Renewable Energy · Strategy · XandY Insights

The most effective ways to persuade people to make sustainable choices

A meta-analysis of 430 field studies shows which interventions change sustainable behavior, which behaviors are most movable, and why the easiest wins are rarely the ones that matter most.

Sustainability issues are among the most pressing challenges we face. They require urgent and effective action to mitigate devastating effects. One of the key challenges is finding effective ways to encourage people to act in more sustainable ways. Some possible interventions that have been suggested include financial incentives for sustainable behavior, social pressure, and public education campaigns.

However, not all interventions are equally effective, and some behaviors may be more difficult to change than others. This highlights the importance of understanding which interventions work best for increasing pro-climate behavior, which behaviors are most movable, and how we can make strategic trade-offs between different intervention types and behaviors. By gaining a better understanding of these factors, we can develop more effective strategies for combating climate change and promoting a sustainable future.

The many ways to increase sustainable behaviors

I recently collaborated on a huge meta-analysis (an analysis of other analyses combined) which compiled the findings from 430 different scientific studies that each tested different ways to encourage people to act more sustainably. These hundreds of "interventions" were diverse. They included education on how to act more sustainably, financial incentives for doing so, feedback about one's past behavior, basic reminders, getting people to set goals to act more sustainably, and communicating social norms. Overall, we found that many of these interventions are effective. Across all 430 studies, the average intervention raised the targeted behavior by roughly 12 percentage points, or about 7 points after correcting for publication bias.

However, that diversity in strategies also came with quite a lot of variation in effectiveness. Our meta-analysis showed that, across those 430 studies, the most effective interventions were those that leveraged social norms. That is, the ones that persuaded people to act sustainably by highlighting the sustainable behaviors or expectations of other people around them.

Which kinds of interventions change sustainable behavior the most? Average effect of six intervention types on climate-relevant behavior in field settings, before publication-bias adjustment 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Effect size (Cohen's d) Social comparison norms, modeling, what others do 0.37 Financial incentives rewards, rebates, pricing 0.32 Appeals values and responsibility; one meta-analysis, no CI 0.28 Commitment goal setting, pledges, implementation intentions 0.27 Feedback information about one's own past behavior 0.16 Education flyers, tips, labels, facts 0.09 BERGQUIST, THIEL, GOLDBERG & VAN DER LINDEN (2023), PNAS · 10 META-ANALYSES, 430 FIELD STUDIES · 95% CIs
Figure 1. Second-order meta-analytic effect sizes by intervention type. Social comparison (norms) and financial incentives lead; education and feedback on their own trail far behind. For scale, an effect of d = 0.31 is about a 12-percentage-point increase in the behavior. Adjusting for publication bias lowers all of these numbers but does not change the order much.

For example, in a classic study that aimed to increase hotel towel reuse, researchers used signs that emphasized messages like "75% of the guests who stayed in this room (#123) participated in our new resource savings program by using their towels more than once. You can join your fellow guests in this program to help save the environment by reusing your towels during your stay." There are now hundreds of different examples of how social norms can be used to change people's behavior.

Another encouraging finding is that the evidence across 430 studies showed that many different types of behaviors are movable. There were interventions that were successful in increasing recycling and use of sustainable transportation, and many others that were successful in decreasing littering, water or electricity use, or food waste. Of all these behaviors, the one most influenced by interventions was littering, whereas transportation behavior was most difficult to change.

So, all of this considered, the big question is: how do we translate these insights into mass-scale change? It is not sufficient to just replicate the procedures of previous interventions without careful attention to the context, uncertainties, and trade-offs. For this, we need strategy.

Making strategic trade-offs when choosing interventions

Strategy is needed to make decisions under uncertainty and to decide how to invest resources. As we explained in another post, we need to manage the Reach × Effect Tradeoff. This is the tendency for high-impact interventions to be difficult to execute on a mass scale, while mass-scale interventions tend to have a small impact per person reached. The meta-analysis found the same pattern: in the subset of very large studies (more than 9,000 participants each), the average effect shrank to about 2 percentage points.

There is an additional layer to this puzzle when we think about reducing climate change: some behaviors are easier to influence than others, and behaviors differ widely in how much they affect the climate. For example, we found that littering and recycling were much easier to influence than transportation behavior, but transportation behavior has much larger climate impacts than littering or recycling. Thus, the importance of changing transportation behavior may be well worth the necessary additional effort.

The behaviors that are easiest to move matter least for the climate Five behavior categories plotted by how movable they are (measured) against their climate impact (illustrative) Hard to move, high impact Easy to move, low impact 0.0 0.2 0.4 0.6 How movable the behavior is: effect size of interventions (Cohen's d, measured) Lower Higher Climate impact of the behavior (illustrative ranking) Transportation d = 0.08 Consumption (food, products) d = 0.20 Conservation (energy, water) d = 0.25 Recycling d = 0.27 Littering d = 0.52 EFFECT SIZES: BERGQUIST ET AL. (2023), PNAS, TABLE 1 · CLIMATE-IMPACT RANKING IS ILLUSTRATIVE, NOT MEASURED
Figure 2. Horizontal position is the measured average effect of interventions on each behavior category. Vertical position is an illustrative ranking of each category's climate impact, drawn from the literature the paper cites, not a measured quantity. Littering is easy to move and matters little; transportation is hard to move and matters most. The strategic question is how much extra effort the upper-left corner is worth.
Interventions built on social comparison moved behavior four times as much as interventions built on education alone (d = 0.37 vs. 0.09).

So what?

The considerations raised above are not just relevant to climate change and the environment. This serves as an excellent case study for understanding which interventions we have at our disposal, which outcomes we care about, and how we make trade-offs between the different interventions and outcomes. Similar approaches can fit a wide range of issues, but especially those that are mired in uncertainty, confusion, or polarization.

Cite this. Goldberg, M. H. (2023). The most effective ways to persuade people to make sustainable choices. XandY Analytics. Retrieved from xandyanalytics.com/most-effective-ways-to-persuade-sustainability

Underlying study. Bergquist, M., Thiel, M., Goldberg, M. H., & van der Linden, S. (2023). Field interventions for climate change mitigation behaviors: A second-order meta-analysis. Proceedings of the National Academy of Sciences, 120(13), e2214851120. doi.org/10.1073/pnas.2214851120. Preregistered; 10 meta-analyses covering 430 unique primary field studies. Overall effect d = 0.31 (95% CI 0.30–0.32), reduced to d = 0.18 after publication-bias adjustment and to d = 0.04 in a subsample of 32 large-scale studies. Data and code: osf.io/6dyq9.