This study’s methodology involves a literature search and a meta-analysis. The literature search involved selecting RCTs that tested the impact of nudges using qualitative filtration criteria. In the meta-analysis, the study outcomes were converted to SMDs, and the SMDs were pooled for relevant categories (such as the target sector and the intervention strategy) to obtain an estimate representing the aggregate impact.

Literature search

This study is a meta-analysis of primary studies that quantitatively measured behavioural changes due to nudge-based interventions. A meta-analysis is a systematic review that pools quantitative outcomes from individual primary studies to calculate a single numerical estimate summarising all the individual outcomes (Harrer et al. 2021). Used to summarise findings in the psychology literature (e.g., Bolier et al. 2013) and behavioural interventions (including nudges) literature in particular (e.g., Khanna et al. 2021, Mertens et al. 2021, and Nisa et al. 2019), a meta-analysis is a key tool for summarising the quantitative results of multiple studies.

We use the eligible study outcomes from two previous meta-analyses, Nisa et al. (2019) on reducing greenhouse gas emissions through nudges targeting indoor energy use, transport, and waste reduction or recycling and from Khanna et al. (2021) on testing non-monetary behavioural interventions to reduce household energy consumption. Additionally, we have considered published literature searched and identified from a bibliographic database (Web of Science) and unpublished grey literature from Google Scholar. We consider only RCTs containing error measurements: standard deviations, standard errors, and confidence intervals.

We used the following search string to perform the literature search: (gas OR wood OR engine OR food OR (public transport) OR energy OR electricity OR electronic OR bus* OR car* OR cook*) AND (social AND (comparison* OR norm*)) OR nudg* OR feedback* OR reduc* OR recycl* OR increas* OR wast* OR promot* OR conserv* OR sav* OR use OR consum* OR effect* OR inform* OR adopt* OR improv*) AND (evaluat* OR assignment OR evidence OR decision* OR interaction* OR case study OR test* OR (field AND (experiment OR trial)) OR intervention OR (randomised controlled).

Adhering to the definition of nudges from Thaler and Sunstein (2021), which is reiterated by Lieberoth et al. (2018), Mertens et al. (2022), Munscher et al. (2016), and Nisa et al. (2019), this study excludes all studies that tested economic incentives or disincentives (taxes, subsidies, and rebates) and regulations (bans and mandates).

The studies that quantitatively gauged behavioural shifts in individuals and collective units (i.e., households and offices) achieved through nudges leading to a lower contribution to air pollution were chosen to be included in this meta-analysis. This study has considered the following behavioural shifts:

• Indoor energy use: purchasing energy-efficient appliances, lowering indoor electricity and gas consumption (i.e., within houses and offices), and adopting clean cooking methods to reduce direct emissions from indoor energy use and indirect emissions from electricity generation.
• Waste burning: increasing the rate of waste recycling or lowering the amount of waste generated to reduce emissions from waste burning.
• Transport: choosing sustainable mobility options (public transport, walking, and cycling) over motorised private vehicles to reduce transport sector emissions.

The meta-analysis consists of only those studies that have measured behavioural change. In the case of the transport sector, as direct measurement of behavioural change is difficult, many studies assess the effectiveness of nudges based on high-quality proxies such as entries in travel diaries. Nudges have been divided into five categories based on the template in Nisa et al. (2019):

• Appeal: requests and pleas to change behaviours based on the shared principles of social cooperation and responsibility. For instance, Abrams et al. (2021) used a plea as a behavioural intervention: “Think about your actions. When the barriers are down, please turn off your engine.”
• Engagement: promoting goal-setting, intention to implement, commitment, and inculcating mindfulness towards air quality improvement. For instance, Franssens et al. (2021) drafted a message – “Naturally, I use public transport. During the week or the weekends, it is natural [that] you travel sustainably.” It was printed on travel card holders and distributed to bus passengers to reinforce their intention to use public transport.
• Environmental alteration: removing external barriers or expediting access to the desired option by altering the physical environment. For instance, in Bensch and Peters (2015), the researchers randomly provided improved cookstoves to some households participating in the baseline survey.
• Information: providing messages with helpful tips, air pollution-related statistics, in-home displays, and energy efficiency labels to change behaviour. For instance, Aittasalo et al. (2012) provided pedometers and logbooks to participants in the treatment group to keep track of the number of steps they were walking each day.
• Social norms: providing comparative information about the air pollution reduction behaviours of other close cohorts, such as other citizens, colleagues, and neighbours. For instance, Delmas and Lessem (2014) tested two treatments that provide feedback on the energy use in dormitory rooms compared to similar rooms on a university campus.

The final sample consists of 102 studies from 1977 to 2023, with 151 individual estimates used for the subsequent analysis.

Meta-analysis

The nature of the reported outcomes in the individual studies is diverse – mean, t-statistic and regression coefficient are the recurring types of statistics used.

As a result, the magnitude of the outcomes and the units of measurement vary greatly. Therefore, the first step was to standardise the outcomes relative to their variability (as measured by a standard deviation or standard error).

For this purpose, the SMD or Cohen’s d has been chosen in this study (Cohen 1988). All the outcomes are presented as positive, as all the interventions targeted positive behavioural shifts to lower emissions. Therefore, the outcomes were multiplied by -1 wherever a reduction was targeted (e.g., in the amount of waste generated, energy consumed, or use of motorised private vehicles) to ensure all the values were positive.

An online tool, Campbell Collaboration’s Practical Meta-Analysis Effect Size Calculator, was used to calculate the SMD and the accompanying variance. The tool requires the outcome and error measure as inputs, producing the SMD and variance as outputs (Wilson 2023).

A random-effects meta-analysis model was chosen to pool the SMDs and compute the consolidated SMDs. Assuming that heterogeneity (i.e., the degree of variation not due to chance in the included outcomes in the meta-analysis) is non-existent would have been erroneous. Hence, the random-effects model was used (Harrer et al. 2021). To compute the pooled SMDs according to the random-effects model, a method that estimates heterogeneity on the same scale as the standard errors of the individual effect sizes is needed. The Empirical Bayes method was chosen because of its higher accuracy vis-à-vis other estimation methods across the entire range of heterogeneity values (Sidik and Jonkman 2007; van der Linden and Goldberg 2020).

The I² statistic was used to gauge the magnitude of heterogeneity. The I² statistic is the percentage of variability in the effect sizes not caused by sampling errors. It helps to ascertain whether the estimated pooled effect size is statistically precise (Deeks et al. 2024). The I² statistic is the most widely used measure of heterogeneity, primarily because it is easy to interpret. An I² value of less than 40 per cent represents low heterogeneity (i.e., high statistical precision), and an I² value of more than 40 per cent represents high heterogeneity (i.e., low statistical precision).

Then, the ‘small-study bias’ was gauged in our dataset. Small-study bias arises when the probability of publishing a reported outcome depends on its magnitude. Researchers only report large and statistically significant outcomes from studies with small sample sizes (Harrer et al. 2021). Hence, the data becomes asymmetric, leading to inflated pooled effect sizes. This study uses a robust quantitative technique to assess whether small-study bias exists.

We use the Pustejovsky–Rodgers modified Egger’s regression test to assess small-study bias quantitatively (Pustejovsky and Rodgers 2019). This test is an improvement over the conventional Egger’s regression between the standardised SMD and the estimate’s precision, correcting for the dependent relationship between standard error and SMD (Harrer et al. 2021). This modified test was chosen because it suits the evaluation of SMDs. The intercept of this regression model was used for this assessment: if it was not statistically zero, then it could be inferred that a small-study bias existed.

In addition, the factors driving heterogeneity across the primary studies were investigated using a metaregression. A meta-regression aims to predict study outcomes using various study characteristics (e.g., publication year, sample selection, etc.) as predictor variables. The statistical significance of the variable coefficients tells whether they are influential enough to dictate the study outcomes and, therefore, impact heterogeneity.

According to Cohen (1988), an SMD of 0.2 is considered small. However, in psychology, it is a meaningful effect size and potentially consequential when scaled to the population level. Therefore, an SMD of 0.2 serves as the benchmark, with an SMD of less than 0.2 indicating low effectiveness and an SMD of more than 0.2 indicating high effectiveness at the population level (Carlo et al. 2005; Funder and Ozer 2019; Gignac and Szodorai 2016; Ozer and Benet-Martinez 2006; van der Linden and Goldberg 2020).

The Government of India’s Mission LiFE presents an opportunity to promote mass behavioural changes that reduce individual emissions, improving air quality. The central government aims to effect a change in demand and supply based on behavioural shifts in consumption patterns through long-term government and industrial policies. The programme aims to make at least 80 per cent of India’s villages and cities environment-friendly and nudge at least 1 billion Indians to practise a sustainable lifestyle by 2027–28 (MoEFCC and NITI Aayog 2022). Therefore, changing individual and community-level consumption patterns is central to Mission LiFE.

Shifting individual consumption patterns is critical to improving air quality in India. Three sectors— indoor energy use (including electricity generation), transportation, and waste burning—are responsible for 52 per cent of India’s annual mean PM_2.5 concentration (Chatterjee et al. 2023). Individual choices primarily drive emissions in these sectors. Therefore, changing individual polluting behaviours can help achieve cleaner air in India in the long term.

Nudges can be an effective tool for shifting consumption patterns. They aim to steer people in a particular direction by changing aspects of the decision-making context in a subtle, unobtrusive, and cost-effective way (Lieberoth et al. 2018). Given their cost-effectiveness, researchers can deploy them as a pilot and then scale them to the regional and national levels based on the emerging evidence.

Even though researchers have tested the impact of nudges on polluting behaviours through RCTs, summarising the available evidence through a meta-analysis can help aggregate information on which interventions work and in which context. This knowledge can help researchers deploy or avoid specific interventions. Nisa et al. (2019) conducted the last comprehensive meta-analysis of such RCTs, summarising the available evidence on the impact of nudges on polluting behavioural patterns using data from studies published until 2017. Given the increased volume of available evidence, the need for an updated meta-analysis covering all the available evidence has also grown. This study aims to fill that gap.

This study finds that engagement (i.e., promoting goal-setting, intention to implement, commitment, and mindfulness of air quality improvement) and providing information (i.e., messages with helpful tips, air pollution-related statistics, in-home displays, and energy efficiency labels) are minimally effective in shifting people’s transportation choices towards more sustainable options. The low effectiveness of engagement and information provision in changing individuals’ transportation choices is estimated with low and high precision, respectively. Information provision is also minimally effective in reducing indoor energy use, according to this study, and the low effectiveness of this strategy is estimated with high statistical precision.

Engagement is highly effective in promoting waste recycling or reduction, but the impact is estimated with low precision. However, all the individual interventions testing the impact of engagement on promoting waste recycling or reduction report having achieved positive behavioural shifts. However, most evidence comes from self-selected samples (i.e., participants explicitly agreeing to participate in the RCT). Therefore, more evidence from the general population is needed before endorsing its effectiveness.

Environmental alterations (i.e., removing external barriers or expediting access to the desired option by altering the physical environment) effectively reduce indoor energy consumption. However, the effect is estimated with low precision. Limited evidence suggests that environmental alteration effectively reduces indoor energy consumption; however, all the tested interventions reported achieving such a reduction. More evidence is needed to bolster this conclusion.

Setting social norms (i.e., providing comparative information about the emissions reduction behaviours of close cohorts such as citizens with a comparable demographic profile, colleagues, and neighbours) is minimally effective in reducing indoor energy consumption; however, the low effectiveness is estimated with low statistical precision.

Although evidence on such interventions is available from six continents, there is limited evidence from the Global South, which consists of Asia, Africa, and Latin America, specifically from India. The scarcity of evidence from the Global South limits the scope of this meta-analysis’s application to India. However, researchers and ULBs can extract relevant information on which strategies work and which don’t to help direct further research in India.

In 2024, the CPCB began implementing a quarterly ranking of cities against annual targets based on their CAPs (CPCB 2024). The metrics include compliance of the CAPs with 14 Mission LiFE indicators that capture individual choices, such as preferring CNG and electric vehicles over petrol and diesel ones and practising waste segregation at home (CPCB 2024; MoEFCC 2023). Therefore, the central government has incentivised cities to promote individual and community-level emissions reduction through this performance-based ranking initiative.

However, Mission LiFE initiatives can only tailor their interventions to change polluting behaviours when researchers and decision-makers know which strategies are most effective in each sector. The results obtained through this meta-analysis will help direct further research in the Indian context to this end, which will serve the dual purpose of increasing the quantum of available evidence and potentially effecting scalable behavioural changes.