Toxic Baits as a Strategy for Controlling Invasive Wild Pigs: Acceptability Among Crop Producers

Abstract

Wild pigs (Sus scrofa) have become a pervasive issue in the United States, causing significant damage to agricultural lands and ecosystems. Toxic baits have been developed as a potential management tool. This study evaluates the acceptability of toxic bait usage among U.S. crop producers and explores influential factors of respondent attitudes. Using data from a survey of crop producers in 11 states, we found that 59.6% of respondents somewhat or completely agreed that the use of a toxic bait in their state was acceptable, and 71.4% of respondents somewhat or completely agreed that they would be comfortable using toxic baits if they were experiencing a problem with wild pigs. Regression model results revealed that larger operations and producers of vulnerable crops were more likely to find use acceptable. Producers who generated income from wild pig hunting on their land tended to be less accepting, while those who generated income from the hunting of other species on their land tended to be more accepting. These findings suggest that the acceptability of a toxic bait is shaped by various economic, operational, and demographic factors but that there would likely be demand among producers of high-value crops facing a wild pig problem.

1. Introduction

Wild pigs (Sus scrofa) were first introduced to the continental U.S. in the 1500s as a food source [1] but are now a significant ecological and economic problem. Known for their adaptability, high reproductive rates, and destructive behaviors, wild pigs cause extensive damage to agricultural crops, livestock, and native ecosystems and even pose risks to human health [2,3]. The economic burden of wild pigs on U.S. crop producers is substantial. A study of wild pig damage to six crops in 11 states found that producers experienced approximately USD 678.8 million in crop losses and related expenses annually, with roughly 30% of the estimate (USD 207.5 million) representing producers’ expenditures to control wild pig populations and mitigate damages [4]. More broadly, total wild pig agricultural damage in the United States has been estimated to exceed USD 2 billion annually [5].

Common methods for managing wild pig populations, such as trapping, hunting, and fencing, are often costly and inadequate due to the species’ elusive nature, adaptability, and large population sizes [3,6]. As a result, there has been growing interest in the development and use of toxic baits as an additional control method. Toxic baits, designed to lethally remove wild pigs through their ingestion of a toxic substance, offer a potentially effective approach for removing large numbers of wild pigs as part of an integrated pest management strategy [7]. However, there are challenges associated with the development of a toxic bait, including the need for target species specificity and a relatively quick and humane death following ingestion of the bait. These challenges are reflected in the limited product development and approval in the United States. Kaput^® Feral Hog Bait was licensed by the U.S. Environmental Protection Agency (EPA) and registered for use in Texas in 2017 to target the growing wild pig problem. The product contains warfarin, an anticoagulant, which is also used as a blood thinner in human medicine. Despite its EPA approval, the manufacturer pulled the product from the Texas market due to strong opposition from conservation groups, meat processors, and hunters, highlighting the importance of understanding public and stakeholder acceptance prior to product rollout [8]. More recently, the product was once again registered in Texas, as well as Oklahoma, in 2024 [9,10]. At present, another toxic bait, Hoggone^TM, formulated with sodium nitrite, is already registered for use in Australia [11] and is being evaluated for potential use in the United States. Its purported advantages include that it is fast acting, humane, and does not bioaccumulate, thereby mitigating secondary risks to scavengers [12]. However, field trials demonstrate nontarget risks, so registration has not been pursued [12,13].

With the potential controversy associated with the use of toxic baits for wild pig control, the investigation of stakeholder perspectives and attitudes concerning this control strategy is a growing area of study. Tucker Williams et al. [14] found that Alabama agricultural producers, hunters, and forestland owners were generally supportive of the use of wild pig toxic bait, though average acceptability scores were higher for a sodium nitrite bait than for a warfarin bait. In another Alabama study, Ellis et al. [15] found that most agricultural producers found the use of wild pig toxic baits to be acceptable, though all other control methods—e.g., aerial gunning and ground shooting—were significantly more acceptable. Other studies, however, have found lower levels of support for wild pig toxic baits. In a nationwide survey of the public, Carlisle et al. [8] found that a slight majority (51%) of respondents viewed the use of a wild pig toxic bait to be unethical, while less than one-quarter (23%) found such usage to be ethical (the remainder were neutral). Primary concerns raised by respondents included collateral harm to nontarget animals (33% of respondents) and the potential suffering of wild pigs (13% of respondents). Additionally, a survey of Texas recreational hunters found that the use of a wild pig toxic bait was acceptable to 43% of respondents and unacceptable to 39% of respondents and that respondents were fairly polarized on the issue based upon a Potential for Conflict Index (PCI₂) analysis [16]. Another study based on the same Texas hunter survey data found that all other wild pig control methods were significantly more acceptable among hunters than toxic bait [17].

This study builds on the growing body of scholarship concerning wild pig toxic baits by addressing a gap in the literature. Specifically, we examined the acceptability of wild pig toxic baits among a stakeholder group that we would expect to have the greatest need for the product: producers of high-value crops with wild pigs present in their operations. Our research objectives were to (i) measure the level of acceptability of a wild pig toxic bait among producers of high-value crops with wild pigs present in their operations in 11 U.S. states and (ii) to understand whether acceptability levels were influenced by producers’ demographic characteristics, livestock operation characteristics (including any wild pig management action undertaken), hunting activities, and the change they preferred for the wild pig population size in their state—referred to as their wildlife acceptance capacity (WAC) for wild pigs [18]. Understanding crop producer acceptability of a wild pig toxic bait is a crucial step for decision-makers, as it affords them insight into possible levels of adoption and usage among these key stakeholders. This insight could, in turn, help them better evaluate whether a toxic bait should be included in their overall strategy for controlling wild pig populations in the event that such a product is approved and registered for use in their state.

2. Materials and Methods

2.1. Data Collection

We collected data for this study using a mail survey of crop producers in 11 states: Alabama, Arkansas, California, Florida, Georgia, Louisiana, Mississippi, Missouri, North Carolina, South Carolina, and Texas. The survey questionnaire was developed by two of the authors and administered by the USDA National Agricultural Statistics Service (USDA-NASS) in 2022 following a round of pretesting with nine agricultural producers from the study area. The states in the study area were chosen because they have medium-to-large wild pig populations and economically important crop production sectors. The questionnaire included items from multiple research studies, including studies that examined wild pig control preferences and crop damage estimates. Regarding the latter, crops of interest were selected based on their high production value [19] and comprised corn (Zea mays), soybeans (Glycine max), wheat (Triticum spp.), rice (Oryza sativa), peanuts (Arachis hypogaea), and sorghum (Sorghum bicolor).

The sampling frame comprised all active producers of at least one of the six crops of interest in the selected states on USDA-NASS’s List Frame, and the sample was selected using a multivariate probability proportionate to size (MPPS) sample design. USDA-NASS mailed questionnaires to all sample members (N = 11,495) on 27 June 2022, and it mailed a follow-up reminder on 18 July 2022. In total, 4534 surveys were returned, with a response rate of 39.4%. Respondents who reported no wild pig presence on their operation (n = 2902) were screened from answering the toxic bait items, resulting in 1632 respondents remaining in the sample for this study.

2.2. Toxic Bait Acceptability

To measure the acceptability of a wild pig toxic bait, we used a questionnaire item that asked respondents to indicate on a 5-point Likert-type scale how acceptable the use of a toxic bait in their state would be to them (1 = “completely unacceptable, 3 = “neutral”, and 5 = “completely acceptable”). Respondents were also provided the following characteristics of the toxic bait to inform their response: (i) it was approved by federal and state authorities, (ii) it could only be used by trained government personnel to assist landowners (with landowners’ permission), (iii) the bait would be placed in a feeder designed to prevent other animals from reaching and consuming the bait, (iv) wild pigs would die within a few hours of consuming the bait, with little pain and suffering, (v) the likelihood other animals would consume the bait and die was very low (but not zero), and (vi) toxic levels of the active ingredient would not accumulate in a wild pig’s muscle or meat tissues, and it was very unlikely there would be any negative effect on humans or animals that consume a wild pig carcass. These characteristics were selected because they generally reflect the profile of a sodium nitrite-based toxic bait, albeit one that has been formulated to significantly mitigate the likelihood of nontarget mortality.

We also included three questionnaire items to gain additional insight into respondents’ views concerning a wild pig toxic bait, but one was excluded from this study because of clear measurement error (The excluded statement read: “I would be opposed to the toxic bait even if published research indicates that it is completely safe for humans, other animals, and the environment”. Analysis revealed that 13% of respondents who somewhat or completely agreed with the foregoing statement also somewhat or completely agreed that the use of a toxic bait in their state would be acceptable to them. We believe the confusion arose because the excluded statement was the only statement that was worded in the negative (i.e., oppositional to a toxic bait)). These items were formulated as statements, and respondents were asked to agree or disagree with the statements using the same 5-point Likert-type scale. The aim of the first was to understand the proportion of respondents who would feel comfortable using the toxic bait if there were a need (“If wild pigs were a problem on my property, I would feel comfortable using the toxic bait on my property”). The aim of the second was to understand perceptions of risk associated with toxic bait usage and the consumption of game meat (“I would feel comfortable eating the meat of a game animal (other than wild pig) harvested in the same area where the toxic bait was used”) (see Appendix A Figure A1). We calculated mean scores for each statement (Appendix A Figure A1), and all sample means were weighted using producer-level weights adjusted for non-response to specific items [4].

2.3. Ordered Logit Model

To determine the factors that influenced respondents’ acceptability of a wild pig toxic bait, we employed an ordered logit regression model [20]. Our dependent variable was constructed from the first toxic bait-related item, which provided an ordinal measure of respondents’ acceptability of toxic bait use in their state. Specifically, the survey prompted respondents to rate their acceptability concerning the use of toxic bait to control wild pig populations in their state on a five-point scale such that the responses were categorical and ordered in nature (see Appendix A Figure A1). We chose to use an ordered logit regression to investigate the relationship between explanatory variables and an ordinal dependent variable because of the inherent ordering in the response category. The ordered logit provides the foundation for our examination of the relationship between respondents’ categorical level of acceptability for toxic bait use, $A_i$, and a set of explanatory variables, ${\mathit{X}}_{\mathit{i}}$. This framework expresses the i-th respondent’s level of underlying acceptability, $A_i^{*}$, to be a linear function of ${\mathit{X}}_{\mathit{i}}$, such that

$$A_i^{*}={\mathit{X}}_{\mathit{i}}\mathit{\beta }+u_i,$$

(1)

where $A_i^{*}$ is an unobserved index of respondent acceptability that is expressed as

$$A_i=\left\{\begin{array}{l}1\mathrm{i}\mathrm{f}{A}_i^{*}\le {\mu }_1\\ 2\mathrm{i}\mathrm{f}{\mu }_1<A_i^{*}\le {\mu }_2\\ 3\mathrm{i}\mathrm{f}{\mu }_2\le A_i^{*}\le {\mu }_3\\ 4\mathrm{i}\mathrm{f}{\mu }_3\le A_i^{*}\le {\mu }_4\\ 5\mathrm{i}\mathrm{f}{{\mu }_4\le A}_i^{*}\end{array}\right.$$

(2)

where the $\mu$’s are unknown threshold parameters that must be estimated with $\mathit{\beta }$. There were five possible levels of acceptability based on the survey question design. The model defines $A_i=1$ as “completely unacceptable”, $A_i=2$ as “somewhat acceptable”, $A_i=3$ as “neutral”, $A_i=4$ as “somewhat acceptable”, and $A_i=5$ as “completely acceptable”.

The probability of observing the j-th level of acceptability, for $j\in [0,\dots ,5]$, corresponds to the probability that the estimated linear function plus a random error term is within a given range of the estimated categorical cutpoints, expressed as

$$p_{ij}=\mathrm{P}\mathrm{r}\left(A_i=j\right)=\mathrm{P}\mathrm{r}\left({\mu }_{j-1}<A_i^{*}={\mathit{X}}_{\mathit{i}}\mathit{\beta }+u_i<{\mu }_j\right)$$

(3)

where the coefficients $\mathit{\beta }$ are simultaneously estimated with the cutpoints $\mu =\left\{{\mu }_1,\dots ,{\mu }_{j-1}\right\}$ using maximum likelihood. The log likelihood function is

$$\ln L={\sum }_{i=1}^N{w}_i{\sum }_{j=1}^5{I}_j\left(A_i\right)\mathrm{l}\mathrm{n}{p}_{ij}$$

(4)

where N is the number of observations, $I_j\left(A_i\right)$ are binary indicator variables equal to one if $A_i=j$ and zero otherwise, the $w_i$ are survey weights, and $p_{ij}$ for the ordered logit specification is $p_{ij}=1/\left[1+\mathrm{e}\mathrm{x}\mathrm{p}\left(-{\mu }_j+{\mathit{X}}_{\mathit{i}}\mathit{\beta }\right)\right]-1/\left[1+\mathrm{e}\mathrm{x}\mathrm{p}\left(-{\mu }_{j-1}-{\mathit{X}}_{\mathit{i}}\mathit{\beta }\right)\right]$. This model is also known as the proportional odds models because $odds\left({\mu }_j\right)$ and $odds\left({\mu }_{-j}\right)$ have the same ratio for all combinations of explanatory variables where $odds\left({\mu }_j\right)=\mathrm{Pr}\left(A_i\le {\mu }_j\right)/\mathrm{P}\mathrm{r}(A_i>{\mu }_j)$. It implies that combining adjoining categories results in a loss of statistical efficiency but does not bias estimated coefficient values.

2.4. Explanatory Variables

Explanatory variables included characteristics of the crop-producing operation (operation size, crop-producing type, presence of organic crops, and insured status of crops), WAC for wild pigs (as indicated by desired population change in state), management actions undertaken (any control and public help), and farm and respondent demographics (primary occupation, decision making, hired manager, place of residence, gender, age, and income). Operation size was measured with a four-level categorical variable (0–499 acres, 500–1499 acres, 1500–2999 acres, and 3000 acres or more). The type of crop-producing operation was ranked by the value of production using a seven-level categorical variable (corn, peanuts, rice, sorghum, soybeans, wheat, and other crops). To measure the presence of organic crops, we used a binary indicator variable equal to 1 if one of the top three producing crops on an operation was certified organic and 0 otherwise. Similarly, the insured status of crops was measured as a binary indicator variable equal to 1 if one of the top three producing crops on an operation was insured and 0 otherwise.

To investigate whether WAC predicted the acceptability of toxic bait use, we constructed a four-level categorical variable that measured each respondent’s preference for wild pig populations to be “Increased/Stay the same”, “Decreased somewhat”, “Decreased greatly”, or “Completely removed”. In addition to respondents’ WAC for wild pigs, we included a binary indicator variable equal to 1 if a respondent was already employing some method of wild pig population control or damage management and 0 otherwise. Additionally, we included a binary indicator variable equal to 1 if a respondent was receiving federal, state, or local government aid to manage wild pigs on their property and 0 otherwise.

To investigate potential hunting-motivated influences, we constructed a four-level categorical variable that combined information about hunting activity and whether an operation realized any net income from hunting wild pigs or other species—e.g., the presence of a guide or outfitting service for paying hunters. This variable grouped respondents into four categories: (1) No hunting/Don’t know, (2) Hunting w/no income from wild pigs or other species, (3) Hunting w/income from other species, and (4) Hunting w/income from wild pigs. The first group of respondents were those who reported that either no hunting took place on their property in the previous year or that they were unaware of any hunting activity. The second group of respondents reported recreational hunting activity—i.e., hunting activity but no net income from hunting was reported. The third and fourth groups of respondents reported both hunting activity and positive net income from hunting, with the former reporting net income from hunting non-pig species and the latter reporting income from hunting wild pigs.

To control for other factors that may also influence producer acceptability of toxic bait use, we included the binary variables “main occupation” (farm or crop work or not), “decision making” (does the respondent makes the majority of decisions about farming or not), “hired manager” (is the respondent a hired manager or not), “lives on operation” (does the respondent currently live on the operation or not), and a binary measure of self-reported gender (female or not). Additionally, the respondent’s age was included in the model as a four-level categorical variable ($<$50 years, 50–59 years, 60–69 years, and $\ge$70 years), while the reported income of the respondent was included as a seven-level categorical variable (cutoffs at USD 25k, USD 50k, USD 100k, USD 250k, USD 500k, and USD 1 million).

2.5. Estimation Strategy

We performed a model selection exercise to determine the combination of explanatory variables that delivered the most parsimonious ordered logit regression model. Specifically, an ordered logit model was estimated for each unique combination of explanatory variables (n = 28,355), and these models were then compared and ordered by a corrected Akaike Information Criterion (AICc) [21]. The AICc potentially improves model selection compared to the traditional AIC in small sample sizes when the ratio of observations to estimated parameters is less than 40, as is the case in our analysis. Akaike weights were then calculated for each model. They can be interpreted as the probabilities of each model’s being the best model in an AIC sense [22]. All analyses were implemented in the R 4.4.2 computing environment using the MuMIn [23] and MASS [24] packages. Our models were estimated using all observations with complete responses to the questionnaire items used for the construction of model variables (N = 1155). All sample statistics and regression results were weighted using producer-level weights adjusted for non-response to specific items [4] to be representative of the major crop types of interest (corn, peanuts, rice, sorghum, soybeans, wheat) in the study area.

3. Results

3.1. Summary Statistics

The demographic characteristics of respondents indicated a diverse yet distinct profile. Most participants were male (94.2%), with females comprising only 5.8% of the sample (

Table 1. Summary statistics for responses with all necessary data variables included in ordered logit regressions.

	Mean	SE
Toxic Bait Acceptability
Acceptability of toxic bait use in respondent’s state
Completely unacceptable	0.145	(0.029)
Somewhat unacceptable	0.052	(0.010)
Neutral	0.170	(0.028)
Somewhat acceptable	0.174	(0.028)
Completely acceptable	0.460	(0.031)
Operation Characteristics
Operation size (acres)
0–499	0.299	(0.033)
500–1499	0.310	(0.030)
1500–2999	0.167	(0.016)
3000+	0.223	(0.022)
Main producing crop type
Other	0.230	(0.030)
Corn	0.291	(0.024)
Soybeans	0.166	(0.030)
Wheat	0.122	(0.020)
Rice	0.044	(0.008)
Sorghum	0.048	(0.016)
Peanuts	0.099	(0.012)
One of top three producing crops is organic
No	0.969	(0.010)
Yes	0.031	(0.010)
One of top three producing crops is insured
No	0.462	(0.033)
Yes	0.538	(0.033)
Wild Pig Acceptance Capacity and Control
Desired change in wild pig population
Increased/Stay the same	0.016	(0.004)
Decreased somewhat	0.030	(0.010)
Decreased greatly	0.285	(0.033)
Completely removed	0.670	(0.033)
Employing any wild pig control
No	0.184	(0.024)
Yes	0.816	(0.024)
Receiving public help for wild pig control
No	0.888	(0.017)
Yes	0.112	(0.017)
Hunting Preferences
Hunting on operation and hunting income
No Hunting/Don’t know	0.199	(0.027)
Hunting w/no income from FS or other species	0.715	(0.028)
Hunting w/income from other species	0.072	(0.012)
Hunting w/income from FS	0.014	(0.005)
Respondent Demographics
Respondent lives on crop operation
No	0.226	(0.026)
Yes	0.774	(0.026)
Respondent’s main occupation is crop operation
No	0.231	(0.029)
Yes	0.769	(0.029)
Respondent is crop operation’s decision maker
No	0.036	(0.011)
Yes	0.964	(0.011)
Respondent is a hired manager for crop operation
No	0.953	(0.010)
Yes	0.047	(0.010)
Respondent self-reported gender
Male	0.942	(0.018)
Female	0.058	(0.018)
Respondent age
<50	0.240	(0.026)
50–60	0.196	(0.021)
60–70	0.366	(0.033)
>70	0.198	(0.025)
Respondent income (USD)
Less than 25,000	0.021	(0.005)
25,000 than 49,999	0.075	(0.016)
50,000 to 99,999	0.229	(0.027)
100,000 to 249,999	0.399	(0.033)
250,000 to 499,999	0.122	(0.016)
500,000 to 999,999	0.057	(0.009)
1,000,000 or more	0.096	(0.020)

NOTES: The means and standard errors reported in this table are calculated using all observations with complete responses to the questionnaire items used for construction of model variables (N = 1155) using sample weights that are adjusted for non-response. Due to rounding, summary statistics may not total exactly 100%.

). Age distribution was skewed toward older individuals, with 36.6% aged between 60 and 70 years and 24.0% under the age of 50. Income levels varied, with the largest proportion of respondents earning between USD 50,000 and 249,999 annually (62.8%), while those earning less than USD 25,000 or more than USD 1,000,000 represented smaller fractions of respondents (2.1% and 9.6%, respectively). Hunting practices were predominantly focused on pigs or other species without generating income (71.5%). Operational data revealed that most respondents resided on their farms or ranches (77.4%) and worked in operations covering 0–499 acres (29.9%). Additionally, 96.8% of farms were categorized as non-organic, and a substantial majority did not identify themselves as hired managers (95.3%). These statistics highlight a predominantly male, older demographic actively engaged in traditional farming and hunting practices, with moderate income levels and a reliance on non-organic farming methods. Many respondents found the use of a wild pig toxic bait in their state to be either completely acceptable (43.7%) or somewhat acceptable (15.9%) (Figure 1). A large majority of respondents (71.4%) also agreed that they would be comfortable using toxic bait on their property if wild pigs were an issue, while a minority (43.6%) indicated they would be comfortable consuming wild game harvested in areas where toxic bait had been used.

3.2. Model Summary Statistics and Selection Results

Our model selection exercise determined that the most parsimonious ordered logit regression model included all explanatory covariates, with an Akaike model weight of 0.685 (Appendix A 

Table A2. Top ten most parsimonious models from ordered logit regression model selection exercise.

	Variables Included in Model
Model Rank	Intercepts (Cutpoints)	Respondent’s Age	Employing any Wild Pig Control	Respondent Is Crop Operation’s Decision Maker	Desired Change in Wild Pig Population	Respondent’s Gender	Respondent Is a Hired Manager for Crop Operation	Hunting on Operation and Hunting Income	Respondent’s Income	One of Top Three Producing Crops Is Insured	Respondent Lives on Crop Operation	Main Producing Crop Type	Respondent’s Main Occupation Is Crop Operation	Operation Size (Acres)	One of Top Three Producing Crops Is Organic	Receiving Public Help for Wild Pig Control	df	logLik	AICc	∆ in AICc	Model Weight
1st	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	x	37	−48,453.36	96,980.78	–	0.6845
2nd	x	x	x	x	x	x	x	x	x	x		x	x	x	x	x	36	−48,455.28	96,982.63	1.841	0.2726
3rd	x	x	x	x	x	x		x	x	x	x	x	x	x	x	x	36	−48,457.76	96,987.58	6.795	0.0229
4th	x	x	x		x	x	x	x	x	x	x	x	x	x	x	x	36	−48,458.52	96,989.10	8.318	0.0107
5th	x	x	x		x	x	x	x	x	x		x	x	x	x	x	35	−48,460.28	96,990.63	9.845	0.0050
6th	x	x	x	x	x	x		x	x	x		x	x	x	x	x	35	−48,460.62	96,991.32	10.531	0.0035
7th	x	x	x		x	x		x	x	x	x	x	x	x	x	x	35	−48,462.86	96,995.78	14.995	0.0004
8th	x	x	x	x	x	x	x	x	x	x		x	x	x		x	35	−48,463.73	96,997.53	16.744	0.0002
9th	x	x	x	x	x	x	x	x	x	x	x	x	x	x		x	36	−48,462.73	96,997.53	16.749	0.0002
10th	x	x	x		x	x		x	x	x		x	x	x	x	x	34	−48,465.52	96,999.10	18.319	0.0001

NOTES: A total of 28,355 unique ordered logit models were estimated with every possible combination of explanatory variables. This table reports the independent variables included in the top ten most parsimonious models along with the model degrees of freedom (df), log-likelihood (logLik), the corrected Akaike information criterion (AICc), the change in the AICc from the previous mode (∆ AICc), and the model weight if employing a composite model. Model estimates reported in the main text were produced by the first model in this table, which includes all possible independent variables.

). Only one alternative model, which excluded a single explanatory variable (whether respondent lived on crop operation), was found to be closely influential, with an Akaike model weight of 0.273. Given the disproportionate influence of the fully inclusive model and to avoid unnecessary complexity, we did not perform any multi-model averaging when generating our model estimates. Model estimated effect sizes were generated via our preferred ordered logit regression model—where all explanatory variables were included—and are presented in

Table 2. Coefficient and odds ratio estimates from ordered logit regression.

	Coef.	SE	p-Value	OR	OR 95% CI
Operation size (acres)
0–499 (reference)	–	–	–	–	–
500–1499	0.22	0.03	<0.001	1.25	[1.18, 1.32]
1500–2999	0.97	0.04	<0.001	2.64	[2.46, 2.84]
3000+	1.17	0.04	<0.001	3.21	[2.99, 3.44]
Main producing crop type
Other (reference)	–	–	–	–	–
Corn	0.34	0.03	<0.001	1.40	[1.31, 1.50]
Soybeans	−0.64	0.04	<0.001	0.52	[0.49, 0.57]
Wheat	−0.33	0.04	<0.001	0.72	[0.67, 0.78]
Rice	−0.26	0.06	<0.001	0.77	[0.69, 0.86]
Sorghum	0.88	0.06	<0.001	2.40	[2.15, 2.69]
Peanuts	0.20	0.05	<0.001	1.22	[1.11, 1.33]
One of top three producing crops is organic
No (reference)	–	–	–	–	–
Yes	−0.28	0.06	<0.001	0.75	[0.66, 0.86]
One of top three producing crops is insured
No (reference)	–	–	–	–	–
Yes	0.43	0.03	<0.001	1.53	[1.45, 1.62]
Desired change in wild pig population
Increased/Stay the same	–	–	–	–	–
Decreased somewhat	0.25	0.10	0.012	1.28	[1.06, 1.56]
Decreased greatly	1.08	0.08	<0.001	2.96	[2.52, 3.48]
Completely removed	2.00	0.08	<0.001	7.41	[6.31, 8.70]
Employing any wild pig control
No (reference)	–	–	–	–	–
Yes	0.32	0.03	<0.001	1.38	[1.31, 1.46]
Receiving public help for wild pig control
No (reference)	–	–	–	–	–
Yes	0.43	0.04	<0.001	1.54	[1.44, 1.66]
Hunting on operation and hunting income
No Hunting/Don’t know	–	–	–	–	–
Hunting w/no income from FS or other species	0.20	0.03	<0.001	1.22	[1.15, 1.29]
Hunting w/income from other species	0.76	0.05	<0.001	2.14	[1.94, 2.36]
Hunting w/income from FS	−0.34	0.09	<0.001	0.72	[0.60, 0.85]
Respondent lives on crop operation
No (reference)	–	–	–	–	–
Yes	−0.05	0.03	0.050	0.95	[0.90, 1.00]
Respondent’s main occupation is crop operation
No (reference)	–	–	–	–	–
Yes	−0.57	0.03	<0.001	0.56	[0.53, 0.59]
Respondent is crop operation’s decision maker
No (reference)	–	–	–	–	–
Yes	0.19	0.06	0.001	1.21	[1.08, 1.36]
Respondent is a hired manager for crop operation
No (reference)	–	–	–	–	–
Yes	0.15	0.05	0.003	1.17	[1.05, 1.29]
Respondent self-reported gender
Male (reference)	–	–	–	–	–
Female	−0.46	0.04	<0.001	0.63	[0.58, 0.69]
Respondent age (years)
<50 (reference)	–	–	–	–	–
50–60	0.48	0.03	<0.001	1.62	[1.52, 1.73]
60–70	−0.11	0.03	<0.001	0.90	[0.85, 0.95]
>70	−0.17	0.03	<0.001	0.84	[0.79, 0.90]
Respondent income (USD)
Less than 25,000 (reference)	–	–	–	–	–
25,000 than 49,999	0.25	0.08	<0.001	1.28	[1.10, 1.49]
50,000 to 99,999	0.60	0.07	<0.001	1.82	[1.58, 2.09]
100,000 to 249,999	0.58	0.07	<0.001	1.78	[1.55, 2.04]
250,000 to 499,999	0.82	0.08	<0.001	2.28	[1.98, 2.65]
500,000 to 999,999	0.43	0.08	<0.001	1.54	[1.31, 1.81]
1,000,000 or more	0.57	0.08	<0.001	1.77	[1.52, 2.06]
Cutpoints ($\mu$’s)
Completely unacceptable|Somewhat unacceptable	1.00	0.12	<0.001
Somewhat unacceptable|Neutral	1.42	0.12	<0.001
Neutral|Somewhat acceptable	2.45	0.12	<0.001
Somewhat acceptable|Completely acceptable	3.34	0.12	<0.001

NOTES: A total of 1155 survey observations were used to estimate model.

.

3.3. Farm Operation Characteristics

Farm size was a significant predictor of toxic bait acceptability. Respondents with larger operations were more likely to support the use of toxic bait in their state on average, with all else equal. Specifically, respondents with operations sized between 1500–2999 acres (OR = 2.64, 95%; CI [2.46, 2.84]) and those over 3000 acres (OR = 3.21; 95% CI [2.99, 3.44]) were more likely to deem the use of toxic bait acceptable compared to smaller operations. The main crop type also influenced respondents’ attitudes, with sorghum-producing farms being the most likely to accept toxic bait use (OR = 2.40; 95% CI [2.15, 2.69]) and those producing soybeans being the least likely (OR = 0.52; 95% CI [0.49, 0.57]), with all else equal. Organic farming had a weak negative association with toxic bait acceptability (OR = 0.75; 95% CI [0.66, 0.86]).

3.4. Wildlife Acceptance Capacity and Hunting Preferences

WAC was estimated to be a significant predictor of toxic bait acceptability. Most notably, respondents who preferred that wild pig populations be “Completely removed” were considerably more likely to be accepting of toxic bait use in their state relative to those respondents who expressed a desire for wild pig populations to be “Increased” or “Stay the same” (OR = 7.41, 95% CI [6.31, 8.70]). The estimated effect size was smaller for those who desired a wild pig population that was “Decreased greatly”, though this group still expressed a markedly increased level of toxic bait acceptability relative to the reference group (OR = 2.96; 95% CI [2.52, 3.48]).

Employing wild pig control measures slightly increased the likelihood of accepting toxic bait use (OR = 1.38, 95% CI [1.31, 1.46]), as did receiving public help for wild pig control (OR = 1.54, 95% CI [1.44, 1.66]). Hunting activity was another strong influence on acceptability. Respondents who allowed hunting in their operation, especially those allowing species other than wild pigs to be hunted for income, were more likely to support toxic bait use (OR = 2.14, 95% CI [1.94, 2.36]). However, those who reported net income from hunting wild pigs, specifically, were less likely to support the use of toxic bait in their state (OR = 0.72, 95% CI [0.60, 0.85]).

3.5. Respondent Demographics

Several demographic factors were significant predictors of toxic bait acceptability. Female respondents were less likely to support toxic bait use compared to males (OR = 0.63, 95% CI [0.58, 0.69]). Age also had an impact, with respondents aged 50–60 years more likely to accept toxic bait use (OR = 1.62, 95% CI [1.52, 1.73]), while those over 70 were marginally less likely to support it (OR = 0.84, 95% CI [0.79, 0.90]). Income was a strong predictor of toxic bait acceptability, with higher income groups being more accepting of toxic bait use on average, with all else equal. Respondents with higher incomes, particularly those earning between USD 250,000 and 499,999, were much more likely to support toxic bait use compared to those earning less than USD 25,000 (OR = 2.28, 95% CI [1.98, 2.65]).

3.6. Respondent Demographics Average Marginal Effects

We present average marginal effects for three explanatory variables that were significant predictors of toxic bait acceptability in our estimated ordered logit regression model (Figure 2). Average marginal effects provide a more intuitive interpretation of estimated effect sizes, reflecting the change in the probability of being completely accepting relative to the reference group. As respondents’ WAC for wild pig populations decreased, the marginal probability of being completely accepting of toxic bait use increased from 4% (p = 0.017, wild pig populations “Decreased somewhat”) to 40% (p < 0.001, wild pig populations “Completely removed”) (Figure 2A). Similarly, as operation size grew, the marginal probability of being completely accepting of toxic bait use increased from 5% (p < 0.001, 500–1499 acres) to 25% (p < 0.001, 3000+ acres) (Figure 2B). Lastly, the estimated effect of hunting activity and income provided intriguing results, though the effect sizes were smaller than the two variables discussed above. Relative to those who reported no hunting activity, respondents who reported recreational hunting activity on their crop operation were 4% more likely (p < 0.001) to completely support the use of toxic bait in their state (Figure 2C). Alternatively, respondents who reported receiving net income from hunting wild pigs were 7% less likely (p < 0.001) to be completely accepting of toxic bait use, while those who reported net income from the hunting of non-wild pig species were 15% more likely (p < 0.001) to be completely accepting.

4. Discussion

The goal for this analysis was to understand acceptance levels for a wild pig toxic bait among a stakeholder group we would expect to be most in need—and therefore most accepting—of the product: producers of high-value crops with wild pigs present on their operations. Most respondents (59.6%) indicated that the use of a toxic wild pig bait in their state would be somewhat or completely acceptable to them. Unsurprisingly, this level of acceptance was higher than studies have found for the U.S. general public (23%) [8] and Texas hunters (43%) [16]. Additionally, it was approximately equal to the level of acceptance (59%) found in a survey of agricultural producers in Alabama by Ellis et al. [18]. That survey, however, included producers of resources that were less susceptible to wild pig damage, such as timber, and roughly half of the respondents reported no wild pig presence in their operations [18]. We, therefore, expected acceptance levels to be higher in the present study. However, when respondents were asked to indicate if they would feel comfortable using toxic wild pig baits on their own property if they had a problem with wild pigs, the percentage of respondents who somewhat or completely agreed with the questionnaire item increased by more than 10 percentage points to 71.4%. This finding suggests that if a toxic bait were available that reflected the criteria specified in the questionnaire (e.g., relatively quick-acting, humane, and very low nontarget risks), there would likely be a relatively high level of demand for the product among crop producers who were experiencing wild pig problems on their crop operation.

This study also revealed insights into the factors influencing the acceptability of toxic baits for wild pig management among respondents. Nearly all explanatory variables we examined were significant predictors of toxic bait acceptability, but three—operation size, WAC, and hunting activity/income—stood out because of their relatively larger effect sizes. Regarding operation size, respondents with larger-scale crop operations were more likely to favor toxic bait usage. This could reflect that the scale of economic impacts from wild pig crop damage was potentially greater in large operations than in small operations. Owners of large operations may also perceive toxic baits as a more efficient and effective strategy compared to trapping, even though previous studies have discovered the effort for both can be similar [25]. This also aligns with previous research indicating that large operations often adopt more aggressive pest management strategies [7].

Respondents with lower WAC for wild pigs, particularly those who desired complete eradication, were unsurprisingly more supportive of toxic bait usage. Indeed, this explanatory variable had the greatest effect size among the variables examined. This is consistent with other studies that have found strong relationships between WAC and management method acceptability [18]. This finding suggests that a producer’s preference for wild pig eradication or significant population reduction could influence the producer’s willingness to adopt more novel control measures. Toxic baits may be perceived by respondents as a tool that can achieve large-scale population reduction more rapidly than traditional methods like trapping or hunting, though respondents may not be accounting for the protocols associated with toxicant usage that may make it a lengthier process such as pre-baiting [25].

Regarding hunting activity and hunting-related income, we found that respondents who generated income from wild pig hunting on their land were less accepting of a toxic wild pig bait, while those who generated income from the hunting of other species were more accepting of a wild pig toxicant. This finding is understandable, as producers who profit from wild pig hunting may view wild pigs as more of a resource than a pest and, therefore, oppose lethal wild pig control measures. It highlights, however, the potential for conflict between wildlife managers who may seek to eliminate or reduce wild pig populations and producers who profit from their presence. Respondents who generated income from the hunting of other species would likely benefit from a reduced wild pig presence on their land, as research has shown that wild pigs affect the presence and behaviors of game species like white-tailed deer [26]. In fact, in a study of wild pig impacts on Texas producers, study participants reported losing white-tailed deer hunting lease revenue because wild pigs negatively affected deer presence on their land [27]. This helps explain why acceptance or use of a wild pig toxicant could be in their best interests.

Finally, we note that crop type had a varying influence on toxic bait acceptability. Producers who reported their primary crop in 2022 as sorghum, corn, or peanuts—crops that are particularly high value and vulnerable to wild pig damage [4]—demonstrated higher acceptance of toxic baits. This likely reflects knowledge or direct experience with the threat wild pigs pose to these high-value and vulnerable crops, motivating producers to seek effective control measures. On the other hand, producers of crops like soybeans and wheat, which research suggests may experience lower levels of wild pig damage [27], generally had lower levels of toxic bait acceptability. This suggests that producers’ attitudes toward toxic bait use are directly influenced by their perceived risk of economic losses due to crop damage.

5. Conclusions

Our findings suggest there would likely be significant demand for a toxic wild pig bait among crop producers experiencing wild pig crop damage, assuming the bait met the characteristics specified in our survey. These characteristics include very low risk to nontarget species and humans who consume the meat of an affected wild pig, as well as a relatively quick and humane death for the wild pig. Demand may be greatest among producers who have large operations and produce high-value and vulnerable crops, such as corn and peanuts. These findings do not consider producer acceptance under different scenarios of bait costs and difficult, time-consuming protocols associated with the use of toxic baits. Findings may be different if producers were informed of the protocols associated with the use of toxic bait [25]. Nonadopters of the product likely include producers who generate income from wild pig hunting on their land. Despite the potential demand for toxic bait, the delay in toxic bait rollout in Texas and the results of previous attitudinal surveys concerning toxic baits suggest that prior public engagement is critical to the success of any future toxic bait rollout [8,16]. Messaging should address the specific concerns of different stakeholder groups, including safety and efficacy, as demonstrated in published studies. It is important to recognize, however, that no amount of outreach and engagement will satisfy all members of the public when it comes to the use of toxic baits. Carlisle et al. [16], for example, found that one of the most common concerns of Texas hunters regarding a toxic bait was the “unknown unknowns”, with respondents pointing to examples of products that were once considered safe and are now believed to pose risks to human and environmental health. While obtaining a public consensus on the use of toxic baits may be unrealistic, engaging stakeholders and allowing for public comment and deliberation are meaningful steps toward generating the level of support necessary to move forward with any management plans that involve the use of toxic baits.