How climate- and sustainability-related food labels change individuals' attitudes, knowledge, norms, and behaviors: A field and survey experiment

Last registered on May 03, 2022

Pre-Trial

Trial Information

General Information

Title
How climate- and sustainability-related food labels change individuals' attitudes, knowledge, norms, and behaviors: A field and survey experiment
RCT ID
AEARCTR-0009331
Initial registration date
April 29, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
May 03, 2022, 9:40 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Primary Investigator

Affiliation
University Bern/ETH Zürich

Other Primary Investigator(s)

PI Affiliation
University of Bern

Additional Trial Information

Status
In development
Start date
2022-05-06
End date
2022-06-17
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Our current food consumption choices have detrimental consequences on the environment. In particular, the consumption of meat products comes with high environmental costs (Godfray et al. 2018; Poore and Nemecek 2018; Springmann et al. 2018). Without substantial reductions in meat consumption, the Paris climate targets are difficult to achieve (Clark et al. 2020). Thus, encouraging individuals to reduce their meat consumption is a promising route to mitigating climate change, especially by lowering short-lived methane emissions that increase the risks of crossing climate tipping points in the near-term (Fesenfeld et al. 2020; Godfray et al. 2018; Poore and Nemecek 2018; Springmann et al. 2018). Yet, although the negative consequences of meat consumption are well understood, meat production and consumption still increase in many countries (FAO 2020).

However, transforming dietary patterns, especially shifting towards more plant-based diets, is not easy. There is a growing body of literature on the different internal and external factors enhancing or inhibiting sustainable food consumption behaviors. Studies have mainly highlighted the role of consumer nudging and information on the environmental and health impacts of excessive meat consumption as a lever for changing dietary patterns (Apostolidis and McLeay 2016, Fesenfeld et al. 2022; Hagmann, Siegrist, and Hartmann 2019; Kamm et al. 2015; Stubbs, Scott, and Duarte 2018; Lemken, Zühlsdorf, and Spiller 2021; van Loo, Hoefkens, and Verbeke 2017; Pohjolainen et al. 2016). Nudging and information techniques seek to steer people's behavior without mandating or forbidding options, such as providing default information on the climate and animal welfare impact of food products in the form of product labels.

However, there are very few real-world examples of retailers that provide climate- and sustainability-related food product labels on most of their food products. Consequently, there is a lack of field experimental evaluations of such labels, especially on the direct labeling effects on consumers' actual food behavior choices as well as potential feedback of such labels. We thus know little about how such labels affect consumer attitudes, knowledge, perceived behavioral control, social norms, behavioral intentions, and actual food behavior choices (Lemken, Zühlsdorf, and Spiller 2021). We also have little evidence on how such labels feed back into the public policymaking process by changing norms and public opinion about policies to transform the food system and reduce meat consumption (Fesenfeld et al., 2022).

Here, we conduct large-scale randomized field- and survey experiments with a representative sample of 2000 Swiss citizens. We collaborate with Migros, one of Switzerland's two largest food retailers, and evaluate their novel M-Check label. The M-Check label is a private labeling initiative that Migros launched in 2021. This label ranks their food products from 1 (worst type) to 5 (best type) stars in terms of animal welfare and CO2 emissions. The rankings were developed and verified in cooperation with external partners. The framing and conjoint experiments will be conducted in two waves with N = 1000 respondents shortly before the real-world information campaign (t1) and N = 1000 respondents shortly after and during the information campaign (t=2). As part of the framing experiment, the respondents will be randomly divided into either the control group getting no additional information or the treatment group getting both, a short information display that matches the information provided by the real-world campaign and a product comparison slide illustrating the M-Check label use. As part of the conjoint experiment, respondents are confronted with sets of randomly varied food policy package designs and choose between these differently designed policy packages.

As part of our two survey waves, we collect various self-stated consumer attitudes, knowledge, perceived behavioral control and norms, behavioral intentions as well as socio-demographic and further control variables. In addition, we receive actual and longitudinal shopping behavior data for those respondents of our survey that provided their consent. This data allows us to compare treatment effects for both stated and revealed preferences and behavioral choices.

In terms of our case selection, we choose Migros as a partner as it is the second-largest food retailer in Switzerland, with a market share of about 35 percent (Statista 2020). The Migros M-Check campaign is a unique case for this study due to the rather sizeable visibility of the labeling initiative for consumers in Switzerland and the potential impact of the label on actual consumer behaviors. Moreover, Migros is one of the first larger supermarket chains worldwide that introduced a climate and animal welfare label on most products. Similar labels are often just applied to a small range of products rather than most of the products offered in a supermarket. Thus, this setting offers a unique opportunity for a real-world experimental evaluation of the effectiveness of climate- and sustainability-related labels and their potential feedback effects on social norms and public opinion about food system transformation.

Building on the theory of planned behavior (Ajzen, 1991) and related dual-processing theories of human decision making (Kahnemann 2011; van Loo, Hoefkens, and Verbeke 2017; Menzel 2013; Smith and DeCoster 2000), we expect that the M-Check label on the climate- and animal-welfare impact of different food products positively affects consumers' attitudes on buying more sustainable food products and consuming less meat. We also expect positive effects of the M-Check label on consumers' sustainability-related knowledge, their perceived social norms, and behavioral control to buy more sustainable food products and consume less meat. Further, we expect that the M-Check label changes peoples’ behavioral intentions and actual shopping behavior, i.e. leading treated consumers to buy more sustainable food products and less meat. Finally, we expect that such voluntary labels by food retailers positively affect citizens' perception of food retailers' sustainability impact and public support for different governmental policies to transform the food system and consume less meat.

Overall, the findings from this study will help to have a clearer picture of what types of sustainability-related labels and information are effective in influencing attitudes, norms, and behaviors. Moreover, the findings of this large-scale real-world experiment will reveal what kinds of food policies are supported, especially when someone has more information on the climate and animal welfare impacts of food consumption. The findings will also give more insight into the perception of private labeling initiatives initiated by economic actors and their influence on citizens' perceived need for more governmental regulation in the food sector. Lastly, the findings and the resulting policy implications can be helpful in future climate- and sustainability-related food labeling schemes.



External Link(s)

Registration Citation

Citation
Fesenfeld, Lukas and Maiken Maier. 2022. "How climate- and sustainability-related food labels change individuals' attitudes, knowledge, norms, and behaviors: A field and survey experiment." AEA RCT Registry. May 03. https://doi.org/10.1257/rct.9331
Sponsors & Partners

Sponsors

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details

Interventions

Intervention(s)
The study is built around the M-Check information campaign launched by Migros, a Swiss food retailer, in May 2022 (calendar weeks 21 – 23), where they promote the newly introduced Migros M-Check label via different platforms, including TV, social media, etc. The M-Check product rankings regarding climate and animal welfare were developed and verified with external partners. The study will be conducted in two waves, with N = 1000 respondents shortly before the real-world information campaign (t1) and N = 1000 respondents shortly after the start of and during the information campaign (t=2). As part of the experiment within the survey, the respondents will be randomly divided into either the control group getting no additional information or the treatment group getting both, a short information display that matches the information provided by the real-world campaign, as well as a product comparison slide illustrating the label use.

In the following part, the choice of a survey experimental design with a field experiment component, as well as the addition of a conjoint experiment will be justified as an appropriate method to answer the research questions. Then the sampling method, the survey procedure and the data analysis methods will be described. This study's quantitative survey experimental part is a randomized controlled experiment with a control group that does not get treated and a group that receives an information treatment consisting of two parts. Participants will be randomly assigned to either the control or the treatment group. Due to randomization, the only systematic difference between the control and the treatment group should be the treatments, which allows estimating a causal effect of the treatments on the dependent variables, if the randomization works as intended and the sample is generally large enough (Stock and Watson 2020).
Further, since the survey experiment is conducted in two waves shortly before/after the real-world information campaign on M-Check by Migros we can use a random event during survey approach (first field experimental component). Due to the close temporal proximity of the data collection in the two waves, the only systematic difference between the representative samples in the first and the second wave of the survey should be the real-world information campaign running before and during the second survey wave. Furthermore, the survey respondents are asked to share their customer account numbers in the survey, such that we can later anonymously use the real-world food consumption purchasing data that is stored within their account to match them with the survey data and monitor potential changes in purchasing due to being assigned to the treatment (second field experimental component). Moreover, the treatment used in the experimental survey part of this study can be classified as a framing treatment. In general, framing is used to highlight or emphasize a specific view on a topic to make people think about this topic from that particular perspective (Chong and Druckman 2007). In this study the information treatment frame is used to make individuals think about the M-Check label and the underlying information on the climate and animal welfare impact of food products, especially meat products and the benefits of meat substitutes. Lastly, a conjoint experiment is added to randomly vary different policy attributes to be able to assess the effects of these attributes on the participants' policy preferences (Hainmueller, Hopkins, and Yamamoto 2014) and to determine the policy packages that are most likely also supported in real-world voting scenarios (Hainmueller, Hangartner, and Yamamoto 2015).

The two randomly varied experimental framing groups (the treatment and the control group in the survey) and the samples before/after the start of the real-world campaign are asked for their knowledge of food-related sustainability, their attitudes, perceived social norms, and behavioral control related to food consumption, their food purchasing intentions, and their policy support for different measures aiming to reduce meat consumption and improve animal welfare.
Intervention Start Date
2022-05-06
Intervention End Date
2022-06-17

Primary Outcomes

Primary Outcomes (end points)
In the following, you can see the key-dependent variables that have to be answered by all survey respondents:
• measuring food-related sustainability knowledge:
- determining the meat/meat substitute product with the lowest climate impact from a selection of 4 items (multiple choice true/false question)
• measuring attitudes:
- the perceived importance of paying attention to the climate impact and animal welfare criteria while shopping, as well as reducing meat consumption (7-point Likert scale from extremely unimportant to extremely important)
• measuring perceived social norms:
- questions on the perceived social norms related to the perceived importance of climate and animal welfare criteria while food shopping of the respondents' family, friends, coworkers, and the general Swiss population (7-point Likert scale from extremely unimportant to extremely important)
• measuring perceived behavioral control:
- the perceived ability/ease of paying attention to climate impact and animal welfare criteria while shopping, as well as to reduce the amount of meat purchased (7-point Likert scale from difficult to easy and from not at all doable to doable)
• measuring behavioral intentions:
- the probability that individuals pay attention to climate impact and animal welfare criteria when food shopping, reduce meat consumption and eat more meat substitutes (7-point Likert scale from extremely likely to extremely unlikely)
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Other dependent variables that have to be answered by all survey respondents include:
• measuring policy support:
- the intentions to support different types of general and more specific food policy measures and instruments (7-point Likert scale from strongly support to strongly oppose)
- the support of more governmental regulation of the food sector (7-point Likert scale from more to less)
- the willingness to pay taxes on meat products (Slider from -2.5% to 100% representing the desired change in the meat tax compared to the current level of 2.5% VAT in Switzerland)
• measuring the perception of Migros and other industry labeling initiatives:
- the perception of Migros as socially responsible, making a contribution to a food sector with a lower climate impact and better animal welfare (7-point Likert scale from fully agree to do not agree at all)
- the perceived effectiveness and credibility of industry labeling initiatives in general (7-point Likert scales from very effective to not effective at all and completely credible to not at all credible)
• We also measure support for differently designed policy packages in a conjoint experiment
The following paragraphs (see section Secondary Outcomes (Explanation) outline the introduction to the experiment, the seven policy package attributes, and the randomly varied policy attribute levels (see further details about the randomization in the section Randomization Method below).

Secondary Outcomes (explanation)
Introduction to conjoint experiment: This next part is about Swiss food and climate policy. Specifically, we will conduct a thought experiment with you in which we ask you to compare and evaluate alternative Swiss agricultural and food policy packages of measures that could be adopted in 2023. Each package of measures consists of various individual measures, all of which aim to reduce greenhouse gas emissions from the food sector to varying degrees. The following are the most important measures currently under discussion to reduce greenhouse gas emissions in the food sector:
o Financial government support for climate-friendly food (e.g., subsidies for plant-based nutrition). Effect: Reduction of prices for climate-friendly food products.
o Standards for producers (e.g., raising animal husbandry standards for meat producers). Effect: Producers are obliged to produce in a more animal-friendly way. This can lead to rising prices for animal products (e.g., meat).
o Taxes on meat products. Effect: The prices for climate-damaging foods, such as meat, rise faster than the prices for less climate-damaging foods.
o Restrictions (e.g., minimum proportion of meat-free dishes in public canteens). Effect: Reduces the consumption/use of particularly climate-damaging foods.
o Consumer information (e.g., mandatory government CO2 and animal welfare labeling on food). Effect: More transparency and more widely accessible information on the impact of food on climate and animal welfare.
o Abolition of state support for meat and feed producers. Effect: Increasing prices for animal products (e.g., meat).

We will now start the thought experiment on Swiss food and climate policy. In four different rounds of voting, we will present you with two alternative packages of Swiss measures side by side. In each round, please carefully compare the two policy packages and indicate which one you like better. Some of the policy packages may be very similar. If you do not support either package, please choose the package you dislike less. You can then rate both packages on a scale from "Fully Disagree" to "Fully Support".

The different randomized policy package attributes and levels are the following:
• Government support
o Strong subsidies (30% price reduction for plant-based food)
o Subsidies (15% price reduction for plant-based food)
o No subsidies
• Standards for producers
o Strong increase in animal husbandry standards
o Increase in animal husbandry standards
o No increase in animal husbandry standards
• Meat tax
o Strong increase in meat tax (30% price increase on meat products)
o Increase in meat tax (15% price increase on meat products)
o No increase in meat tax
• Restrictions
o At least 75% meat-free dishes in public canteens
o At least 50% meat-free meals in public canteens
o At least 25% meat-free meals in public canteens
o No restrictions
• Consumer information on climate impact of food products
o Mandatory government climate labeling for food products
o Voluntary government climate labeling for food products
o No state climate compatibility labeling
• Consumer information on animal welfare impact of food products
o Mandatory state animal welfare labeling for foodstuffs
o Voluntary national animal welfare labeling for food products
o No government animal welfare labeling
• Reduction of government support for meat and feed producers
o Complete abolition of subsidies
o Halving of subsidies
o No reduction of subsidies

Experimental Design

Experimental Design
Here, we conduct large-scale randomized field- and survey experiments with a representative sample of 2000 Swiss citizens. In our experiments, we seek to answer the following questions:

1. How do private firms’ carbon and animal welfare food product labeling initiatives affect individuals…
a.…food-related sustainability attitudes?
b.…food-related sustainability knowledge?
c....perceived behavioral control to change their food behavior?
d....perceived social norms to change their food behavior?
e....behavioral intentions to change their food behavior?
f.....actual food shopping behavior?
g.....perceptions of private firms' sustainability efforts in the food sector?
h.…support for (differently designed) governmental food policies?

We collaborate with Migros, one of Switzerland's two largest food retailers, and evaluate their novel M-Check label. The M-Check label is a private labeling initiative that Migros launched in 2021. This label ranks their food products from 1 (worst type) to 5 (best type) stars in terms of animal welfare and CO2 emissions. The rankings were developed and verified in cooperation with external partners. The framing and conjoint experiments will be conducted in two waves with N = 1000 respondents shortly before the real-world information campaign (t1) and N = 1000 respondents shortly after and during the information campaign (t=2). As part of the framing experiment, the respondents will be randomly divided into either the control group getting no additional information or the treatment group getting both, a short information display that matches the information provided by the real-world campaign and a product comparison slide illustrating the M-Check label use. As part of the conjoint experiment, respondents are confronted with sets of randomly varied food policy package designs and choose between these differently designed policy packages.

As part of our two survey waves, we collect various self-stated consumer attitudes, perceived norms, and behavioral intentions as well as socio-demographic and other control variables. In addition, we receive actual and longitudinal shopping behavior data for those respondents of our survey that provided their consent. This data allows us to compare treatment effects for both stated and revealed preferences and behavioral choices.

In terms of our case selection, we choose Migros as a partner as it is the second-largest food retailer in Switzerland, with a market share of about 35 percent (Statista 2020). The Migros M-Check campaign is a unique case for this study due to the rather sizeable visibility of the label initiative for consumers in Switzerland and the potential impact of the label on actual consumer behaviors. Moreover, Migros is one of the first larger supermarket chains worldwide that introduced a climate and animal welfare label on most products. Similar labels are often just applied to a small range of products rather than most of the products offered in a supermarket. Thus, this setting offers a unique opportunity for a real-world experimental evaluation of the effectiveness of climate- and sustainability-related labels and their potential feedback effects on social norms and public opinion about food system transformation.

Linking the theory of planned behavior (Ajzen, 1991) with dual-processing theories of human decision making (Kahnemann 2011; van Loo, Hoefkens, and Verbeke 2017; Menzel 2013; Smith and DeCoster 2000), we build an argument about how climate- and sustainability-related food labels can change peoples' attitudes, knowledge, norms, and actual behaviors and potentially feed back in the policymaking process by altering public opinion about governmental food policies.

While the theory of planned behavior (Ajzen, 1991) emphasizes the importance of individuals' attitudes, perceived norms, and behavioral control in changing behavioral intentions and actual behaviors, dual-processing theories of human decision making also emphasize the less conscious and peripheral route of information processing (the experiential system) that leads to rather fast decisions based on decision heuristics and learned behavior (Kahnemann 2011; van Loo, Hoefkens, and Verbeke 2017; Menzel 2013; Smith and DeCoster 2000).

Based on the combination of both theories we derive the following testable hypotheses on the effects of our experimental interventions on the M-Check label (see section Intervention above).

H1: The positive impact of the M-Check label on participants' food-related sustainability attitudes (i.e., buying more sustainable food products and consuming less meat) will be greater for those in the M-Check information treatment group than for those in the control group.

H2: The positive impact of the M-Check label on participants' food-related sustainability knowledge will be greater for those in the M-Check information treatment group than for those in the control group.

H3: The positive impact of the M-Check label on participants' perceived behavioral control to change their food behavior (i.e., buying more sustainable food products and consuming less meat) will be greater for those in the M-Check information treatment group than for those in the control group.

H4: The positive impact of the M-Check label on participants' perceived social norms to change their food behavior (i.e., buying more sustainable food products and consuming less meat) will be greater for those in the M-Check information treatment group than for those in the control group.

H5: The positive impact of the M-Check label on participants' intentions to change their food behavior (i.e., buying more sustainable food products and consuming less meat) will be greater for those in the M-Check information treatment group than for those in the control group.

H6: The positive impact of the M-Check label on participants' actual food behavior changes (i.e., buying more sustainable food products and consuming less meat) will be greater for those in the M-Check information treatment group than for those in the control group.

H7: The positive impact of the M-Check label on participants' perceptions of private firms' sustainability efforts in the food sector will be greater for those in the M-Check information treatment group than for those in the control group.

H8: The positive impact of the M-Check label on participants' support for (differently designed) governmental food policies to foster sustainability in the food sector will be greater for those in the M-Check information treatment group than for those in the control group.

We also expect that the treatment effects of the M-Check information treatment group on behavioral change intentions, actual behavior changes and policy support are significantly mediated via changes in participants' attitudes, knowledge, perceived behavioral control, and social norms.
Experimental Design Details
Randomization Method
Framing experiment:
The participants are randomly divided into the control and the treatment group using the randomizer feature on Qualtrics, the online survey software used to design the survey and collect the data. Further, it is specified in the Qualtrics randomizer tool that the participants should be distributed evenly into both groups, ensuring that approximately the same number of respondents will be randomly assigned to the control and the treatment group.

Conjoint experiment:
In the conjoint experiment, we used a customized javascript code in Qualtrics to randomly vary the attribute levels of different food policy packages consisting of seven different types of policies (see section Secondary Outcomes above). We ask respondents to evaluate profiles that combine multiple randomly assigned attributes. We used a conjoint design of fully randomized paired profiles in which each respondent was shown profiles of two different hypothetical policy packages displayed side by side. Hence, each policy measure constituted an attribute in the package to which it belonged, and the attribute values were randomly assigned such that the two policy packages in each pair differed in one or more attribute values. This paired-profiles design was chosen because research suggests it performs well at reducing social desirability bias and replicating real-world behavior (Hainmueller et al, 2015).

To help participants understand how the relevant policy measures function, before showing the pairs of policy packages, we showed a page with brief descriptions of each policy measure (see section Secondary Outcomes above).
Randomization Unit
Framing experiment: The unit of randomization will be individual respondents.
Conjoint experiment: The unit of randomization will be individual respondents and policy attributes within the conjoint experiment.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
2,000 respondents, we do not use cluster sampling
Sample size: planned number of observations
Framing experiment: 2,000 individuals living in the French- and German-speaking parts of Switzerland (1000 respondents in first wave t1, 1000 respondents in second wave t2); for all respondents that provided consent in the survey, we receive longitudinal (daily) actual shopping behavior data (on product level) from Migros. Thus, we estimate a very large number and fine-grained observations on respondents' shopping behavior. Conjoint experiment: 2000 respondents (1000 respondents in first wave t1, 1000 respondents in second wave t2), each conducting four rounds of choosing and rating one out of two randomly varied policy packages. In total, the number of observations in the conjoint experiment is thus 2000*4*2 = 16000 observations.
Sample size (or number of clusters) by treatment arms
Framing experiment:
Control Group: 1000 respondents (500 respondents in first wave t1, 500 respondents in second wave t2)
Treatment Group: 1000 respondents (500 respondents in first wave t1, 500 respondents in second wave t2)
Conjoint experiment:
2000 respondents (1000 respondents in first wave t1, 1000 respondents in second wave t2), each conducting four rounds of choosing and rating one out of two randomly varied policy packages. In total, the number of observations in the conjoint experiment is thus 2000*4*2 = 16000 observations.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
IRB

Institutional Review Boards (IRBs)

IRB Name
University of Bern, Faculty of Business Economics and Social Sciences Ethics Commission
IRB Approval Date
2022-05-02
IRB Approval Number
N/A (IRB is still under review)
Analysis Plan

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials