksjoner versus klimatilpasningstiltak. Samfunns0konomen 3, 18-32.

Henwood, K., Pidgeon, N., Parkhill, K., 2014. Explaining the 'gender-risk effect'in risk perception research: a qualitative secondary analysis study. Psyecology 5, 167-213.

Hulme, M., 2009. Why We Disagree About Climate Change: Understanding Controversy, Inaction and Opportunity. Cambridge University Press.

Kahan, D.M., Braman, D., Gastil, J., et al., 2007. Culture and identity-protective cognition: explaining the white-male effect in risk perception. J. Empir. Legal Stud. 4, 465-505.

Kerbrat-Orecchioni, C., 2002. Présupposé, présupposition. In: Charaudeau, P., Maingueneau, D. (Eds.), Dictionnaire d'analyse du discours. Seuil, Paris, pp. 467-469.

Kotchen, M.J., Boyle, K.J., Leiserowitz, A.A., 2013. Willingness-to-pay and policy-instrument choice for climate-change policy in the United States. Energy Policy 55, 617-625.

Krosnick, J.A., MacInnis, B., 2013. Does the American public support legislation to reduce greenhouse gas emissions? Daedalus 142, 26-39.

Krosnick, 1999. Survey research. Annu. Rev. Psychol. 50, 537-567.

Leiserowitz, A., 2005. American risk perceptions: is climate change dangerous? Risk Anal. 25, 1433-1442.

Lorenzoni, I., Pidgeon, N., 2006. Public views on climate change: European and USA perspectives. Clim. Change 77, 73-95.

Lorenzoni, I., Leiserowitz, A., de Franca Doria, M., et al., 2006. Cross-national comparisons of image associations with global warming and climate change among lay-people in the United States of America and Great Britain. J. Risk Res. 9, 265-281.

McCright, A.M., 2010. The effects of gender on climate change knowledge and concern in the American public. Popul. Environ. 32, 66-87.

Morton, T.A., Rabinovich, A., Marshall, D., Bretschneider, P., 2011. The future that may (or may not) come: how framing changes responses to uncertainty in climate change communications. Global Environ. Change 21 (1), 103-109.

Ostrom, E., 2010. Polycentric systems for coping with collective action and global environmental change. Global Environ. Change 20, 550-557.

Rathzel, N., Uzzell, D., 2009. Changing relations in global environmental change. Global Environ. Change 19, 326-335.

Roberts, M.E., Stewart, B.M., Tingley, D., et al., 2014. Structural topic models for open-ended survey responses. Am. J. Pol. Sci. 1064-1082.

Roberts, M.E., Stewart, B.M., Tingley, D., 2015. Navigating the local modes of big data: the case of topic models. In: Alvarez, R.M. (Ed.), Data Science for Politics, Policy and Government. Cambridge University Press.

Shwom, R., Bidwell, D., Dan, A., et al., 2010. Understanding US public support for domestic climate change policies. Global Environ. Change 20, 472-482.

Skjervheim, 0., H0gest0l, A., 2015. Norwegian Citizen Panel 2015: Fifth Wave Methodology Report.

Smith, N., Leiserowitz, A., 2014. The role of emotion in global warming policy support and opposition. Risk Anal. 34, 937-948.

Tvinnereim, E., Fl0ttum, K., 2015. Explaining topic prevalence in open-ended survey questions on climate change. Nat. Clim. Change 5, 744-747.

Tvinnereim, E., Ivarsflaten, E., 2016. Fossil fuels, employment, and support for climate policies. Energy Policy 96, 364-371.

Tvinnereim, E., Steinshamn, S.I., 2016. Folkelig aksept for klima- og energitiltak i Norge. Samfunns0konomen 130, 77-86.

Whitmarsh, L., 2008. What's in a name? Commonalities and differences in public understanding of 'climate' change and 'global warming'. Public Underst. Sci.

Wolf, J., Moser, S.C., 2011. Individual understandings, perceptions, and engagement with climate change: insights from in-depth studies across the world. Wiley Interdiscip. Rev. Clim. Change 2, 547-569.