What makes food policy acceptable? Package design, true-cost information, and public support in Switzerland

Last registered on September 09, 2026

Pre-Trial

Trial Information

General Information

Title
What makes food policy acceptable? Package design, true-cost information, and public support in Switzerland
RCT ID
AEARCTR-0019505
Initial registration date
September 02, 2026

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
September 09, 2026, 3:44 PM EDT

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

Locations

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

Request Information

Primary Investigator

Affiliation

Other Primary Investigator(s)

PI Affiliation
Unisanté

Additional Trial Information

Status
In development
Start date
2026-09-02
End date
2026-10-30
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Food systems have far-reaching impacts that result in public health costs and environmental damage which are not reflected in retail prices, but are borne implicitly by society. True Cost Accounting has emerged as a framework for making these costs visible, and is increasingly invoked to justify food policies that mitigate and internalize the public health and environmental health costs in food prices. Yet operationalization of true cost food policies to date has focused almost entirely on consumers- on true-cost labels and true pricing at the point of purchase - while a central element of its proposed theory of change remains untested: whether communicating these costs to the public affects support for the policy instruments that would mitigate and internalize the true cost of food in practice. On 27 September 2026, Swiss citizens vote on the «Initiative für eine sichere Ernährung (Ernährungsinitiative)», which calls for greater domestic food self-sufficiency and a shift toward plant-based production and consumption. This pre-registered study leverages that vote to examine what makes true cost food policy acceptable to voters along two dimensions: how a policy is designed, and what people know about the problem it addresses. In the second wave of a two-wave online survey with approximately 2,200 eligible Swiss voters, participants are randomly assigned to one of three conditions: a neutral control, a qualitative description of the food system's health and environmental impacts, or the same description with those impacts quantified as costs in Swiss francs per person per year. Following treatment, participants rate seven food policy measures individually and then evaluate policy packages in a conjoint experiment in which each measure varies in stringency and revenues from these measures vary in their use. This design allows us to estimate which policy package design features drive acceptability, whether true-cost information raises support for policy instruments, and, exploratorily, whether any such effect operates by correcting mistaken beliefs. The study speaks to literatures on policy packaging, information provision, and the political feasibility of food systems transformation.
External Link(s)

Registration Citation

Citation
Dassesse, Kate and Laurence Jeangros. 2026. "What makes food policy acceptable? Package design, true-cost information, and public support in Switzerland." AEA RCT Registry. September 09. https://doi.org/10.1257/rct.19505-1.0
Experimental Details

Interventions

Intervention(s)
Participants are randomly assigned to one of three conditions in the second wave of a two-wave panel survey. The intervention varies the information respondents receive about the environmental and health impacts of the Swiss food system before they evaluate food policy measures and packages.

Participants in the control condition read a short text on a food-related topic unconnected to the environmental, health or price dimensions of the food system. Participants in the qualitative information condition read a text describing the health and environmental impacts generated by the Swiss food system, without any monetary figures. Participants in the true-cost information condition read a text with the same content, in which those impacts are presented as costs and quantified in Swiss francs per person per year, with an uncertainty range and the source of the estimate.

The three texts are matched on length, paragraph structure, the presence of a quantity, and the citation of an institutional source. None of them makes a claim about the effectiveness of any policy measure.
Intervention Start Date
2026-09-30
Intervention End Date
2026-10-30

Primary Outcomes

Primary Outcomes (end points)
1) Acceptability of individual policy measures. Following the treatment, respondents rate seven food policy measures on a 7-point Likert scale (1 = strongly oppose, 7 = strongly support). The measures correspond to the seven conjoint attributes.

2) Package acceptability. In the conjoint experiment, the primary outcome is the probability that a given package profile is chosen. Two quantities are estimated from it: a) average marginal component effects (AMCEs), used to test which package design features drive acceptability; and b) marginal means, used to compare attribute effects across the three conditions.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
1) Package acceptability ratings. Respondents rate each package on a 7-point Likert scale. Unlike the forced choice, this is an absolute rather than relative judgement and therefore supports estimation of a treatment main effect on package acceptability.

2) Referendum acceptance. After each choice task, respondents indicate whether they would accept their chosen package in a popular vote. This provides an absolute measure of support that maps onto the institutional setting in which the study is embedded.

3) Perceived effectiveness and perceived importance. Two 7-point items measure whether respondents consider a levy on high-impact foods effective in reducing the environmental and health burden of the food system, and how important they consider it that policy accounts for food costs not reflected in retail prices. Both items are worded so as to be self-contained and equally interpretable in all three arms.

4) Beliefs about hidden costs. Respondents estimate the environmental and health costs of the Swiss food system not included in retail prices, in Swiss francs per person per year, using an open numeric field. Beliefs are elicited before randomisation and again after the conjoint, using the same scope and units with lightly varied surface wording. A bracketed fallback is displayed only to respondents declining the numeric field.

5) Placebo belief. Respondents estimate public spending per person per year on culture, in both the pre- and post-treatment blocks. The treatment is not expected to affect this quantity.

6) Perceived credibility of the information. Respondents rate how credible they find the text presented to them, on a 7-point Likert scale, in all three conditions. It is reported descriptively and used to interpret the treatment effects.

7) Comprehension. One single-best-answer item per arm, matched across arms on number of response options, placed after the information treatment. Analysis is intention-to-treat; comprehension is reported descriptively and used only in a pre-specified secondary analysis restricted to respondents answering correctly.

Secondary Outcomes (explanation)

Experimental Design

Experimental Design
This study employs a randomized controlled trial (RCT) embedded in the second wave of a two-wave online survey experiment fielded around a Swiss popular vote on the Ernährungsinitiative. In Wave 2, and after all outcomes of the companion study have been collected, participants estimate the health and environmental costs of the Swiss food system. They are then randomly assigned to one of three conditions (control, qualitative information, true-cost information) that vary whether these impacts are presented and, if so, whether they are quantified in Swiss francs per person per year (see Interventions). Following treatment exposure, participants rate seven food policy measures individually and then evaluate policy packages in a conjoint experiment, in which each measure varies across three levels of stringency and revenues vary in their use. Participants subsequently re-estimate the hidden costs of the food system, allowing pre- and post-treatment beliefs to be compared.
Experimental Design Details
Not available
Randomization Method
Randomisation to the information treatment is executed by the survey platform Qualtrics, at the point of entry to the treatment module, with the aim of achieving balanced group sizes across the three conditions. Randomisation is independent of all measured covariates; balance on key pre-treatment covariates (e.g., age, gender, education, political orientation) will be verified in the final data, and any imbalances will be reported in the analysis.

Conjoint attribute levels are drawn independently by the survey platform for each package in each task, according to a uniform distribution over the three levels of each attribute
Randomization Unit
The individual respondent.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
Not applicable, the sampling design is not clustered.
Sample size: planned number of observations
The target sample size is 2,200 respondents completing Wave 2 of the panel. Wave 1 is fielded before the federal vote on the Ernährungsinitiative of September 27, 2026; Wave 2 is fielded after the vote. The final achieved sample depends on panel attrition between waves; wave-to-wave attrition will be reported together with a balance table on Wave-1 covariates by Wave-2 treatment arm. Each respondent completes four conjoint choice tasks, each comprising two package profiles, yielding approximately 8,800 choice observations. Each respondent additionally provides seven individual-measure acceptability ratings, yielding approximately 15,400 measure-level observations.
Sample size (or number of clusters) by treatment arms
Information treatment. Respondents are randomly assigned to one of three conditions with the aim of achieving balanced group sizes:
- Control: approximately 733 respondents
- Qualitative information: approximately 733 respondents
- True-cost information: approximately 733 respondents
Actual sizes may vary slightly as a result of randomisation and completion patterns.

Conjoint experiment. All respondents completing Wave 2 take part in the conjoint experiment. Because attribute levels are randomised within respondent, all 2,200 respondents contribute to the estimation of every attribute effect, irrespective of information-treatment assignment.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
All figures are minimum detectable effects at 80% power, α = 0.05 two-sided, at N = 2,200 completing Wave 2 (approximately 733 per condition). Estimates are simulation-based. Conjoint estimands use a within-task-differenced linear probability model with respondent-clustered CR1 standard errors; the clustered estimator was verified by simulation to have a standard error 1.41 times the classical one, and this factor is applied throughout. Individual-measure acceptability (inverse-covariance-weighted index across seven items, 7-point scale). Pairwise contrast between two conditions: 0.146 SD, equivalent to approximately 0.22 scale points if the index standard deviation is 1.5. Comparison of the two information conditions pooled against control: 0.127 SD, approximately 0.19 scale points. Conjoint attribute effects (AMCEs on choice probability, pooled across conditions, all 2,200 respondents contribute): 3.6 percentage points. Comparison of attribute effects across conditions (difference in marginal means between two conditions): 4.6 percentage points for a single pre-specified comparison, rising to 6.2 percentage points under Bonferroni correction across the fourteen attribute-level comparisons. Package acceptability rating (7-point scale, eight profile ratings per respondent, assumed respondent-level ICC of 0.5): 0.110 SD, approximately 0.16 scale points.
IRB

Institutional Review Boards (IRBs)

IRB Name
ETH Zürich Ethics Commission
IRB Approval Date
2026-08-25
IRB Approval Number
Project 26 ETHICS-286