Voters’ preferences for a meat tax: Evidence from a referendum choice experiment

Last registered on November 30, 2021


Trial Information

General Information

Voters’ preferences for a meat tax: Evidence from a referendum choice experiment
Initial registration date
November 24, 2021
Last updated
November 30, 2021, 3:35 AM EST



Primary Investigator

University of Hamburg

Other Primary Investigator(s)

PI Affiliation
University of Hamburg

Additional Trial Information

In development
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
The animal farming industry is in the public eye. Consumption of meat and dairy products and its consequences are subjects of discussion in society and politics alike.
The public debate focuses on two aspects: (1) livestock sector’s impact on climate and environment and (2) farm animals’ husbandry and living conditions. 14.5% of all human-induced greenhouse gas emissions are attributed to the livestock sector (Gerber et al. 2013). Moreover, husbandry conditions, especially in intensive livestock farming, are increasingly discussed in public and even questioned by meat eaters (Eurobarometer 2016).
The production and consumption of animal products generate negative externalities regarding the environment and animal welfare in society (Springmann et al. 2016; Lusk 2011). Politics are increasingly confronted with the deficiencies of the system and start to take them into account, e.g. in the European Commission’s “Farm to Fork” strategy (European Commission 2020).
In Germany, policy makers discuss potential regulations for both issues. In 2021, a CO2 price for fossil fuels in the heating and transportation sector has been introduced in Germany (Bundesregierung 2021). In this regard, the German Green Party suggested a climate charge on animal products (Maurin 2019). While they could not come through with it, such a tax is not off the table yet. The German Ministry of Food and Agriculture, on the other hand, proposes a transformation of the German livestock sector towards higher animal welfare standards. Apart from plans of introducing a state animal welfare label, an expert commission also suggested to implement an animal welfare consumption tax, a so called “Tierwohlabgabe”, to raise funds for supporting farmers who rebuild their stables and farms to provide more animal welfare (BMEL 2021).
Both debates are conducted rather separately from each other, but target the same product and industry. Policy changes to internalize either externality would impact prices for animal products. We are thus interested in how voters perceive the two potential tax schemes.

There is a broad literature on taxing certain food products, such as sugar or fat (e.g., Berardi et al. 2016; Colchero et al. 2016; Cornlesen & Carreido 2015; Smed 2012). Regarding the taxation of meat or animal products, the effects of certain tax schemes on demand, greenhouse gas (GHG) emissions and/or societal health have been modeled for several countries (e.g., Funke et al. 2021 for the world; Moberg et al. 2021 and Säll et al. 2020 for Sweden; Bonnet et al. 2018 for France; Dogbe & Gil 2018 for Catalonia Spain; Springmann et al. 2018 for world regions; Chalmers et al. 2016 for Scotland; Edjabou & Smed 2013 for Denmark; Wirsenius et al. 2011 for the EU). From this research follows that specifically taxing meat could have a strong steering effect.
Nonetheless, implementing such taxes is clearly challenging from a political point of view. Numerous studies, both surveys and choice experiments, thus examine people’s preferences regarding (carbon) tax schemes in general, but also in sub-sectors such as transportation and animal products (e.g., Baranzini et al. 2021; Douenne & Fabre 2020; Fesenfeld et al. 2020; Grimsrud et al. 2019; Hagmann et al. 2019; Hardisty et al. 2019; Baranzini & Carattini 2017; Brännlund & Persson 2012; Sælen & Kallbekken 2011; Hsu et al. 2008). Thereof, certain policy specifications are found to increase people’s support for such taxes: (1) refraining from calling the charge a “tax”, (2) earmarking tax revenues for environmental purposes, (3) redistributing tax revenues preferably as lump-sum payments, but earmarking still preferred over redistribution in direct comparison, (4) progressive taxing, and (5) clearly explaining tax impact/effectiveness (Carattini et al. 2018; Klenert et al. 2018).
However, we are not aware of any research directly comparing preferences of introducing a tax on the same product, namely meat, but for different arguments. Our first research question is hence: Do support rates for a meat tax differ depending on whether it is levied for environmental or animal welfare reasons?

The support for a meat tax should clearly be linked to the tax rate. We therefore investigate people’s support in a referendum-like choice experiment letting them vote for either the status quo or a new meat tax policy with increasing tax levels. With such a referendum scheme, hypothetical bias is limited to a certain extent as the decision is concrete and realistic (Carson & Groves 2007). Furthermore, we inform participants that the outcome of the referendum will be forwarded by us to the committee of the Federal Parliament of Germany in charge for the respective tax type.
For the policy proposals shown to participants, we apply the above mentioned five success factors. In addition, we only consider consumption tax schemes, i.e. only consumers of the product are charged for the amount they buy. However, we are further interested in people’s preferences regarding the degree of differentiation of the consumption tax.
The consumption could either be a fixed and flat amount charged on every kilogram sold, independent from the type of meat or animal welfare level provided such as the proposed “Tierwohlabgabe” or the “EEG-Umlage” (charge for renewable energies) used to finance subsidies for renewable energy in the electricity sector. Another option would be to implement a more differentiated scheme in line with a Pigouvian tax to bring consumption to a socially optimal level (Pigou 1920). Meat types with higher underlying carbon emissions are charged a higher tax than those with lower emissions. The same principle could be applied to animal welfare with products produced under higher animal welfare standards burdened less than those produced under lower levels.
Theoretically, both schemes are likely to affect demand for meat. The flat taxes are expected to mainly affect how much meat is consumed (see e.g., Nordgren 2012 for GHG tax on meat). The taxes differentiating between different levels of externalities associated with the product are expected to affect both the level of meat consumption and also the composition of meat products consumed (see e.g., Wirsenius et al. 2011 for GHG weighted tax on meat). We do not examine actual consumption behavior in our experiment. Instead, we are interested in people’s preferences for certain tax schemes.
Our second research question is thus: Do support rates differ depending on the degree of differentiation of the meat consumption tax?

To identify drivers of the latter, we exogenously vary the salience of the expected effect on the composition of meat products consumed by randomly switching the order of the referendum choices and a belief elicitation task. The latter asks respondents whether they believe a specific tax scheme will change the level and composition of meat consumption both for themselves and for other participants of the survey. The salience of the behavioral response is larger for those participants who answer the belief elicitation before the referendum task. Thinking through potential effects of a tax scheme might serve as a proxy for experiencing the effects of that scheme.
Our third research question is: Does the salience of own expectations about the tax’s effect on consumption behavior impact support rates?
External Link(s)

Registration Citation

Perino, Grischa and Henrike Schwickert. 2021. "Voters’ preferences for a meat tax: Evidence from a referendum choice experiment." AEA RCT Registry. November 30.
Experimental Details


The experiment comprises a between-subjects design.
There are four experimental groups that are asked for their support for a meat consumption tax scheme. The tax schemes are: (1) flat tax for carbon emissions, (2) differentiated tax for carbon emissions, (3) flat tax for animal welfare, (4) differentiated tax for animal welfare.

In each of the experimental groups, we additionally randomize the sequence of the referendum task and a belief elicitation task, increasing the total number of treatments to eight. The belief elicitation task asks respondents to state their expectations about how the tax scheme presented to them would change their own behavior in terms of total meat consumption and the composition of their meat consumption. They are also asked to predict the answers of the other survey participants who are presented the same tax scheme. The answer to the latter is incentivized.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Referendum choice experiment on six different tax schemes.

The key outcome variable is a subject's vote for the proposed tax scheme in a choice set. Variable "choice" is equal to 1 for voting for the proposal and 0 for voting against it.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Belief elicitation task regarding the proposed tax scheme’s impact on meat consumption behavior:
- Beliefs regarding change in (1) own total meat consumption and (2) composition of own meat consumption operationalized as the direction of change in the share of each of the four different meat/animal welfare types (a-d)
- Beliefs regarding change in (3) other participants’ total meat consumption and (4) composition of other participants’ meat consumption for each of the four different meat/animal welfare types (a-d)

Outcome variables 1 and 2a-d are measured on an ordinal scale with answers 1 = decrease, 2 = increase, 3 = stay the same.

Outcome variables 3 and 4a-d are measured on a continuous scale as the percentage of other participants choosing answers 1/2/3 in questions 1 and 2a-d. The percentages for answers 1/2/3 need to sum up to 100%. In each of the five questions (3, 4a-d), subjects have to estimate two variables: the percentage of “decrease” and the percentage of “increase”. The percentage of “stay the same” is the residual which is bound to be non-negative and weakly below 100%.

The elicitation for beliefs regarding others’ consumption change is incentivized. For each of the five questions (3, 4a-d), the absolute difference in percentage points between the two percentages provided by a subject (for “decrease” and “increase”) and the two true percentages based on answers to questions 1 and 2a-d is calculated. Participants get a bonus of 0.10€ for each value they estimate if the sum of the absolute difference between their estimate and the true value is smaller than 2 percentage points. Total bonus payments per subject are hence between 0€ and 1€ in steps of 0.10€.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The referendum choice experiment is part of an online survey.
Respondents are representative for the German population in terms of age, gender and region of living (on a federal state level).
Subjects are randomized into one of four experimental groups (see Interventions section above). They receive six choice sets that differ in the tax amount proposed. The proposed meat tax amount increases from the first to the last choice set.

The design is supposed to test the following hypotheses:
Previous choice experiments find that public support for a new carbon tax policy declines with increasing tax rates (Baranzini & Carattini 2017; Gevrek & Uyduranoglu 2015; Brännlund & Persson 2012). We therefore expect a similar effect for the policy proposal of a meat tax.

Hypothesis 1: The support rate for a meat tax decreases with increasing tax levels, independent from tax type and purpose.

Regarding the effect of the different tax purposes on support rates, the hypothesis is not equally straightforward. There is ample research studying which arguments convince people to consume less meat apart from the impact of nudges or social norms. In incentive compatible experiments, carbon labels are found to reduce meat consumption (Camilleri et al. 2019; Perino et al. 2014). Comparing nutritional and animal welfare appeals, the latter is more effective in reducing meat consumption (Palomo-Vélez et al. 2018). Survey results indicate that information on animal welfare and health have a stronger effect on consumption intentions than climate change arguments (Cordts et al. 2014). In a panel food diary study comparing health, climate and animal welfare arguments, Perino & Schwirplies (2021) find that only information on animal welfare significantly reduce red meat consumption, while climate change aspects only shape reduction intentions. Several choice experiments examine the effect of nutritional, carbon and animal welfare labels on people’s meat consumption choices. Findings indicate that consumers have a high willingness-to-pay (WTP) for nutritional arguments, but also for animal welfare with varying degree (Carlsson et al. 2021; Apostolidis & McLeay 2019; Gracia & de-Magistris 2016; Grunert et al. 2014; Van Loo et al. 2014). WTP for carbon labeling on meat products is, however, rather low (Koistinen et al. 2013). Thus, animal welfare aspects seem to be of higher importance for consumers as is meat consumption’s impact on the environment.

In addition, the current economic situation with its increasing energy prices in combination with the introduced carbon price for the heating and transport sector in Germany in the beginning of 2021 might lead to additional refusal of taxing meat for its emissions.

Hypothesis 2: The support rate is higher for an animal welfare tax than for a CO2 tax across all tax levels.

Finally, we look at the degree of differentiating the meat tax. We do not want to identify the effectiveness of differentiated vs. non-differentiated tax rates here. Instead, we are interested in how people perceive the different tax schemes. The introduction of an explicit meat tax can be perceived as invading people’s privacy regarding food choices. The degree of differentiating can nevertheless appear differently.

On the one hand, in case of a differentiated tax rate, people might perceive to have more influence on their actual tax payments because their freedom of choice is preserved as they can actively choose meat products with lower taxes (e.g., meat type chicken or highest animal welfare level). With a non-differentiated tax rate, they have to pay the same amount of money no matter which option they pick. This might resemble a penalization for consuming meat more generally. Preferences for more autonomy of decisions are in line with libertarian views (e.g., Thaler & Sunstein 2008 in the discussion on nudging) and psychological evidence of reactance in case of reduced or threatened freedom of choice (Brehm 1966).

On the other hand, a non-differentiated tax rate could appear to have only a revenue generation purpose for the state, while a differentiated one is supposed to interfere with people’s individual consumption behavior. In this view, the latter seems more paternalistic. German policy makers also rather implement or discuss non-differentiated taxes, such as the charge for renewable energies (“EEG-Umlage”) or the animal welfare consumption tax (“Tierwohlabgabe”). As politicians are elected representatives, they might anticipate or reflect people’s preferences for a non-differentiated tax rate. So the impact of differentiation on support rates is not clear ex ante.

Hypothesis 3a: The support rate depends on the degree of differentiation.

Finally, we expect the salience of people’s beliefs regarding the expected tax effect to impact support rates. Research on the acceptance of congestion charges, waste taxes and a carbon tax finds that trial periods increase support and people update their beliefs regarding tax schemes (see Carattini et al. 2018 for a review). We do not implement a trial period in our experiment. Instead, we use the belief elicitation task to induce subjects to actively think about what effect on consumption behavior they expect from the tax scheme both for themselves and the general population. We assume that the functioning of the differentiated tax rate is more difficult to process for people compared to a non-differentiated one. Thus, if subjects are asked to reflect upon their beliefs before they vote, they might become more aware of the effect a differentiated tax rate could actually have.

Hypothesis 3b: The dependence of the support rate on the degree of differentiation is more pronounced when the salience of behavioral responses to the tax is high.

Hypothesis 3c: If the salience of behavioral responses to the tax is high, the support rate for a differentiated tax rate increases more strongly than for a non-differentiated tax rate in a difference-in-difference comparison.
Experimental Design Details
Randomization Method
Randomization will be done by Sawtooth Software used for programming survey and experiment. Each respondent is allocated to one of the eight combinations (as described in "Experimental Design" section) based on a random draw from numbers 1 to 8.
Randomization Unit
Unit of randomization will be individual respondents.
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
3,000 individuals, design is not clustered
Sample size: planned number of observations
3,000 individuals, living in Germany
Sample size (or number of clusters) by treatment arms
(1) Flat tax for carbon emissions + belief elicitation first: 375
(2) Flat tax for carbon emissions + belief elicitation second: 375
(3) Differentiated tax for carbon emissions + belief elicitation first: 375
(4) Differentiated tax for carbon emissions + belief elicitation second: 375
(5) Flat tax for animal welfare + belief elicitation first: 375
(6) Flat tax for animal welfare + belief elicitation second: 375
(7) Differentiated tax for animal welfare + belief elicitation first: 375
(8) Differentiated tax for animal welfare + belief elicitation second: 375
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

Institutional Review Boards (IRBs)

IRB Name
Dean's Office of the Faculty of Economics and Social Sciences at University of Hamburg
IRB Approval Date
IRB Approval Number
N/A, copy of the document attached


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


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials