Introduction

Red meat consumption negatively impacts both the environment and consumers’ health. Red meat production is a major contributor to global greenhouse gas emissions, land clearing, and biodiversity loss, and red meat consumption has been found to increase the risk of several noncommunicable diseases including type 2 diabetes, cardiovascular disease, and colorectal cancer [1]. The United States is among the world’s top producers of greenhouse gas emissions and top consumers of red meat, with nearly half of Americans consuming red meat on a given day [2]. However, a recent survey found that approximately 70% of Americans are concerned about addressing global climate change [3], and another survey found that 40% of Americans were trying to reduce their red meat consumption [4], signaling that implementing population-wide interventions to reduce red meat consumption may be possible.

Public policies are a critical tool for reducing red meat production and consumption because they can affect entire populations and stimulate changes in supply and demand. The landmark 2019 EAT-Lancet Commission report noted that a full range of population-level interventions—from “soft” interventions such as consumer awareness campaigns to “hard” interventions such as fiscal measures—will be essential to transform food systems and diets to achieve targets for health and sustainability [5]. Warning labels and taxes are 2 policies that have been effectively used to reduce consumption of tobacco, sugar-sweetened beverages, and other unhealthy foods [6,7]. Modeling studies have found that meat taxes would reduce consumption and lead to reduced mortality and healthcare costs, increased quality of life, and higher productivity [8–14]. Similarly, one previous study evaluated graphic warning labels on meat products and found that a disgust-eliciting image under the guise of trivia reduced participants’ intention to consume meat compared to an explicit warning label [15]. Little experimental research, however, has examined how these policies affect purchases of red meat, and to our knowledge, no policies have been implemented to target red meat reduction.

Empirical data in this area are important because it is unclear whether taxes and warning labels on red meat would have similar effects as they have on other products like sugary drinks, tobacco, and alcohol. In the US, meat is typically consumed as the main part of a meal. Because of its more central role in the American food basket, it is possible that consumers are less likely to view red meat products as “discretionary” or “nonessential” than they would with sugary drinks, and, thus, they may be less likely to change their behavior in response to taxes or labels. On the other hand, relative to sugary drinks, where the majority of Americans are aware of common health harms associated with consumption [16], a much smaller proportion are aware of the harms associated with red meat [17]. This suggests that warning labels that educate consumers about these health and environmental harms may reduce consumption.

Health behavior and economic theory and research indicate that these policies could lead to reductions in purchases through multiple pathways. Meat taxes reduce consumption by increasing prices and, thus, reducing demand for these products [18]. Warning labels reduce consumption by increasing attention, helping people understand the healthfulness or the environmental impacts associated with the product, increasing risk perceptions about harms associated with consumption, and changing behavioral intentions, leading to reductions in selecting meat products for purchase [19].

Although prior research suggests that warnings and taxes reduce purchases of products like sugary drinks, tobacco, and alcohol, it is not clear whether these policies will have similar effects on red meat. In the US, meat is typically consumed as the main part of a meal [20]. Because of its more central role in the American food basket, it is possible that consumers are less likely to view red meat products as “discretionary” or “nonessential” than they would with sugary drinks, tobacco, or alcohol, and thus they may be less likely to change their behavior in response to taxes or labels. On the other hand, relative to sugary drinks, where the majority of Americans are aware of common health harms associated with consumption [16], a much smaller proportion are aware of the harms associated with red meat [17]. This suggests that warning labels that educate consumers about health and environmental harms may reduce consumption.

To address these gaps and inform policymaking, the main study question was, will taxes and warning labels reduce red meat purchases? To address this question, we tested the impact of taxes, warning labels, and taxes combined with warning labels on red meat purchases among US red meat consumers using a randomized controlled trial in a naturalistic online grocery store. Secondary objectives included testing the impact of these policies on nutrient purchases, psychological outcomes, and policy support.

Methods

Stimuli

We developed one health (“WARNING: Eating red meat increases your risk of colon cancer and rectal cancer”) and one environmental (“WARNING: Eating red meat harms the environment”) warning (Fig 1). The warning labels were designed based on previous research on warning efficacy [19,22] and nutrient warning labels in effect globally [23]. We tested several health and environmental warnings for red meat [24,25] and selected one for each topic (health and environment) based on perceived message effectiveness, evidence for the claims, literacy, and political feasibility through consultation with nutrition and environmental scientists and public health lawyers. The tax level of 30% was determined as a tax level likely to reduce health and environmental harms [8,10].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Warning labels used in trial.

https://doi.org/10.1371/journal.pmed.1004284.g001

Results

A total of 3,518 participants completed the shopping task (Fig 2). Sample characteristics are shown in Table 1. Most participants reported doing all of their households’ grocery shopping, and more than half the sample reported shopping for groceries online within the last month. About 85% of the sample consumed red meat 2 or more times per week. Descriptive statistics on outcomes can be found in S3 and S4 Tables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Participant flow chart (final sample analyzed = 3,518).

https://doi.org/10.1371/journal.pmed.1004284.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic characteristics by condition (n = 3,518).

https://doi.org/10.1371/journal.pmed.1004284.t001

In the control condition, an average of 39% (95% confidence interval [CI]: 38%, 40%) of the products participants selected contained red meat. Relative to the control group, participants in the warning label, tax, and combined conditions purchased a lower percent of red meat items, 36% (95% CI [35%, 37%], p = 0.001), 34% (95% CI [33%, 35%], p < 0.001), and 31% (95% CI [30%, 32%], p < 0.001), respectively (Fig 3, Panel A, and S5 Table). When comparing differences between conditions, participants in the tax condition purchased a lower percentage of products containing red meat than did people in the warning label condition (p = 0.022), and participants in the combined condition purchased a lower percentage of products containing red meat than did participants in either the warning or tax condition (p < 0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

(A) Percent of red meat items purchased during the shopping task, by condition; (B) count of red meat items purchased during shopping task, by study arm. *Asterisk denotes statistically significant differences between each intervention and the control at the 5% level. ^aShared superscript indicates significant difference between interventions at the 5% level. ^bError bars show 95% confidence interval (CI). ^cP values were calculated using fractional probit regression models for percent and Poisson regression models for count.

https://doi.org/10.1371/journal.pmed.1004284.g003

Results for count outcomes followed a similar pattern. Each condition reduced the number of items containing red meat purchased relative to the control. The average count of purchased items containing red meat in the shopping basket for the control condition was 3.5 items (95% CI [3.4 items, 3.6 items]), compared to 3.2 items (95% CI [3.0 items, 3.5 items]) in the warning label condition (p = 0.002) and 3.1 items (95% CI [2.8 items, 3.3 items]) in the tax condition (p < 0.001). The combined led to the smallest number of items containing red meat purchased (2.7 items, 95% CI [2.6 items, 2.9 items]) relative to the control, warning label, and tax conditions (p < 0.001 for all 3 comparisons) (Fig 3, Panel B, and S5 Table).

The impact of the combined condition on red meat purchases was moderated by education and age (S6 and S7 Tables). Relative to the control, the combined condition led to fewer red meat purchases for all education levels except those with a graduate degree. For example, among participants who had a high school degree, relative to the control, the combined condition led to an 8.7 percentage point reduction in red meat purchases (95% CI [−11.6 percentage points, −5.7 percentage points]). In contrast, among those with a graduate degree, the combined condition had no impact on red meat purchases. Additionally, relative to the control, the combined condition led to larger reductions in red meat purchases for the youngest age group (−11.0 percentage points, 95% CI [−13.8 percentage points, −8.3 percentage points]) than for the oldest age group (−5.0 percentage points, 95% CI [−8.2 percentage points, −1.8 percentage points]). There was no evidence of moderation for the other conditions or by other variables, and the pattern of moderation results for count of red meat purchases was similar (S6 and S7 Tables).

The conditions also affected secondary behavioral outcomes, with effects varying by nutrient (Table 2). Relative to the control, the combined condition led to fewer calories purchased (-311.9 calories, 95% CI [-589.1 calories, -34.7 calories], p = 0.027). Both the tax and combined conditions led to lower saturated fat purchased compared to control (p = 0.01 and p = 0.001, respectively), as well as reductions in the total weight of products that contain red meat (−321 g, 95% CI [−399 g, −242 g] and −421 g, 95% CI [−501 g, −342 g], respectively, from a baseline of 1,603 g in the control condition). There were no significant impacts on sodium purchased.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Impact of warning label, tax, and combined warning label and tax on calories, saturated fat, and sodium purchased (n = 3,518).

https://doi.org/10.1371/journal.pmed.1004284.t002

Of the 3,518 participants who completed the shopping task (i.e., checked out and did not time out of the shopping task), only 9 did not respond to the post-shopping task questionnaire. Average total spending was $31.68 for those who responded to the questionnaire and $32.09 for those who did not, suggesting similar shopping behaviors between the 2 groups. Conditions that included a warning (i.e., the warning and combined conditions) decreased the perceived healthfulness and environmental sustainability of red meat. By contrast, policies that included a tax (i.e., the tax condition and combined condition) increased perceived cost (Table 3). No condition affected intentions to reduce red meat consumption in the next month.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Impact of warning label, tax, and combined warning label and tax on psychological outcomes ^a.

https://doi.org/10.1371/journal.pmed.1004284.t003

In the control condition, policy support was low for taxes (21.4% of participants in favor and 55.7% against) and higher for health and environmental warning labels on red meat (42.1% and 40.2% of participants in favor and 28.5% and 32.2% against, respectively). Exposure to the tax led to lower support for taxes and higher opposition to taxes. Exposure to the taxes also led to lower support for environmental warning labels and increased opposition to health warning labels. In contrast, exposure to warning labels led to increased support for health and environmental warning labels; there was no effect on support for taxes or opposition for taxes or labels (S8 Table).

Process measures indicated the store was acceptable to participants. The majority reported that the online experimental store was easy to use, they could easily find all food and beverage items they were looking for, there were enough options, the store felt like a real online grocery store, and the items they selected in the experimental store were similar to their usual grocery purchases (S1 Fig).

Discussion

In a randomized trial, we found that both warning labels and a tax led to modest reductions in purchases of items containing red meat. The largest impact came from combining both warnings and the tax, resulting in an approximately 21% to 26% relative reduction of red meat purchases across outcomes. This reduction in purchases of items containing red meat translated into a reduction in calories and saturated fat of the shopping basket, but not sodium.

Our findings add to the growing body of evidence that taxes and warning labels applied to foods reduce purchases. Although limited research has explored the impact of these policies in the context of environmentally harmful products, our results are in line with nascent evidence that environment-focused fiscal and labelling policies may shift food purchasing behaviors. For example, a recent study by Wolfson and colleagues that found that, among US consumers, climate impact labels—and particularly negatively framed “high-impact” labels—led to an increase in sustainable selections from a fast food menu [30]. Our results showed that taxes also reduced red meat purchases and that these reductions were larger than for warnings, albeit by a small margin. These results are in line with a choice experiment conducted in Australia, which reported that price changes had a larger effect on selection of healthy and sustainable food options than did labels or logos [31]. These results are also in line with economic theory, which suggests that educational interventions alone will be less likely to lead to meaningful reductions in red meat consumption than a tax [32].

Warnings and taxes alone led to small reductions in red meat purchases, and the combined label and tax condition had the largest impact on red meat purchases. These results are consistent with a recent United Kingdom–based study in a simulated lunchtime canteen, which also found that labeling and taxes were most effective at shifting choice to meals with a lower carbon footprint when used in combination [33]. A Canadian study of sugar reduction policies also found additive effects on sugar purchases when taxes and front-of-package labels were implemented together [34]. One explanation is that these policies operate through distinct and complementary pathways: Warning labels can inform consumers and increase risk appraisals, whereas taxes incentivize consumers to select lower-cost (untaxed) items. Indeed, we also found that warning labels changed perceptions of the healthfulness and sustainability of red meat, whereas taxes influenced perceptions of cost. Together, these results suggest that the US could achieve the greatest public health and environmental impact by implementing warnings and taxes together.

In addition to reducing red meat purchases, warning labels and taxes also reduced purchases of nutrients. Specifically, the combined tax and warning label condition led to reductions in both total energy and saturated fat, while taxes reduced saturated fat, and no condition impacted sodium. These results highlight the importance of understanding how these interventions affect substitutions across the food basket and ultimately influence the overall nutritional quality of diets. Moreover, a better understanding of substitutions is needed, as net health and environmental effects of policy actions to reduce red meat will vary depending on whether participants substituted to poultry, a plant-based meat alternative, beans, vegetables, or another food. Indeed, a recent simulation study in New Zealand found that while all meat replacement diets were associated with health gains and greenhouse gas reductions, replacement with minimally processed plant-based foods led to greatest benefits for all outcomes relative to ultraprocessed plant food alternatives, cellular meats, or diets based on other recommendations (e.g., EAT-Lancet) [35]. Future research will be critical for understanding how to design policies to not only discourage red meat intake but also promote healthy and sustainable substitutions.

We found that policies may have a differential impact for some subpopulations. For example, the combined tax and warning label condition led to a larger reduction in red meat purchases for younger versus older adults. One possibility is that younger people have greater knowledge of and concern about sustainability issues with regard to food [36,37] and thus are more responsive to policies. We also found some evidence of higher impact in the combined condition for lower-educated populations compared to those with a graduate degree. A recent study found that higher-educated US adults were more likely to report efforts to reduce red meat consumption [4], suggesting those with a graduate degree are already making efforts to reduce and may therefore be less likely to be further influenced by red meat policies. Other research has shown that education is associated with baseline knowledge about the health and environmental harms of eating red meat [17]; thus, policies such as warning labels, which seek to increase awareness, may have little impact among highly educated groups. The findings of a greater potential impact among participants with lower education suggests that such policies could promote health equity by leading to greater dietary change in lower educated groups, who tend to suffer a disproportionate burden of diet-related diseases, including cardiovascular disease [38–40] and cancer [41,42].

This study used both health and environmental warnings in the labeling policy condition, so we were unable to disentangle the effects of the 2 types of warnings. Our previous research has found that health warnings and “combined” warnings that include both health and environmental messages are perceived as more effective than warnings that include only environmental messages [24,25] suggesting that the health warning may have been responsible for most of the observed reductions in red meat purchases in this study. In contrast, a study in the UK found that nutrition labels (Nutriscore) did not impact the healthfulness of purchases in an online experimental store but that environmentally focused “ecolabels,” either alone or in combination with nutrition labels, improved the sustainability of purchases [43]. Future research should examine whether health and environmental warnings have a differential impact on purchasing behaviors, and whether either type of warning is more likely to lead to reactance (message rejection) or label fatigue.

This study has strengths and limitations. One strength was the use of a naturalistic online grocery store designed to mimic a real-world US retailer. The use of these online stores provides a major advantage in that participants can choose from a wide assortment of real products in a setting that looks and feels like a real store. A recent validation study found that expenditures in the naturalistic online grocery store used in this study were moderately-to-strongly correlated with expenditures made in a real online grocery store and that 95% of participants reported making similar purchases in the naturalistic online store as they did in a real online store [44]. However, in this study, it is unclear the extent to which aspects of this experiment (e.g., the use of a prespecified shopping list, lack of financial transactions) may have altered purchasing behaviors and attenuated or amplified the impact of the policies. For example, it is possible that in the real world, where habits, preferences, and other concomitant secular trends could influence consumer response, the effects of the policies could be attenuated. Or, real-world effects could be stronger because people are spending real money.

In addition, our study sample differed in sociodemographic characteristics from the US population (e.g., our sample included somewhat more women and respondents who identified as non-Hispanic white), though it is unclear whether this was due to the online panel, the fact that we included only meat consumers, or both. A related limitation is that, while online food retail is growing, only about 30% to 40% of Americans shop for groceries online [45,46]. It is not clear whether online food shopping behavior differs from in-person shopping behaviors, and, thus, it is unclear how generalizable our results are to consumers who do not shop online. Future research could address these limitations by testing these policies in a real-world brick-and-mortar retail setting where participants use their own shopping lists, spend their own money, and can take home products for consumption.

This study has several public health and policy implications. The results show that taxes and warning labels reduce red meat purchases among red meat consumers, suggesting that adoption and implementation of these policies could yield both health and environmental co-benefits. However, the likelihood of policy adoption is unclear. While several countries in Europe have recently considered adopting meat taxes [47–49], policy discussions on red meat in the US have been limited. A recent legal analysis found that while public policies to encourage meat reduction are legally feasible in the US, political feasibility is highly uncertain given the economic and political power of meat industry lobbies [50]. Policymakers may wish to consider the potential benefits to consumers, along with the legal and political feasibility, when weighing regulatory options for reducing intake of red and processed meat. Another barrier to policy adoption is public support: In this study, overall support for both policies was low (<50% of participants), and similar to a previous study [51], support was lower for taxation than labeling. However, some evidence suggests that this barrier could be overcome through different types of messaging about the policy. For example, one study in Germany found that consumers were more likely to support a meat tax if it was justified by animal welfare than by climate change mitigation [52]. Research will be necessary to understand what type of framing or justification is likely to elicit higher support for meat reduction policies in the US.

Conclusions

In a randomized controlled trial using a naturalistic online grocery store, we found that taxes and warning labels reduced purchases of red meat purchases as well as calories and saturated fat. The condition that included both taxes and warning labels had the biggest impact, suggesting this combined approach would achieve the greatest co-benefits for health and environment.

Supporting information

Acknowledgments

The authors thank Pasquale Rummo for assisting with the development of the online grocery store. The authors thank David Gonzalez, Jamie Halula, Sophia Hurr, Keilei Latragna, Rhea Naik, and Carla Seet for their help with cleaning the dataset and reviewing the online grocery store. They would also like to thank Dhara Buebel for assisting with the manuscript preparation, Emily Busey for designing the online grocery store logo, Christina Chauvenet for assistance with project management, Bridget Hollingsworth for providing guidance on tagging red meat products, Yiqing Zhang for providing expertise on the data scraping process, and Wilma Waterlander for providing general feedback for the online store development process. Finally, the authors thank Susan Jebb and Brian Cook for contributing their expertise from the development of the Woods supermarket platform.