Climate change is probably the most important policy issue of our time. The set of expected outcomes range from social disruptions at a global scale to an existential threat for humankind. Since it is a consequence of human activity, both the problem and its response feature behavioural components, from the low salience of carbon emissions to the design of policies, accounting not only for self-interest, but also for emotions, social norms and values. With a global scale and scope, meaningful action on climate change requires a trans-disciplinary approach, behavioural science enmeshing itself with other sciences, and also in the very design of the discussion and decision processing.
Unfortunately, behavioural science is late coming in this endeavour. So far, climate policy has been mostly driven by climate science models with little to no behavioural component, and behavioural public policy has been convoked only as a last-step solution to entice piecewise less environmentally harmful behaviours, or to understand and appease reactance to the policies prescribed by climate science. Hopefully, this is evolving: the 2024 call for tender by the DITP, which purported to cover all behavioural public policy consulting services for France central administrations, used preparedness and adaptation to storm surges as the main exercise case in the submission process.
This chapter aims to provide you with an overview of the topics and issues of behavioural climate change policies. Rather than a selection of interventions, my directing thread is an analysis of climate change policy as a behavioural problem — Behaviour as a Lens, as in (Hallsworth 2023aHallsworth, Michael. 2023a. A Manifesto for Applying Behavioural Science. The Behavioural Insights Team. https://www.bi.team/publications/a-manifesto-for-applying-behavioral-science/.) — to assess how well we are currently responding to this challenge.
While reading this chapter, you should pay close attention on how the behavioural concepts introduced in previous chapters are used: I constructed this text as much as a case study as a traditional lecture.
Before I start an analysis which owes much to (Linden et al. 2021Linden, Sander Van Der, Adam R. Pearson, and Leaf Van Boven. 2021. “Behavioural Climate Policy.” Behavioural Public Policy 5 (4): 430–38. https://doi.org/10.1017/bpp.2020.44.), let me state my starting point. In what follows I’ll take for granted that climate change is real, that it overwhelmingly stems from human activity, that expected consequences are dire49 I do not deny that some people and areas may benefit from climate change (e.g. ice-free shipping lanes in the Arctic). I contend that losses will vastly outweigh these gains, and that most of these gains, for example arable land in Siberia, will fail to materialize., and that time to act is short. Basically, I’ll take the IPCC reports50 Link to the latest versions at face value. I’ll also take three possibly more contentious stances:
This being said, let us proceed to a behavioural diagnosis of the challenges facing climate-relevant policies and behaviour.
The purpose of this section is to highlight the behavioural dimension of climate change policies. Human-caused climate change is by definition the result of human behaviours. The reviews I read are mainly focused on the problem of behavioural change, i.e. what prevent people from adopting less harmful behaviours. Comparatively little thought seems to be devoted to why they have environmentally harmful patterns in the first place.
A common approach to the behavioural challenge of climate change is to frame it as a procrastination problem. This approaches underlines that behavioural changes required to lower emissions entail immediate, observable costs, such as reduced consumption, higher costs of locally sourced items and services, foregone opportunities of work and leisure53 Fewer plane trips to fancy academic conferences in fashionable destinations, for example., and additional taxes to fund the adaptation of the existing material and social infrastructure to the expected consequences of climate change. Because a continuous increase in consumption and opportunities was largely taken for granted, loss aversion magnifies the subjective assessment of these costs.
On the other hand, the benefits of behaviours and policies limiting climate change and adapting to its consequences (climate policies for short) will take place in a future which is remote, both in terms of subjective horizon54 A feature which may be magnified by the ageing of richer societies: people may care less about what will likely happen after their death. and in terms of geography. People now bearing the brunt of the consequences of climate change live far away from those who contribute the most to it. Furthermore, these benefits lack salience, since a catastrophe averted is, by definition, not observable. This results in a strong incentive to procrastinate.
An additional factor to explain the lack of action lies in the complexity of designing climate policies. Apart from the largely unpopular carbon tax, there are few straightforward and efficient measures. At all scales, seemingly simple solutions turn out to have harmful unintended consequences. For example, the plastic bags ban has resulted in a proliferation of water-consuming cotton tote bags55 Evidence for this oft-cited claim is actually hard to come by. Most sources lack strong evidence of the substitution effect, and impact figures almost always quote the 2018 Danish study (The Danish Environmental Protection Agency 2018The Danish Environmental Protection Agency. 2018. Life Cycle Assessment of Grocery Carrier Bags. No. 1985. Ministry of Food and Environment of Denmark. https://www2.mst.dk/udgiv/publications/2018/02/978-87-93614-73-4.pdf.)., and the reliance of coal power in China makes the replacement of fuel furnaces by China-made solar panels a debatable change.56 The IEA says it still makes sense to use China-made PV panels, which offset their production emissions over 3-8 months of use, per (IEA 2022IEA. 2022. Special Report on Solar PV Global Supply Chains. International Energy Agency. https://www.iea.org/reports/solar-pv-global-supply-chains.). In our experiment on wood burning (Abdel Sater et al. 2024Abdel Sater, Rita, Mathieu Perona, Elise Huillery, and Coralie Chevallier. 2024. “The Power of Personalised Feedback: Evidence from an Indoor Air Quality Experiment.” Behavioural Public Policy, December 10, 1–38. https://doi.org/10.1017/bpp.2024.46.), we discovered that the CO2 emission benefits of wood burning for heating were conditional on forests management, equipment quality, and anyway came at a cost for human health, through particulate matter emission. There is an ongoing debate on the actual merits of electric cars, given their consumption of rare earths, the non-reparable design of some models57 Telsa famously welds batteries to the chassis, making them impossible to replace, and the energy mix at their point of production, and final disposal.
Behaviourally speaking, this situation is very similar to the more familiar situation of house buying: too much information, too many trade-offs, resulting in an information overload which favors either the status quo, or choices on irrelevant features, such as the public profile of the person advocating for a given policy.58 In France, I’d suggest to call that the Jancovici effect.
The combined effect of procrastination and information overload contribute to the widely observed intention-action gap in individual climate behaviours: even among people with a strong awareness of climate change and a professed willingness to act, behaviour changes are much more modest than what would be reasonably achievable. As recalled by (Dablander et al. 2025Dablander, Fabian, Florian Lange, Cameron Brick, and Adam R. Aron. 2025. “Expressing Intentions Is Not Climate Action.” Proceedings of the National Academy of Sciences 122 (28): e2512457122. https://doi.org/10.1073/pnas.2512457122.), meta-analyses show that intentions explain only 18 to 28% of the variance in actual behaviour, and too few studies measure actual behaviours.
At an individual level, climate action looks like a prisoners’ dilemma. As heartwarming at it may be, the fire-fighting hummingbird image illustrates that individual actions have a negligible impact. Any meaningful result must stem from widespread behavioural change, and there will inevitably be free-riders. Climate policies can thus be framed as a collective action problem.59 Wikipedia page A significant feature in this collective action approach is that individual behaviour changes are not only low-impact, they are also often low salience.
Garbage sorting is a prominent example: close to no-one sees you carefully sorting between glass, plastics and organic garbage, and it is uncommon to boast about it.60 You also need to trust that the sorted garbage is properly recycled or processed down the line, which is not the case for most plastics. Similarly, taking a train rather than a plane, or refraining from travelling to far-away destinations is not as easily observable as the other person’s Instagram posts about their surfing session in Australia. This lack of salience may lead to a misperception of the scale of actual behavioural change, putting brakes of the behavioural cascade which underlie society-wide changes (see again (Frank 2020Frank, Robert. 2020. Under the Influence. Princeton University Press. https://press.princeton.edu/books/hardcover/9780691193083/under-the-influence.) for more details on behavioural cascades).
On the payoff side of the collective action frame, the populations contributing the most to climate change are generally not among the most affected. This is of course true on a global scale, but also on a national one. In France, the homes repeatedly flooded in the Somme area were low- and middle-income dwellings. Paris is protected by a flood control system, including artificial lakes far upstream. There is thus a significant disalignment between whose behaviour needs to change the most and whose situation will benefit the most from that change.
Fairness concerns also loom large in the collective action approach. Some population with highly unsustainable behaviours, e. g. farmers in France, also face strong constraint on their livelihoods, being one bad crop away from bankruptcy, which encourages short-term thinking (see the chapter on scarcity mindset) and discourages change. The Yellow vests protests provided another example. Many protesters were painfully aware that their “countryside house with two cars” lifestyle is not even financially sustainable in a world of increasing energy prices. However, they felt they had no other option for decent housing and healthy environment for raising their children.61 Another example is the choice of diesel cars in France. For a long time, diesel motors have been pushed on consumers as more eco-friendly (they emit less CO2 than petrol engines) and cheaper. People understandably felt flouted when taxes of diesel were aligned with those on petrol, and diesel cars were the first banned out of cities due to their high PM emissions. These examples highlight the fact that policies efficient from an environmental and economic point of view (farming regulations, carbon tax) may need to compensate people for the consequences of their own debatable choices — which likely offends the sense of fairness and desert of other parts of the population.
This fairness motive is also a sticking point in global climate negotiations. The picture of responsibilities for climate change is not the same if you look at current emission at a country level or at an individual level, or if you take a long-term view, including past emissions since it is the stock of carbon in the atmosphere which matters.
The very fact of climate change is challenged by a vocal minority in most countries. Behaviourally speaking, we observe there many forms of motivated reasoning, from selective disregard of sources to overcritical screening of climate change proponents’ arguments. These instances of motivated reasoning take place in a public landscape where climate issues are increasingly politicized. They thus become part of in-group thinking, where people are influenced by what they think is the dominant opinion of their group (injunctive norms). These perceptions may be severely biased, for example when a minority within the group can credibly claim they express the norm. This further hinders between-groups cooperation through the representation of the other groups as more distant and adversarial than they really are.
Several studies have thus shown that in the US, both Republican and Democrats overestimate the span of the ideological divide between average registered voters of each party on key issues such as abortion, gun control or climate policies (Kleinfeld 2023Kleinfeld, Rachel. 2023. Polarization, Democracy, and Political Violence in the United States: What the Research Says. Carnegie Endowment for Internation Peace. https://carnegieendowment.org/research/2023/09/polarization-democracy-and-political-violence-in-the-united-states-what-the-research-says?lang=en.).
I the first version of this lecture, I followed my own framework for behavioural analysis, and started with the structural obstacles below before embarking on the section you just read. Indeed, there are structural obstacles hampering behavioural science results from entering the policy making process, and if I do not deem them insurmountable, they represent a significant challenge. However, students told me that starting by the headwinds felt too discouraging, hence the inversion between the behavioural and structural section of the analysis.
To date, most models of climate change make only crude assumption on how human behaviours respond to changing climate conditions, and macroeconomic models are only beginning to include environmental constraints and dynamics.63 See for example the burgeoning literature on environmental-dynamic stochastic equilibrium (E-DGSE) models. Part from this stems from academic incentives: cross-discipline endeavours are harder to publish in field-specific journals, and thus represent riskier career bets. But most of the difficulty lies in the fact that the results of behavioural science do not interface readily with the kind of models used in climate sciences. While we have robust results on specific behaviours and specific environments, we are far from a general theory of human behaviour whose relevant features could be abstracted in quantitative relations between e. g. average temperature and variance, and levels of consumption or consumption-derived emissions. Barring a general theory, our piecewise evidence does not reach that sector-specific level of generality. This makes our body of knowledge difficult to embed in wider climate models.
Policy-wise, the inclusion of behavioural science into policy design is hampered by doubts about the reliability, scientific and practical, of behavioural science. After a time a great enthusiasm for behavioural solutions, the reproductibility crisis has dampened spirits, casting doubts on the effect size which can be expected from behavioural interventions. The credibility of behavioural intervention also suffers from the typically short horizon of their evaluation. This leads to uncertainties about their long-term effects. Most studies use a narrow set of metrics used, which may oversee such phenomena as moral licencing, which leads to adverse spillover effects between behaviours.64 “I use my bike to commute, so I can make one more flight for my vacations.” Such backlash effects on other behaviours are rarely including in interventions’ assessments.
Another strand of criticism argues that behavioural intervention crowd out more meaningful structural or systemic change. The question there is less the absolute return on investment of behavioural intervention than the idea that they provide an excuse for not entering into more ambitious but politically costly changes. A related criticism is that they put the onus of responsibility on individuals, whose decision margin and impact may be slim, at the expense of governmental and corporate responsibility.
In this section, I cover some examples of how behavioural insights are leveraged to contribute to climate policies. In this domain, it is easy to loose sight of the forest for the trees. Each intervention raises new, narrow but legitimate questions, which tend to distract from the overarching one: how well do the current intervention match the above diagnosis?
It should also be noted that many papers test a wide range of interventions, for example (Sinclair et al. 2025Sinclair, Alyssa H., Danielle Cosme, Kirsten Lydic, et al. 2025. “Behavioral Interventions Motivate Action to Address Climate Change.” Proceedings of the National Academy of Sciences 122 (20): e2426768122. https://doi.org/10.1073/pnas.2426768122.), spanning several of the obstacles indentified earlier.
By and large, this is the most natural area of intervention for behavioural science. Indeed, all but one chapter in (OECD 2017OECD. 2017. Tackling Environmental Problems with the Help of Behavioural Insights. Organisation for Economic Co-operation and Development. https://www.oecd-ilibrary.org/environment/tackling-environmental-problems-with-the-help-of-behavioural-insights_9789264273887-en.) refer to this type of interventions. The famous OPOWER experiment (Allcott 2011Allcott, Hunt. 2011. “Social Norms and Energy Conservation.” Journal of Public Economics, Special Issue: The Role of Firms in Tax Systems, vol. 95 (9): 1082–95. https://doi.org/10.1016/j.jpubeco.2011.03.003.) showed both promise and pitfalls in this area. One the one hand, low-cost feedback equipment can entail meaningful changes in consumption patterns. On the other hand, rebound and boomerang effects can negate the positive impact, and should be dealt with in the very design of the intervention, as we saw in the section on symbolic rewards (Schultz et al. 2007Schultz, P. Wesley, Jessica M. Nolan, Robert B. Cialdini, Noah J. Goldstein, and Vladas Griskevicius. 2007. “The Constructive, Destructive, and Reconstructive Power of Social Norms.” Psychological Science 18 (5): 429–34. https://doi.org/10.1111/j.1467-9280.2007.01917.x.).
Personalized feedback is effective under many conditions. It leads to reductions in consumption, sometimes modest, but which seem to be sustained at least as long as the feedback remains in force. However, this approach often lacks economies of scale: to provide feedback detailed enough to be convincing, you need a detailed and reliable measurement of the behaviour you want to influence, or its consequences. Electricity consumption was the proverbial low-hanging fruit: most homes in rich countries are now equipped with smart meters, which can provide high-frequency information about consumption levels, and even infer the kind of appliances used by their pattern of power use. It is much more costly and cumbersome to equip even voluntary households with shower meters, particulate matter meters (and we saw in our discussion of (Sater et al. 2024Sater, Rita Abdel, Mathieu Perona, Elise Huillery, and Coralie Chevallier. 2024. “The Power of Personalised Feedback: Evidence from an Indoor Air Quality Experiment.” Behavioural Public Policy, December 10, 1–38. https://doi.org/10.1017/bpp.2024.46.) how it proved impossible to distinguish PM emission form wood burning from emission due to other sources), mobility monitoring65 A few years ago, I had to turn off the autodetection of a cycling monitoring app. I wanted it to automatically log my daily bike commutes. I discovered it also logged bus trips, because the average speed and start/stop pattern of a bus in Paris looks like that of a bike., etc.
Even the rise of smart home devices is of little use, since there is no government or industry standard to date, and a high turnover of companies which mean a lot of legacy equipment. On a broader scale, the kind of infrastructure to get good measurement of only a reasonable selection of climate-relevant behaviour amounts to a worrying surveillance network from which a lot of personal data could be easily extracted.
Triggering changes of social norms is the second common strategy in this area. One key problem is that we are generally not in the case we saw with binge drinking, where the undesirable norm was the exception and the desirable norm the rule, and thus the intervention boiled down to correcting misperceptions. Again, energy use is a low-hanging fruit: because of the energy crisis in the 1970s and 1980s, many people in WEIRD countries grew up with the idea that conservation of energy was important.66 Of course, for most people in non-WEIRD country, it is a daily imperative.
The same does not hold for most other behaviours we may want to change: in most cases, the dominant norm is unsustainable. According to recent estimates, the average CO2-equivalent footprint in terms of greenhouse emission gases is an average 9.8t per person in France. This is five times the 2 tons estimate to reach the +1.5°C target of the Paris Agreement, and this carbon budget is currently eaten up by mobility-related emissions alone.67 Computations for the Secrétariat au Développement Durable. On the subject of carbon budget, see (Henriet and Schubert 2021Henriet, Fanny, and Katheline Schubert. 2021. La transition énergétique. Rue d’Ulm / Cepremap.)
As (Sparkman et al. 2021Sparkman, Gregg, Lauren Howe, and Greg Walton. 2021. “How Social Norms Are Often a Barrier to Addressing Climate Change but Can Be Part of the Solution.” Behavioural Public Policy 5 (4): 528–55. https://doi.org/10.1017/bpp.2020.42.) argue, the same holds for most climate-relevant consumer behaviours: mobility, but also meat consumption, digital devices turnover, meat consumption, or fertility. The difficulty of appealing to social norms in this area is compounded by two behavioural elements:
In this kind of situation, dynamic norms, highlighting a rate of change, may be more convincing than static ones. For example,
The share of students ordering a vegetarian main course at the University cafeteria doubled between 2021 and 2024.
probably gets more attention and traction than:
There are now 10% of students who order a vegetarian main course at the University cafeteria.
This fictional example illustrates both the power and the pitfall of a dynamic norm. Using rates, you may show attention-grabbing figures for the rate of change, giving the impression that a massive change is underway, and appealing to a kind of injuctive norm (other people of my group are changing behaviour fast, I’d better jump on the bandwagon). However, the strength of the effect largely relies on the base rate neglect fallacy of your target population, which has only a hazy notion of how prevalent the behaviour is. Of course, this is dangerously close to manipulation69 One could argue this is an example of Isaac Asimov’s “The closer to the truth, the better the lie, and the truth itself, when it can be used, is the best lie.”., and make your intervention vulnerable to an exposition of the actual base rate — including through casual observation “but I don’t see anyone ordering veg meals!”. For this reason, dynamic norms should probably reserved for cases where the sustainable norm is both at a sufficient baseline level, and evolving fast enough to publish a sufficient rate of change.
Other behavioural levers include:
Cognitively speaking, many decision-making processes are suboptimal. In cultural evolutionary terms, the gains of making a “good” decision on its own terms are often balanced against the gains of making a decision which agrees with the narrower group interests and beliefs, such skewing our deliberative skills towards the latter ((Mercier and Sperber 2019Mercier, Hugo, and Dan Sperber. 2019. The Enigma of Reason. Harvard University Press.), again). Furthermore, we tend to pay little attention to the amount of variability in judgement, both for the same individual at different points of time, of between individuals ((Kahneman et al. 2021Kahneman, Daniel, Olivier Sibony, and Cass R. Sunstein. 2021. Noise: a flaw in human judgment. William Collins.)). Thus, decision process design could greatly benefit form a behavioural approach, if only to move away from the information deficit - rational choice framework. As (Fischhoff 2021Fischhoff, Baruch. 2021. “Making Behavioral Science Integral to Climate Science and Action.” Behavioural Public Policy 5 (4): 439–53. https://doi.org/10.1017/bpp.2020.38.) notices, communicating scientific results to the wider public is a scientific area in itself. Behavioural science has been quite successful in doing it, with concepts like nudges becoming common currency, but much less so in contributing to the understanding of this endeavour. The authors also underline that there is a gap between the information scientists are able to provide and decision-maker needs, as well as actual choices people face (typically, information overload when trying to choice the less damaging consumption choice). They argue that behavioural science could also help by making explicit some key choice parameters (risk appetite, present preference, confidence in future technologies) which frame the debate.
The cognive overload issue is the main concern of (’Arvai and Gregory 2021’Arvai, Joseph, and Robin Gregory. 2021. “Beyond Choice Architecture: A Building Code for Structuring Climate Risk Management Decisions.” Behavioural Public Policy 5 (4): 556–75. https://doi.org/10.1017/bpp.2020.37.), who underline that to move beyond a patchwork of interventions on choice architecture (energy efficient defaults in buildings, meat-free offers in restaurants), we need to design decisions processes able to tackle more complex decisions than our lunch menu, for example when choosing an energy mix with a regards to CO2 emissions, but also to local industrial needs (can usage cope with price or generation volatility?), local employment, skills, use of other rare resources, and so on. Typically, the observed behaviour in front of this type of decision is tradeoff avoidance: decision-maker select an option which outperforms the others in a single, salient dimension, since it is easier to justify, among a lot of the kind of motivated reasoning we saw earlier. The authors’ approach is to help people making their objectives and values explicit, and link as much as possible values-based objectives (I care about local employment) with science-based insights (this technology fosters more high-quality local jobs). To this end, they build a sequential and transparent framework for decision-making:
A core challenge for behavioural climate policy is to leave a standalone position — with results easier to showcase on the academic market — for a deep embedding in climate policy in general. This implies a strong, often tough, dialogue with other disciplines but also civil servants and decision-makers. A dialogue in which we need to bring all the behavioural tools of our trade.
8.4.1 Social sciences arrived late in the field
The way we approach public policies issues often displays a strong path dependency. (Fischhoff 2021Fischhoff, Baruch. 2021. “Making Behavioral Science Integral to Climate Science and Action.” Behavioural Public Policy 5 (4): 439–53. https://doi.org/10.1017/bpp.2020.38.) argues that in the US, social sciences have been quickly excluded from initial governmental research groups on climate change, in the early 1980s. Economics joined the IPCC in 1993, and with Kenneth Arrow, its contribution focused on the discount rate to use ((Cherrier and Garcia Duarte 2024Cherrier, Beatrice, and Pedro Garcia Duarte. 2024. “How the "Ramsey Formula" Came to Define Time Discounting in Economics (1950-2000).” SSRN Scholarly Paper No. 4973044. Rochester, NY, Pre-published September 1. https://doi.org/10.2139/ssrn.4973044.)62 The authors show that this foray into climate models had a significant retro-action on macroeconomic methods: it stabilised the Ramsey formula as the canonical time discount approach., see (Charpentier 2024Charpentier, Arthur. 2024. “Discounting the Future?” Billet. Freakonometrics, December 5. https://doi.org/10.58079/12u55.) for a summary). The whole debate, which peaked in 2006 through the UK-commissioned Stern report (Stern 2007Stern, Nicholas. 2007. The Economics of Climate Change: The Stern Review. Cambridge University Press. https://doi.org/10.1017/CBO9780511817434.), rose to public prominence because of the dispute between its lead author and fellow economist William Nordhaus about the sensitivity of the conclusions to the discount rate used. Basically, Stern argued that IPCC models should use rational choice theory, while Nordhaus favoured rates derived from actual (much more narrow-sighted) behaviours. While economists saw it as a debate between normative (rational choice) and descriptive economics, behavioural insights — to begin with the fact that people may display different time preferences depending on the circumstances — were sorely absent.