Claiming confidence: U.S. consumer trust in meat and poultry claims

Abstract

In the fall of 2024, the Food Safety and Inspection Service updated its guidelines for verifying animal-raising and environment-related claims used on meat and poultry products. Notably, revisions strongly emphasized the inclusion of third-party certifications in the supporting documents. Recent inquiries into claim validity, however, have uncovered improper approvals and compliance issues, prompting increased concerns regarding consumer confidence. This study examined consumer trust across nine claim categories explicitly defined within the guidelines. Data were collected from 1,485 respondents through a web-based survey. A best–worst scaling approach identified the most and least trusted claim categories; these data were analyzed for the entire sample and stratified by household shopping roles and meat consumption frequency. Additional questions assessed knowledge of and trust in certifying agencies (eg, U.S. Department of Agriculture [USDA]) and third-party certifiers. The results reveal varying levels of trust, with consumers trusting breed and source/traceability claims most, whereas trust in animal welfare and environment-related claims was lowest. Nonetheless, all nine claim categories received generally low levels of trust. Household food-shopping roles and meat-consumption frequency had little impact on trust. Findings also show that consumers primarily trust the USDA to certify claims, and third-party certifications can enhance trust. This study offers new, quantified insights into consumer trust in food claims for meat and poultry products. While addressing trust issues, the results can guide efforts to improve consumer confidence in animal-raising and environment-related claims through targeted interventions and better labeling practices.

Consumer trust across all animal-raising and environment-related claims is relatively low. When forced to make trade-offs, consumers, on average, trust Breed and Source/Traceability claims most; however, Animal Welfare claims are least trusted.

Introduction

In August 2024, the United States Department of Agriculture’s (USDA) Food Safety and Inspection Service (FSIS) released updated guidelines for verifying animal-raising or environment-related labeling claims for meat, poultry, and egg products. The guideline revision was part of a broader effort focused on strengthening the documentation of animal-raising claims, announced by the USDA in June 2023 (USDA 2023). The update replaces the 2019 document, making both minor and major revisions to the guidelines (USDA 2024a). The new document aims to improve guidance for verifying claims and to strengthen the recommended documentation (USDA 2024b).

The release of the updated guidelines follows growing concerns about the FSIS’s handling of livestock claims. Recent investigations into the FSIS and its approval process highlighted the seriousness of these issues. A 2020 audit by the USDA Office of Inspector General (2020) examined FSIS’s controls over meat, poultry, and egg product labels and found that “…meat, poultry, and egg product labels may reflect inaccurate statements and claims made by establishments.” The inquiry examined application files for two types of label submissions, namely labels requiring formal review and generic labels submitted voluntarily. The audit revealed that, among the approved label applications reviewed, 15% of the required label applications and 18.3% of the generic label applications were inadequate for approval. Similarly, a study by the Animal Welfare Institute (2023) requested application files for 97 animal-raising claims and found that 85% lacked sufficient evidence for approval, with FSIS unable to provide nearly half of the applications. Compliance issues have intensified concerns about the validity of animal-raising claims. A USDA study, conducted by FSIS in partnership with the USDA’s Agricultural Research Service (ARS), evaluated the accuracy of negative antibiotic use claims by testing liver and kidney samples from cattle across 34 states. The findings showed that 20% of samples intended for the “raised without antibiotics” market tested positive for antibiotic residues (USDA 2024b).

Major updates to the guidelines included updated documentation for substantiating claims related to living or raising conditions, such as “pasture fed” and “meadow raised” (USDA 2024c). In response to concerns about negative antibiotic use, strong recommendations were introduced for routine sampling programs to test antibiotic residues. Similar suggestions were made for environment-related claims, encouraging the use of scientific data or studies to support them (USDA 2024b). Minor changes included examples of diet claims and a more detailed discussion of the use of animal-raising and environment-related claims on multi-ingredient products. Additionally, further clarification was provided about the use of “vegetarian-fed” and “no animal byproducts” claims. To make the guidelines more specific, the revision separated claims about animal welfare and environmental stewardship into separate sections (USDA 2024c).

The most notable addition to the guidelines was the strong emphasis on using third-party certification to substantiate animal-raising and environmental-related claims (USDA 2024c). As explicitly stated in the document, certifying third parties must be independent of the establishment seeking certification and be evaluated by FSIS to ensure suitability for the respective claim. Additionally, producers may also use USDA audit-based programs, such as the Process Verified Program administered by the Agricultural Marketing Service, to ensure products comply with standards. While not required for approval, the frequent and consistent encouragement to include third-party certification in supporting documents suggests that doing so would increase the likelihood of claim approval. According to the guideline, this emphasis on third-party certification comes from the FSIS’s limited jurisdiction (USDA 2024c). Using independent third-party certification would also better protect consumers from misleading or false claims by ensuring that establishments adhere to proper on-farm practices.

Given recent FSIS guideline updates and the increasing controversy surrounding animal raising and environmental-related labeling claims, this research assesses consumer trust in claim categories for meat and poultry products. Consumer trust is measured through survey-based assessments, and comparative analyses are used to identify differences in trust among contextually different claims. Additionally, we further disaggregate the data into consumer segments defined by household food-shopping roles and meat-consumption frequency, and within-group comparisons are conducted to evaluate potential associations between consumer characteristics and trust levels. Finally, we examine consumer knowledge of and trust in certifying entities, as well as the impact of third-party certification on consumer trust, to determine familiarity with the claim approval process and test for underlying factors influencing trust. This study provides quantified insights into consumer trust in animal raising and environmental-related labeling claims to assist policymakers and industry stakeholders in maintaining strong relationships with their consumer base.

Background on consumer trust in labeling claims

Understanding consumer trust in labeling claims is critical because it influences purchasing decisions, market demand, and perceptions about the credibility of both government and, potentially, producers. If trust in labeling claims erodes, consumers may be less willing to pay a premium for credence attributes, thereby reducing the value of differentiated products. Moreover, inaccurate labeling by a single entity can undermine trust across the entire market, affecting all producers. By assessing how trust varies across market segments, this study aims to provide insight into where economic losses may arise due to low trust, misinformation, or a lack of certification clarity.

A substantial body of research on food labeling has frequently examined consumer perceptions (Van Kleef et al. 2005; Benson et al. 2018; Franco-Arellano et al. 2020) and how labels influence purchasing behaviors (McFadden and Huffman 2017; McFadden and Lusk 2018; Tønnesen et al. 2022; Gorton et al. 2023). Consumer trust in food products and various food actors largely depends on belief in the integrity of the entity (Macready et al. 2020; Wu et al. 2021), and the source communicating product quality can also significantly influence consumer confidence in food labels (Wu et al. 2021). Recent research revealed that only 24% of U.S. adults have a high level of trust in the information provided about food sourcing and production (University of Minnesota 2022), and confidence in food safety among Americans dropped to 62%, down from 70% in 2022 (International Food Information Council 2024). Food fraud issues can undermine trust in both the food supply and the regulatory agencies responsible for oversight (Yiannaka 2023); however, consumer confusion about food-related claims can also affect trust.

Other research has contributed to the literature by assessing consumer knowledge of food-related claims (Williams 2005; Wong et al. 2013). Many of these studies have shown that consumers are often confused about or misinterpret the true meaning of claims. For instance, consumers have displayed confusion about the differences between “natural” and “organic” claims (Abrams et al. 2010; Kuchler et al. 2020). Similar conclusions have been reached regarding animal welfare claims. Although consumers show a growing preference for better animal welfare practices and welfare labels (Ortega and Wolf 2018; Ochs et al. 2019; Thibault et al. 2022), there is also confusion about what these claims mean and which practices they represent (Lim and Page 2022). For example, consumers are more willing to pay for the “no added hormones” claim, even for meat products from species where the use of hormones is banned by law (Yang et al. 2020). One study found that only 47% of respondents are at least somewhat trusting of animal welfare certifications, while 16% are somewhat or completely distrustful (Lusk and Polzin 2023). While animal treatment labels have attracted increased public scrutiny, consumers may be willing to pay a premium for animal-welfare labels when the claim is verified by a trusted organization (Spain et al. 2018).

Consumer confidence has been measured more directly for more commonly used claims, such as “organic” and “natural.” For example, consumers have shown skepticism or distrust toward “natural” (Abrams et al. 2010), “health” (Lusk and Polzin 2023), and “fat” claims (Chan et al. 2005). This skepticism can negatively affect consumer evaluations of food products (Fenko et al. 2016). However, consumer trust in both labels and claims may largely depend on the perceived reliability of the certifying organization (Janssen and Hamm 2012; Rupprecht et al. 2020; Wang et al. 2020).

Materials and methods

Data collection

This study received approval from the University of Arkansas Institutional Review Board (protocol #: 2411570610). The survey was created in online software developed by Qualtrics and distributed to a panel of 1,538 respondents in December 2024. All respondents were recruited through the data collection platform Prolific. Research shows that Prolific maintains a reliable pool of verified respondents who exhibit high attentiveness and completion rates, as well as low dishonesty; as a result, respondents recruited via Prolific tend to provide high-quality data (Peer et al. 2017; Douglas et al. 2023). The target sample size was 1,500 respondents, which, given the U.S. population, provides a +/−3% margin of error at a 99% confidence level. Ultimately, 1,538 responses were received. To ensure all respondents had baseline familiarity with meat products, a screening question asked how often they eat meat with meals. Of the initial sample, 47 observations did not consume meat and were excluded. An additional six observations were omitted due to incomplete responses or failing attention check questions. Thus, there were 1,485 usable responses.

After giving informed consent and answering a qualifying question about meat consumption, respondents completed best-worst scaling (BWS) questions to assess which claim categories were the most and least trusted. As a secondary measure of trust, respondents were also asked how often they believed each claim category was valid when shopping. Additionally, questions were included to evaluate knowledge about and trust in certifying agencies and to assess whether third-party certification would improve trust in meat and poultry claims. Respondents completed the survey by providing information on demographic characteristics, including their role in household food shopping. More details about these questions are provided in the subsections below.

Trust in livestock claims questions

A BWS approach was first used to examine consumer trust in animal-raising and environment-related labeling claims, as described by Finn and Louviere (1992) and later refined by Marley and Louviere (2005). This version of choice modeling is commonly used to elicit preferences and attribute rankings (Lusk and Briggeman 2009; Caputo and Lusk 2020; Luke and Tonsor 2025) and has become increasingly popular in recent years (Hollin et al. 2022; Schuster et al. 2024). In a typical BWS approach, respondents see a set of alternatives and are asked to select their most preferred (“best”) and least preferred (“worst”) options from that set. This process is often repeated with different sets of alternatives across several choice sets. BWS is useful because it extends ranking analysis by capturing both the best and worst choices. These two extreme selections provide more detailed insights into consumer preferences than just choosing the top option alone (Louviere et al. 2013).

The BWS in this study used nine livestock claim categories as alternatives. Respondents were provided with both the claim category, and an example claim for clarity. The nine claim categories and example claims included in this study were drawn directly from the FSIS guidelines (USDA 2024c). The specific claim categories and example claim used in the BWS are shown in Table 1.

Table 1.

Claim categories and example claims presented in the best–worst scaling.

Claim category	Example claim
Animal welfare	“Humanly raised”
Breed	“Certified angus beef”
Diet	“Grain fed”
Environment-related	“Sustainably farmed”
Living or raising conditions	“Cage free”
Negative antibiotic use	“No added antibiotics”
Negative hormone use	“Raised without added hormones”
Organic	“Organically raised”
Source or traceability	“Source verified and traceable to [Name of Farm Origin]”

A process analogous to a balanced incomplete block design (BIBD) was used to develop the BWS choice sets. This design is commonly used in BWS (Jin et al. 2019; Caputo and Lusk 2020; Luke and Tonsor 2025), and some evidence suggests it is the most frequently used (Hollin et al. 2022). The BIBD design framework uses incomplete choice sets because a subset of alternatives is shown in each question, yet the design ensures balance in the number of times an alternative appears across choice sets and co-appears with other alternatives across choice sets (Louviere et al. 2013).

In this study, respondents were presented with 12 choice sets, each containing three claims. Thus, each of the nine claim categories appeared four times and co-appeared with the other eight claims once. Respondents were tasked with selecting the claim category they “trust the most” and “trust the least” from the three claims listed within each choice set. The order in which claims appeared within a choice set was randomized to reduce order effects. An example of a BWS question is shown in Appendix Figure S1 (see online supplementary material for a color version of this figure).

As a secondary measure of trust across the nine claim categories, respondents were asked to estimate how often they believe a claim is valid while shopping. Response options provided frequency ranges in 25% increments, from “0% to 25% of the time” up to “76% to 100% of the time.” This was asked through three questions, each featuring three claims, to minimize straightlining, with the order of claims within each question randomized. While the BWS approach forces respondents to indicate relative trust in claims, this secondary measure allows respondents to select similar levels of trust.

Consumer trust in certifying entities questions

Respondents were asked questions to assess their knowledge about the agencies that certify claims and their trust in these agencies. The first question asked respondents to identify which entities certify claims on meat and poultry products. Response options included the following: the food company, the United States Department of Agriculture (USDA), the United States Food and Drug Administration (FDA), the United States Centers for Disease Control and Prevention (CDC), a third-party certifier, and an “I am not sure” option was also provided. Respondents could select multiple options, except if “I am not sure” was selected. Immediately afterward, respondents were asked to identify the most trusted entity to certify claims on meat and poultry product labeling. The same entities from the previous question were provided as response options; however, the “I am not sure” option was omitted.

Given the FSIS’s emphasis on using third parties to certify claims, additional questions were asked specifically about whether third-party certification would increase trust for each of the nine claim categories. Response options were a simple “Yes” or “No.” As with the secondary trust questions, the claim categories were split into three questions to reduce straightlining, and the order of claims was randomized to reduce order effects.

Analysis of BWS data

We begin with a descriptive analysis of the trust in claim categories from the BWS data. As Louviere et al. (2013) explained, a BWS score is the average difference between how many times a claim is chosen as most trusted and least trusted by respondents. BWS scores could range from −4 to +4 in this study because each claim category appeared four times across the 12 BWS questions. BWS scores can then be standardized by dividing each score by the number of times it appeared (ie, each claim appeared four times), which provides a clear visual summary of BWS response patterns.

To provide a more advanced analysis of the BWS data, we combine two frameworks outlined by Islam et al. (2023) and Lusk and Briggeman (2009). These methods are based on the Random Utility Model (RUM), which assumes that individuals choose an alternative from a set that maximizes utility (McFadden 1972). Formally, indirect utility can be defined as $U_{\mathit{njs}}=V_{\mathit{njs}}+{\epsilon }_{\mathit{njs}}$, where $U_{\mathit{njs}}$ denotes the utility that the n^th respondent receives from selecting alternative j from choice set s, and ${\epsilon }_{\mathit{njs}}$ is an error term representing the unobservable component of utility. A BWS Case 1, the design used in this study, allows for the claims to be included as single regression parameters when modeling the observable portion of utility, which can be specified as:

$$V_{\mathit{njs}}={\beta }_{1}B\mathit{reed}+\mathrm{ }{\beta }_{2}D\mathit{iet}+{\beta }_{3}L\mathit{ive}\mathrm{ }or\mathrm{ }\mathit{Raising}\mathrm{ }\mathit{Conditions}\hspace{1em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+{\beta }_{4}N\mathit{egative}\mathrm{ }\mathit{Antibiotic}\mathrm{ }\mathit{Use}+{\beta }_{5}N\mathit{egative}\mathrm{ }\mathit{Hormo}\mathrm{n}e\mathrm{ }\mathit{Use}+{\beta }_{6}S\mathit{ource}\mathrm{ }or\mathrm{ }\mathit{Traceability}+{\beta }_{7}O\mathit{rganic}+{\beta }_{8}E\mathit{nvironment}\mathrm{ }\mathit{Related}.$$ (1)

The explanatory variables are indicator variables for the claim categories (eg, Breed), and ${\beta }_k$ are the coefficients to be estimated for the $k^{th}$ claim category. The Animal Welfare claim category was used as the base for estimation.

Assuming the unobserved utility component is independently and identically distributed (IID) following a type 1 extreme value, the likelihood of choosing claim $i$ as best and claim $k$ as worst takes the form of the multinomial logit (MNL) equation:

$$\mathit{Pr}\left(i=\mathit{most}\mathrm{ }\mathit{trusted},\mathrm{ }k=\mathit{least}\mathrm{ }\mathit{trusted}\right)=\frac{\hspace{0.17em}\text{exp}\hspace{0.17em}\left(V_i-V_k\right)}{{\sum }_{x,y|x\ne y}\exp \left(V_x-V_y\right)},$$ (2)

where $Pr$ represents the joint probability of selecting claim categories $i$and $k$ as the most and least trusted, respectively, given the choice set. Consistent with Islam et al. (2023), the numerator represents the difference in utility of the most and least trusted selections, while the denominator sums over all possible selection combinations within the choice set.

The MNL model assumes preferences are the same across a given sample, and the assumption of independence of irrelevant alternatives (IIA) holds. Using the approach defined by Lusk and Briggeman (2009), we account for preference heterogeneity across respondents by incorporating a random parameter logit (RPL) model:

$${\stackrel{\sim }{V}}_{nj}={\overline{V}}_j+{\sigma }_j{\mu }_{nj},$$ (3)

where ${\stackrel{\sim }{V}}_{nj}$ denotes the utility received from the level of trust placed in category $j$ by respondent $n$, ${\overline{V}}_j$ and ${\sigma }_j$ denote the mean and standard deviation of $V_j$, and ${\mu }_{nj}$ represents the individual-specific random error term that is normally distributed with a mean of 0 and standard deviation of 1. Consistent with Lusk and Briggeman (2009), we substitute equation 3 into equation 2 and estimate the RPL using simulated maximum likelihood estimation. All models were estimated in Stata 17 using the cmxtmixlogit command, with 500 Hammersley draws and default starting values.

RPL estimates are difficult to interpret directly (Caputo and Lusk 2020; Luke and Tonsor 2025). Therefore, preference shares, or what we call trust shares, are estimated for the nine claim categories, similar to Lusk and Briggeman (2009). The trust share$$of the $j$^th claim category is calculated as:

$$S_j=\frac{\hspace{0.17em}\text{exp}\hspace{0.17em}({\widehat{V}}_j)}{{\sum }_{k=1}^J\hspace{0.17em}\text{exp}\hspace{0.17em}\left({\widehat{V}}_k\right)},$$ (4)

which can be interpreted as the percentage of the population expected to select that claim as the most trusted (Luke and Tonsor 2025). Direct comparisons can be made between trust shares; for example, a claim category with a trust share three times the value of another can be said to be trusted three times more (Lusk and Briggeman 2009).

Trust shares are nonlinear functions of the RPL estimates, similar to elasticities discussed by Krinsky and Robb (1986). The Krinsky-Robb simulation method can be used to estimate trust shares and confidence intervals from RPL estimates using BWS data (Bir et al. 2019; Liu et al. 2025). We implemented this method by using 1,000 random draws from the multivariate normal distribution of the RPL estimates and the corresponding variance-covariance matrix. Trust shares and differences in trust shares were calculated for each draw, generating an empirical distribution. Then, we formally tested differences across trust shares by comparing the proportion of simulation draws in which one share exceeded another; this approach is similar to how Poe et al. (2005) tested differences in the empirical distributions of willingness-to-pay, which is also a nonlinear function. Using the Poe test to examine differences across all trust shares requires 36 multiple comparisons; thus, we adjusted the P-values using a Bonferroni correction to control the family-wise error rate at 5%.

Estimation for the secondary trust measure

The response options for the secondary measure of trust, which asked respondents how often they believed a claim was valid when shopping, were frequency ranges in 25% increments. Frequency distributions for each claim category can be reported; however, average belief levels for each claim category can also be estimated using interval regression (IR) models. Thus, an IR model was estimated for each of the nine claim categories, which can be generally specified by:

$${LB}_{nj}^{\mathrm{*}}={\beta }_{0j}+{\epsilon }_{nj},$$ (5)

where ${LB}_{nj}^{*}$ is the latent level of belief provided by the n^th respondent for claim category j, ${\beta }_{0j}$ represents the estimated average belief level for category j, and ${\epsilon }_{nj}$ is a normally distributed error term.

Stratification analysis

To examine differences among consumer groups, trust shares were categorized by household shopping role and meat consumption frequency. The household shopping role questions offered three response options: more than half, around half, and less than half. The meat consumption frequency question had five options: daily, several times per week, once a week, several times a month, once a month, and a few times per year. Most respondents ate meat daily or several times per week; therefore, those who ate meat once a week or less were grouped together. As a result, there were three groups based on household shopping role and three based on meat consumption frequency.

After consumer groups were segmented, six additional RPL models were estimated to simulate trust shares, as in equation 4 for the pooled model. Poe tests were conducted to examine within-group differences in trust shares, using Bonferroni-adjusted P-values to account for the 36 comparisons. Between-group differences were examined using simulated 95% confidence intervals, as in Liu et al. (2025).

Results

Summary statistics for respondent characteristics, including the variables used for stratification analyses, are available in Appendix Table S1 (see online supplementary material). Around 73% of respondents reported being responsible for more than half of the household food shopping, 19% were responsible for around half, and 7% were responsible for less than half. Nearly half (49%) of respondents reported eating meat daily, approximately 40% ate meat several times per week, and the remaining 11% ate meat at least once a week or less.

Figure 1 shows the standardized BWS scores. Breed is, on average, the most trusted category with a standardized BWS score of 0.40, followed by Source/Traceability (score = 0.28) and Diet (score = 0.08) claims. These were the only three categories with positive standardized BWS scores. The other six categories all have negative standardized BWS scores; however, the score for Organic is close to zero (score = −0.01). The Organic claim was followed by Negative Antibiotic Use (score = −0.03), Negative Hormone Use (score = −0.08), Living/Raising Conditions (score = −0.09), Environment-Related (score = −0.13), and Animal Welfare (score = −0.42) claims.

Figure 1.

The standardized best–worst scaling (BWS) scores for meat and poultry claim categories. Note: Scores reflecting individual-level trust in food labeling claims are calculated as the difference between the most (+1) and least (−1) trusted selections per respondent. Each claim appeared four times, yielding raw scores from −4 to +4, which are standardized by the number of appearances and averaged across respondents.

Estimates for the pooled RPL model and resulting trust shares are reported in Table 2. Recall that the Animal Welfare claim category was used as the base in estimation and normalized at zero. Given that the Animal Welfare claim was determined to be the least trusted category in the standardized BWS scores, it is unsurprising that all estimated coefficients are positive and significant at a P < 0.01.

Table 2.

Random parameter logit model estimates and trust shares for meat and poultry claim categories.

Claims category	RPL estimates		Trust share (%)
	Mean coefficient	Standard deviation
Breed	2.793***	2.759***	32.4a
	(0.117)	(0.102)	[29.6, 35.2]
Source or Traceability	2.231***	2.326***	22.3b
	(0.101)	(0.083)	[19.9, 24.8]
Diet	1.455***	1.534***	9.5c
	(0.076)	(0.059)	[8.3, 11.0]
Organic	1.154***	1.803***	9.0c
	(0.075)	(0.069)	[7.7, 10.4]
Negative Antibiotic Use	1.004***	1.693***	7.6cd
	(0.076)	(0.068)	[6.6, 8.8]
Negative Hormone Use	0.969***	1.657***	7.2cd
	(0.075)	(0.069)	[6.2, 8.4]
Living or Raising Conditions	0.944***	1.449***	6.1de
	(0.067)	(0.063)	[5.1, 7.1]
Environment-Related	0.794***	1.315***	4.8e
	(0.060)	(0.055)	[4.1, 5.6]
Animal Welfare	0.000	0.000	1.2f
			[1.0, 1.3]
Number of observations	1,485
Pseudo log-likelihood	−28,688

Note: Trust shares can be interpreted as the percentage of the population expected to select the respective claim as most trusted. The Animal Welfare category was used as the base in the estimation and normalized at zero. Robust standard errors are reported in parentheses.

^***Denotes statistical significance at a P-value less than 0.01. 95% confidence intervals for trust shares are reported in brackets. Mean trust shares are statistically different at P < 0.05 unless followed by a letter in common.

The simulated trust shares and 95% confidence intervals are reported in the last column of Table 2. More than half of trust shares are associated with either Breed (32%) or Source/Traceability (22%) claims. Next, Diet (10%), Organic (9%), Negative Antibiotic Use (8%), and Negative Hormone Use (7%) carried moderate shares. Living/Raising Conditions (6%), Environment-Related (5%), and Animal Welfare (1%) claims have the lowest trust shares. The letters next to trust shares indicate the multiple-comparison results; shares that share a letter are not significantly different at P < 0.05. Notably, Breed was the most trusted claim, and Animal Welfare was the least trusted claim. Source/Traceability was the second most trusted claim, and the trust shares for the remaining claims overlapped with at least one other claim.

Trust shares across the stratified consumer groups are reported in Table 3. RPL model estimates for the different consumer groups are presented in Appendix Tables S2 and S3 (see online supplementary material); the standard deviations associated with all random parameters are statistically significant, suggesting meaningful heterogeneity in trust across respondents and indicating that the assumption of homogeneous preferences inherent in the MNL model is not supported by the data. The stratified results generally match those from the pooled model, but there are some differences. For example, Source/Traceability had a higher trust share than Breed for the Less than Half group in the shopping role stratification. The letters next to trust shares indicate the within-group multiple-comparison results. An interesting finding from the within-group results is that there is less overlap in trust shares for the More than Half group in the shaping role stratification than for the other groups. For example, there is not a difference, at P < 0.05, between the trust shares for Breed and Source/Traceability for any groups except the More than Half group.

Table 3.

Trust shares for meat and poultry claim categories segmented by household food shopping role and frequency of meat consumption consumer groups.

Claim category	Trust shares by household food shopping role (%)			Trust shares by meat consumption frequency (%)
	More than half	Around half	Less than half	Daily	Several times per week	Once a week or less
Breed	32.6a	30.0a	25.6a	33.2a	29.9a	30.2a
	[29.6, 35.6]	[22.8, 37.0]	[17.0, 35.0]	[29.1, 37.4]	[25.7, 34.9]	[23.7, 36.3]
Source or Traceability	21.2b	21.9a	28.2ab	23.9a	22.7a	19.1ab
	[18.6, 24.0]	[17.6, 26.6]	[10.1, 44.3]	[20.4, 27.5]	[19.4, 26.2]	[14.8, 23.8]
Diet	10.3c	9.5b	11.7ab	8.9b	10.0b	8.4c
	[8.8, 11.8]	[7.1, 12.4]	[5.0, 21.3]	[7.3, 10.9]	[8.2, 12.1]	[6.2, 11.0]
Organic	9.1cd	10.1b	5.5b	8.7b	9.3bc	10.0bc
	[7.5, 10.9]	[7.0, 13.8]	[2.9, 9.2]	[7.0, 10.5]	[7.6, 11.4]	[7.2, 13.1]
Negative antibiotic use	7.7 cd	7.8b	8.3b	7.7b	8.1bc	9.8bc
	[6.5, 9.0]	[5.8, 10.2]	[4.0, 13.6]	[6.1, 9.6]	[6.3, 10.0]	[7.3, 12.5]
Negative Hormone Use	6.6d	5.4b	8.4b	6.7bc	6.4bc	7.0c
	[5.5, 7.8]	[3.8, 7.5]	[3.7, 14.3]	[5.4, 8.3]	[4.9, 8.2]	[4.6, 10.1]
Living or Raising Conditions	6.6de	6.8b	6.8b	6.0bc	6.7bc	6.8c
	[5.4, 7.8]	[4.4, 9.6]	[3.4, 11.7]	[4.6, 7.6]	[5.2, 8.1]	[4.6, 9.6]
Environment-Related	4.7e	7.2b	4.4b	3.9c	5.5c	6.4c
	[4.0, 5.6]	[4.7, 10.5]	[2.1, 7.9]	[3.1, 4.9]	[4.3, 6.8]	[4.1, 9.5]
Animal Welfare	1.2f	1.4c	1.2c	1.0d	1.4d	2.2d
	[1.0, 1.4]	[1.0, 1.8]	[0.7, 1.9]	[0.8, 1.2]	[1.1, 1.7]	[1.6, 3.0]

Note: Trust shares can be interpreted as the percentage of the population falling within each group that would be expected to select the respective claim as most trusted. 95% confidence intervals for trust shares are reported in brackets. Within each column, mean trust shares are statistically different at P < 0.05 unless followed by a letter in common.

Trust shares and 95% confidence intervals for the household food shopping role groups are shown in Figure 2. Although trust shares vary across groups, all 95% confidence intervals overlap. The 95% confidence intervals are largest for the Less than Half group. Thus, there appears to be more noise in trust and distrust as household shopping responsibility decreases. Breed and Source/Traceability claims have the highest shares, with 32% and 30% of the More than Half and Around Half groups, respectively, trusting Breed claims the most. About 22% of respondents in both groups chose Source/Traceability claims. The Less than Half group differs from the others, with the highest share (28%) trusting Source/Traceability claims most. A slightly lower share (26%) selects Breed claims as the most trusted. Diet and Organic claims each have shares of 9% to 10% in the More than Half and Around Half groups. Meanwhile, these categories account for 12% and 5% of trust in the Less than Half group. Trust shares for Negative Antibiotic Use, Negative Hormone Use, and Living/Raising Conditions claims range from 6% to 8% across all groups. Environment-related claims account for 4% of shares in the More than Half and Less than Half groups, whereas the Around Half group has a slightly higher share of 7%. Trust shares for Animal Welfare claims are 1% for all groups.

Figure 2.

Trust shares for meat and poultry claim categories stratified by household food shopping role. Note: Error bars represent 95% confidence intervals.

Figure 3 illustrates the trust shares and 95% confidence intervals for the meat consumption groups. Trust shares associated with Breed claims ranged from 30% to 33% across all groups. While still secondary to Breed claims, shares for Source/Traceability claims showed more pronounced differences between the Daily (24%) and Once a Week or Less (19%) groups. Conversely, trust shares for Organic, Diet, or Negative Antibiotic Use claims remained relatively similar, ranging from 8% to 10% across all groups. Similar but slightly lower shares were observed for Negative Hormone Use and Living/Raising Conditions claims. Environment-Related and Animal Welfare claims received the lowest trust shares, with shares of 2% or less for Animal Welfare claims across groups.

Figure 3.

Trust shares for meat and poultry claim categories stratified by meat consumption frequency. Note: Error bars represent 95% confidence intervals.

To provide further insight into consumer trust in animal-raising and environment-related claims, Table 4 shows how often respondents perceive a claim seen in a grocery store as valid. Consistent with previous results, we find that Breed claims are the most commonly believed by consumers. On average, consumers find these claims valid 63% of the time, with 72% of respondents choosing a response option greater than 50%. Source/Traceability and Diet claims also have average belief levels above 50%. The lowest belief frequencies are for Animal Welfare and Environment-Related claims. Consumers believe Animal Welfare claims are valid only 40% of the time, with 70% of respondents selecting a response below 50%. Overall, for six of the nine claim categories, the average predicted belief levels range from 45% to 50%.

Table 4.

Consumer belief levels in the validity of a claim displayed on a meat or poultry product label.

Claim category	Proportion of respondents
	0% to 25% of the time	26% to 50% of the time	51% to 75% of the time	76% to 100% of the time	Average predicted belief level
Breed	8.08	19.39	35.62	36.90	63.30
Source or Traceability	14.07	26.73	32.46	26.73	55.92
Diet	14.28	33.40	39.39	12.93	50.71
Organic	18.38	34.14	32.86	14.61	48.88
Negative Antibiotic Use	21.28	37.10	30.30	11.31	45.87
Negative Hormone Use	21.95	36.43	30.44	11.18	45.66
Living or Raising Conditions	25.12	34.34	29.63	10.91	44.52
Environment-Related	23.10	41.14	27.00	8.75	43.31
Animal Welfare	29.70	40.61	23.10	6.60	39.59

Note: The average predicted belief levels were estimated with 1,485 observations.

The response proportions for questions regarding knowledge of and trust in certifying entities are shown in Table 5. We find that 80% of respondents identify the USDA as the responsible entity for certifying labeling claims found on meat and poultry products. An additional 50% of respondents attribute responsibility to the FDA, while 14% selected that the food company was responsible. The remaining categories (ie, the CDC, a third-party, and “I am not sure”) were chosen by around 10% of respondents. As shown in the second column of Table 6, consumer trust in entities to handle certification followed a similar pattern. Approximately 62% of respondents report trusting the USDA to certify these claims. Interestingly, only 11% of respondents report trusting a third party to be responsible for certification.

Table 5.

Consumer knowledge of and trust in entities certifying meat and poultry labeling claims.

Entity	Proportion of respondents
	Entity responsible for certifications	Entity most trusted to certify
USDA	80.20	61.68
FDA	50.30	21.89
The food company	14.28	1.89
A third-party	11.72	10.71
CDC	10.71	3.84
I am not sure	8.42

Note: Number of observations was 1,485. In the first column, respondents were given the option to select multiple options, therefore proportions will not equal 100%. If “I don’t know” was selected, the respondent was not allowed to select any others.

Table 6.

Impact of a third-party certification on consumer trust in a claim displayed on a meat or poultry product label.

Claim category	Proportion of respondents
	Increased trust	No change
Animal welfare	65.52	34.48
Breed	67.68	32.32
Diet	64.04	35.96
Environment-related	66.13	33.87
Living or raising conditions	67.34	32.66
Negative antibiotic use	63.97	36.03
Negative hormone use	63.97	36.03
Organic	66.67	33.33
Source or traceability	67.47	32.53

Note: Respondents were asked whether a third-party certification would increase trust in each of the respective claims categories. Yes or no response options were provided. Number of observations was 1,485.

The response proportions for the survey question regarding the impact of a third-party certification on consumer trust are reported in Table 6. We find that a third-party certification would have a mostly positive effect on consumer trust. Contrary to the variability in trust across claims observed so far, two-thirds of respondents agree that a third-party certification would increase their trust in a claim on the label of a meat or poultry product, regardless of the claim presented. About one-third of respondents agree that this would not affect trust, regardless of the claim category.

Discussion

The updated guidelines for substantiating animal-raising and environment-related claims arrive amid growing public scrutiny of these voluntary marketing statements. Issues related to non-compliance and regulatory oversight have raised doubts about both the accuracy of the claims and the agency’s competence. As a result, consumer perceptions of these claims might decline. The FSIS, however, has taken steps to strengthen the documentation required to substantiate a claim by emphasizing the importance of third-party certification. We surveyed 1,485 respondents in December 2024 using a BWS approach. The survey measured consumer trust across nine different claim categories outlined in the FSIS guideline and further examined how third-party certification influences trust. Additionally, we assessed consumer awareness and trust in certifying organizations. Our research findings highlight four key points.

First, there is a disproportionate level of trust among the nine claim categories related to animal raising and the environment. Breed and Source/Traceability claims are among the most trusted claims used on meat and poultry products. In contrast, consumers trust Animal Welfare claims the least, though still more than the less-trusted categories like Environment-Related.

Second, while there are slight differences in trust shares across consumer groups, household shopping role or meat consumption frequency has little to no clear impact on trust across claims. Trust levels vary across the three household shopper groups; however, the results do not support a definitive conclusion. Small trends are observable based on meat consumption frequency. Relative trust in ethical and environmental claims (such as Animal Welfare, Environment-Related, Organic) increases as meat consumption decreases. Conversely, trust in more objective claims (like Breed and Source/Traceability) tends to decline among infrequent meat consumers. Nonetheless, the slight differences and the large confidence intervals indicate that these findings should be interpreted cautiously.

Third, adding third-party certification is likely to boost consumer trust in animal-raising and environment-related claims, although the extent of this improvement remains uncertain. This pattern was consistent across all nine claim categories. The final key finding is that consumer trust in animal-raising and environment-related claims is generally low. For six of the nine categories, consumers believe these claims are valid less than 50% of the time. Breed, Source/Traceability, and Diet claims have higher belief levels but are still trusted less than 65% of the time.

The distrust shown in our findings reflects how much food influencers can impact trust in food products (Wu et al. 2021). Issues related to food litigation and reports of food malpractice have been common in the media recently. For claims about animal raising and the environment, reports of non-compliance and government mishandling continue to emerge. As Truong et al. (2021) suggest, these issues can hurt consumer confidence in the entire food system. Additionally, our findings may indirectly reveal problems with misaligned consumer expectations or confusion about the meaning of claims. Past research shows that consumers often lack basic knowledge about on-farm practices and animal-raising claims (Powers et al. 2020; Yang et al. 2020; Adam and Bruce 2023). Many industry stakeholders have also voiced concerns about this. Among the many comments received by FSIS in response to the 2019 guideline update, many expressed concerns about a mismatch between producer and consumer expectations. This disconnect could mislead consumers (USDA 2024b). Ultimately, ambiguity in claims can cause consumer confusion and skepticism, whereas more transparent labeling can enhance product value (Sirieix et al. 2013). Our study seems to support this idea. Simpler and more objective claims (eg, Breed, Source/Traceability) were often rated as most trustworthy. Meanwhile, claims related to detailed on-farm practices and broad statements (such as Animal Welfare, Environment-Related) were often seen as least trustworthy. Since most respondents in our study come from urban or suburban areas (see Appendix Table S1—see online supplementary material), some likely lacked knowledge of animal-raising practices.

There appears to be a need to build or rebuild consumer trust in claims related to animal raising and the environment. The recent update to the substantiation guidelines could be a significant step toward addressing the lack of consumer confidence. Consistent with previous findings (Spain et al. 2018), our results indicate that involving a neutral party in the certification process can positively influence consumer trust. In the short term, this should be considered a reasonable action, and FSIS should continue to emphasize the importance of including this certification in supporting documents. However, additional measures might be needed to fully address consumer perceptions.

One potential area of focus may lie in our findings about trust in certifying entities. Previous research has shown that consumer trust in food products is closely linked to confidence in the certifying organizations (Janssen and Hamm 2012; Rupprecht et al. 2020). Surprisingly, we find that trust in animal raising and environment-related claims is relatively low; however, respondents generally trust the USDA to certify labeling claims found on meat and poultry products. This conflicting result suggests that the distrust of animal raising and environment-related claims may not be solely attributable to issues with the certifying body. Instead, the root of this trust issue could stem elsewhere. Another key element of consumer trust is confidence in the trustworthiness of food actors (Macready et al. 2020; Wu et al. 2021). As consumers often doubt the regulating agency’s ability to oversee on-farm activities and ensure that farmers meet certification standards (Truong et al. 2021), issues may lie in their trust of producers. To enhance consumer confidence in the accuracy of labeling claims, future policies could focus on strengthening oversight of on-farm practices. This idea is partly supported by the FSIS’s recognition of its limited jurisdiction. Beyond serving as an additional layer of assurance, third parties can enhance consumer confidence by enforcing more consistent regulations of animal-raising methods.

Furthermore, the results highlight the importance of improving the understanding of claim meaning. Educating consumers about the on-farm activities involved in producing products with animal-raising or environmentally related claims may help reduce confusion about the implications of these claims. Given the varying levels of trust across different categories in our study, specific areas for targeted intervention can be identified. In addition to consumer education, developing standardized definitions could be beneficial. As argued by Tonkin et al. (2016), consumers may believe that generalized claims are intentionally meant to mislead. Consequently, the presence of such claims can erode consumer trust, even if they comply with regulations. Establishing fixed standards and clear definitions for generalized terms would help minimize the risk of misaligned expectations and unwarranted skepticism.

This study has several limitations that should be acknowledged. First, although the BWS approach requires respondents to make explicit trade-offs, this was a hypothetical task. As with other stated-preference methods, responses may be subject to hypothetical bias, and the results should be interpreted as reflecting stated trust rather than observed behavior. Second, responses may be influenced by social desirability bias, particularly for claims related to environmental practices or animal welfare. Although these claims were, on average, less trusted, respondents may still have overstated or understated trust. Furthermore, surveys can be subject to measurement error and limitations in external validity. For example, respondents may not fully understand survey questions, and the results may not be generalizable to other populations or over time, as trust can change. Another limitation is that we do not explicitly account for consumer exposure to claim categories. For example, some claims have been present in the marketplace for an extended period (eg, breed), while others have emerged only in recent years (eg, environmental impacts). As a result, observed variations in consumer trust may be partially attributed to differences in claim familiarity. Future research could incorporate direct measures of claim exposure to better isolate the role of consumer experience in shaping perception.

Future research could examine the reasons for the varying levels of distrust across these categories. Identifying the source of consumer skepticism will enable more targeted policy actions. Additionally, this research offers a framework for similar studies in other areas of food labeling. While we focus on one aspect of food labeling, issues of consumer confidence in the overall food supply may persist. Very little literature has examined the levels of trust in food labeling, and further research may be needed to assess the extent of this issue. Although there may be no quick fix to restore consumer trust, small measures such as encouraging third-party certification can help rebuild confidence in the food they purchase. In light of the recent FSIS guideline revision, this research contributes to the discussion of food labeling by assessing consumer trust in animal-raising and environmental labeling claims. We hope to have provided policymakers and producers with insights into a potential concern in the meat and poultry industry.

Glossary

Abbreviations

ARS

agricultural research service

BIBD

balanced incomplete block design

BWS

best-worst scaling

FSIS

food safety and inspection service

IIA

independence of irrelevant alternatives

IID

independently and identically distributed

MNL

multinomial logit

RPL

random parameter logit

RUM

random utility model

CDC

United States Centers for Disease Control and Prevention

USDA

United States Department of Agriculture

FDA

United States Food and Drug Administration

Acknowledgments

This work was funded by the Tyson Endowed Chair in Food Policy Economics.

Author contributions

Andrew Dilley (Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft), Brandon Ray McFadden (Conceptualization, Data curation, Funding acquisition, Methodology, Supervision, Visualization, Writing—review & editing), James Mitchell (Conceptualization, Methodology, Supervision, Writing—review & editing), and Jada Thompson (Conceptualization, Methodology, Supervision, Writing—review & editing)

Supplementary data

Supplementary data is available at Journal of Animal Science online.

Conflict of interest statement. The authors declare no real or perceived conflicts of interest.