Research on predictors of uses of animals has surged over the past twenty years (A. Feltz & Feltz, 2021; Ruby, 2012). For example, those who think eating meat is normal, necessary, nice, and natural (i.e., the 4Ns) are more likely to consume more animal products than those who do not (Piazza et al., 2015). Research also suggests that knowledge of animal related issues can predict animal related outcomes. For example, those who know more about animals used as food also consume fewer animal products (A. Feltz et al., 2024; S. Feltz & Feltz, 2019). Concern for animal welfare is also commonly assessed. Those who care more about animal welfare are more supportive of animal protections (Rothgerber, 2015). Attitudes about uses of animals, such as acceptability of hunting animals for sport, predict some behavioral outcomes like consuming fewer animal products (Herzog et al., 2015).

Some evidence suggests that there might be important differences in views of farmed animal welfare between residents of high and low agriculture states. The Humane Society of the United States has listed which states have laws specifically protecting factory farmed animals¹1

https://www.humanesociety.org/blog/california-which-passed-worlds-strongest-farm-animal-protection-law-and-nations-first

We aimed to provide direct evidence about potential differences concerning farmed animal health and welfare between residents of high animal agriculture and low animal agriculture states. We also provide evidence about what more proximal cognitive mechanisms might predict those differences (e.g., knowledge and one’s values) and determine if those predictors were influenced by state of residency. The predictors selected were based on the Framework for Skilled Decisions (Cokely et al., 2018). While the Framework for Skilled Decisions includes several elements (environment, cognitive abilities, deliberation, and confidence), this Framework suggests that many outcome variables can be predicted by what one knows about an issue along with relevant values. This Framework has been successfully applied to related issues involving consumption of animals and animal products (A. Feltz et al., 2024; Feltz et al., 2022), consumer views of plant-based products (S. Feltz & Feltz, 2019), views about natural resource management (Tanner & Feltz, 2022), views about endangered species (Offer-Westort et al., 2020), among a host of other outcomes. For example, those who know more about animals used as food and have weaker value-related justifications for consuming meat tend to consumer fewer animal products than those who know less or have stronger value-related justifications for consuming meat (A. Feltz et al., 2024).

Overall, we expected that residents of high animal agriculture states would have views less supportive of factory farmed animal welfare compared to views of residents in low animal agriculture states (e.g., less favorable views of animal advocacy agencies, more approving of body part removal and killing practices on industrialized farms. See below for more specific predictions). Our approach would also allow us to see if residency in those states moderated the expected relations between the predicter variables (i.e., knowledge and values) and views concerning farmed animal welfare (e.g., body part removal practices).

To help reduce the complexity and number of analyses, we first factor analyzed the items that we created for this study. Exploratory factor analyses suggested that a number of the sets of questions could be meaningfully captured by one or two factors (all the exploratory factor analyses are provided in the supplemental materials) and are summarized in Table 1. To determine if presentation order of the blocks of items influenced responses, we conducted a MANOVA with order of presentation as the fixed factor and all the variables listed above as the dependent variables. The MANOVA indicated a significant difference between the two orders, F (23, 1961) = 3.47, p < .01, ηp2 = .04. A subsequent inspection of the ANOVA tables for each variable indicated small to trivial differences on the dependent variables (ηp2s ≤ .01). Consequently, while order of presentation may have influenced responses, order was not a strong influence. For ease of analyses, we did not include order of presentation in any subsequent analyses.

While our hypotheses were theory driven, they were not preregistered, so we adopted a cautious analytic strategy. After conducting the exploratory factor analyses, we divided the full data set into two sets of data—a test set and a validation set. Participants were randomly assigned to only one of the test set (N = 992) or validation set (N = 993). For the test set, there were in 497 participants from low animal agriculture states and 495 from high animal agriculture states. For the validation set, there were 498 from low animal agriculture states and 495 from high animal agriculture states. See Table 2 for means and standard deviations of the dependent variables in the test set (all analyses and results for the validation set are provided in the supplemental materials).

The means for the test set, t-tests using state of residency as the independent variable, and effect sizes are reported in Table 2. We found no measurable differences between the two sets of states with respect to knowledge of factory farming, government obligations to farmed animals, state responsibility for animal welfare, farmer responsibility for animal welfare, distributors’ responsibility for farmed animal welfare, consumers’ responsibility for farmed animal welfare, killing practices, mental state attributions, plant-based school advocate favorability, personal connection to animals, or personal activism.

Compared to residents of low animal agriculture states, those in high agriculture states had lower scores on the Knowledge of Animals Used as Food scale, lower estimates of percent of farmed animals on industrialized farms, lower frequency of eating meat substitutes, lower ratings for government obligations to human welfare, lower ratings for federal responsibility for farmed animal welfare, lower favorability of animal protection advocates, and lower favorability of plant based advocates. Compared to those in low animal agriculture states, residents in high animal agriculture states had higher ratings for removal practices on industrialized farms, had higher 4Ns scores, and consumed more animal products. Each of these effects was replicated in the validation set suggesting that the small differences were robust (for similar analyses and pattern of results using residence in urban, suburban, or rural settings as the independent variable instead of state of residency, see the supplemental materials).

Our second set of analyses was aimed at modeling these data along the lines of the Framework for Skilled Decisions (Cokely et al., 2018). This model gives central, non-redundant roles to knowledge and values in predicting outcome variable (for overall correlations, see Table 3). Previous research suggested that the Knowledge of Animals as Food scale tends to predict outcomes related to consuming animals and animal products while the Knowledge of Factory Farming scale tends to predict animal welfare and protection type outcome variables (A. Feltz et al., 2024). We used those two measures as the knowledge measures for the relevant outcome variables. For the values predictor variable, we used the 4Ns since this follows previous research (A. Feltz et al., 2024; Piazza et al., 2015).

Additionally, we were interested in determining if residency in a high or low animal agriculture state would moderate some of the relations between relevant knowledge and the 4Ns with the outcome variables. So, for each of the outcome variables, we constructed a hierarchical linear regression. In the first step, we included only the relevant knowledge scale and the 4Ns. In the second step, we included state of residency, the 4Ns, the relevant knowledge scale, and the interaction terms of both the state of residency by 4Ns and the state of residency by the relevant knowledge scale. The model in the first step would allow us to determine the non-redundant predictive power of knowledge and the 4Ns along the lines of the Framework for Skilled Decisions. The model in the second step would allow us to determine if those relations were moderated by state of residency. For most of the outcome variables in the test set, the 4Ns and relevant knowledge were significant, non-redundant predictors and state of residency did not moderate those relations. Most of these results were replicated in the validation set data (see the Supplemental Materials for full analyses). In general, those who were more knowledgeable had views more supportive of factory farmed animal welfare compared to those who were less knowledgeable. Endorsement of the 4Ns were negatively related to views supportive of factory farmed animal welfare (for similar analyses and pattern of results that use political orientation as a potential moderator, see the supplemental materials).

This study examined whether residents in high and low animal agriculture states differed in their views on farmed animal welfare. In a large-scale study that included a direct replication, we found no measurable differences between residents of high and low animal agriculture states in the U.S. regarding knowledge of factory farming, responsibility for farmed animal welfare attributed to the state, farmers, distributors and consumers, killing practices, mental state attributions, plant-based school advocate favorability, personal connection to animals, or personal activism. Compared to residents of low animal agriculture states, those in high agriculture states knew less about animals used as food, had lower estimates of percent of farmed animals on industrialized farms, ate meat substitutes less frequently, had lower ratings for government obligations to human welfare, lower ratings for federal responsibility for farmed animal welfare, lower favorability of animal protection advocates, and lower favorability of plant-based advocates. Compared to those in low animal agriculture states, residents in high animal agriculture states were more accepting of body part removal practices on industrialized farms, had higher scores on the 4Ns, and consumed more animal products. In addition, knowledge of animals used as food/factory farmed animals or the 4Ns were significant predictors of all the other outcome variables we gathered regardless of state of residency.

The results of our study were largely in line with the existence of real, but small, differences between residents of high and low animal agriculture states. For example, previous research suggests that Southerners tend to be slightly more speciesist (i.e., view the interests of non-human animals differently than the identical interests of humans) than Westerners in the U.S. (Wulderk & Anderson, 2023). For some of the outcome variables in our study, residents of high animal agriculture states were slightly more supportive of factory farming activities than those in low animal agriculture states. However, the magnitude of the differences between residents of those states was relatively small. The average effect size for the significant differences was equal to a Cohen’s d of 0.18 (the average Cohen’s d for all variables was 0.13), which is typically thought to be small (Cohen, 1988). There were also no qualitative differences in the dependent variables we gathered. For example, residents of high animal agriculture states were slightly more supportive of body part removal practices on industrialized farms compared to residents of low animal agriculture states. However, the overall means were fairly strongly not supportive of those body part removal practices for residents in both high and low animal agriculture states. So, while there was a measurable quantitative shift in views between residents of high and low animal agricultures states, residents of both states were on average strongly opposed to body part removal practices. Even if there may be some small quantitative shifts in some views about animals based on state of residency, the predictor variables of those views remained invariant and consistent with previous research and theory.

Whether the significant differences we found between residents of high and low animal agriculture states are practically important depends largely on one’s goals. To illustrate, if one is interested in having the most accurate descriptive or predictive models of views related farmed animal welfare between residents of high and low farmed animal agriculture states, then these differences may be important for those kinds of models. However, if one is interested in intervening to affect change in individuals, then the differences observed in our studies may not be of practical importance. As such, if an intervention works in one state, there is reason to think that intervention will work in all states and animal advocates may be able to ignore the small differences related to state of residency. Of course, there are other goals that researchers may have that may factor in determining the importance of those small effects.

There are several limitations of this study. First, we only sampled 5 low and 5 high animal agriculture states in one country. While this was by design to approximate an extreme groups analysis, there may be something idiosyncratic about those groups of states that could account for the overall differences, or lack of differences, we observed. This worry may be mitigated to some extent because while there were some mean differences between residents of those two sets of states, the predictor variables were invariant with respect to states. The invariance of predictor variables suggests that the mean differences were not something idiosyncratic about states but rather about differences in more proximal causes such as knowledge and values that are likely to be similar across states. We also only measured a handful of uses of animals. There are many other uses of animals that could, potentially, display larger differences across state of residency (e.g., animals used for entertainment; medical research; hunting). While we have no evidence to suggest that those other uses would have larger difference, those potential difference could also be explored. Additionally, we gathered a rich data set with many possible relations that we did not test. For those who are interested in those kinds of analyses, the entire data set is available (see Feltz & Dillard, 2024). Finally, other potential theoretical models could be applied to these data. While we modeled the data in accordance with the Framework for Skilled Decisions, subsequent work could evaluate whether there are other useful models to help understand the effects reported here.

In conclusion, we think the message this study is clear. Residents of high and low animal agriculture states are more similar than they are different with respect to views about factory farmed animal welfare. While there may be some other important individual differences to document (see, for example, one’s readiness to change behaviors involving animals, Hoang et al., 2023), state of residency seems to have relatively small impacts on views about farmed animal welfare.