Towards an Integrated Orobanche Management: Understanding Farmers’ Decision-Making Processes Using a Discrete Choice Experiment

Abstract

Controlling the Orobanche weed parasite is a major challenge for farmers, and the individual application of various management practices has not yet proven to be successful in addressing this issue. To develop an effective strategy for managing this parasitic weed, an Integrated Orobanche Management (IOM) approach has become a priority. Using a Discrete Choice Experiment (DCE) methodology, we analyze the trade-off in farmers’ preferences between different attributes of IOM scenarios and estimate their willingness to pay (WTP). A sample of 124 Tunisian faba bean farmers participated in the study. The findings indicate that Tunisian farmers are open to adopt an IOM that includes Orobanche-tolerant faba bean varieties, and that the cost of technical package does not seem to be an obstacle. Nevertheless, farmers feel to be rewarded for delaying the sowing date from November to December. Furthermore, the study highlights that farmers show no clear preferences for the use of herbicide, specifically glyphosate, as well as for the practice of intercropping with fenugreek. While increasing faba bean yields remains a priority, farmers are willing to pay more for IOM scenarios that reduce the Orobanche plant shoot count. In conclusion, there is significant heterogeneity in farmers’ preferences, their financial situation, and the severity of Orobanche infestation significantly influencing their decision. Policy recommendations are derived from our results.

1. Introduction

Broomrapes species (Orobanche crenata, O. cumana, O. foetida, Phelipanche ramosa, etc.) are parasitic weeds that affect a wide range of important crops, including faba bean, chickpea, pea, lentil, sunflower, tomato, tobacco, eggplant, potato, carrot, and other crops. These aggressive and destructive weeds have a significant impact on agriculture, particularly in Middle East and North Africa (MENA) regions, including Tunisia [1,2,3]. The parasite depends entirely on its host for nutrition, leading to significant losses in both yield and quality of the affected crops. Despite the availability of various control technologies, including preventive measures [1], genetic approaches [3,4,5], biological [6,7,8], physical [9,10], chemical [11,12], and agronomic methods [13,14,15,16,17], the rapid spread of Orobanche remains a major threat. This is due to the diversity of host crops and Orobanche species, which continue to challenge farmers worldwide. Accordingly, the parasite affects agriculture on multiple fronts—economically, socially, and environmentally [2,18].

Abang et al. [2] and Abbes et al. [4] highlighted that while individual control measures may lead to improvement in crop yields, a reduction in infestation rates, and some decrease in Orobanche seed banks in the soil, none of the currently available control methods have proven to be completely effective methods in combatting Orobanche. This may be attributed to the limited knowledge and adoption of diverse control methods by farmers, many of whom continue to rely on ineffective management strategies or have ceased cultivating infected crops altogether. The reasons for this include the fact that not all control methods are universally acceptable, applicable, or effective, and many of the technologies developed have not been effectively disseminated [2]. Therefore, it is crucial to consider the socio-demographic characteristics of farmers and the socio-economic features of individual farming systems in order to implement effective management practices for combating the Orobanche parasite.

To develop an effective approach for controlling Orobanche and minimizing the associated damages and crop losses worldwide, particularly in MENA regions, including Tunisia, it is crucial to meet three key objectives in all farming systems: (i) preventing the spread of Orobanche seeds into non-infested areas; (ii) reducing the Orobanche seed bank in the soil; (iii) preventing Orobanche reproduction [2,19,20,21,22,23,24].

To address this challenge, an integrated weed management (IWM) program is essential for effectively combating Orobanche, which could be termed Integrated Orobanche Management (IOM). IWM is an approach that combines multiple methods simultaneously, such as chemical, biological, and mechanical techniques, specific crop management practices, and preventive measures, to effectively manage and reduce the impact of weeds in agriculture systems [25,26]. However, the adoption of these strategies has been slow among farmers. The reasons for farmers’ reluctance to embrace IWM, particularly in the case of Orobanche, are not well understood and remain underexplored [1]. Few systematic surveys have been conducted to help decision-makers, researchers, and extension services understand farmers’ perception of the Orobanche problem from the perspective of farming systems and from their socio-economic context. This understanding is crucial for defining, planning, and implementing an effective IOM strategy [27,28,29,30].

Rogers [31] commented that promoting preventive practices is often ineffective because these methods do not provide immediate benefits in the short-term. Furthermore, Llewellyn et al. [32] mentioned that marketing strategies for agrichemical products and herbicide product performance guarantees have reduced the perceived need for alternative weed control methods, like IWM.

From a socio-economic perspective, Sirinivas et al. [1] found that the low-level adoption of IOM among FCV tobacco farmers is partly due to a lack of understanding of the biological cycle of Orobanche and ignorance of existing effective management practices to control the parasite. They also pointed out that weak adoption of IWM is due to the low level of scientific orientation and failure to address the specific needs and socio-economic issues of farmers. Consequently, to increase farmers’ knowledge and improve their adoption of recommended Orobanche management practices, it is essential to consider factors such as farmers’ age, education level, landholding size, access to training, extension services, and their overall scientific orientation. In the same vein, Chèze et al. [33] pinpoint that the farmer’ willingness to change practices is significantly influenced by their income, particularly if they earn some money from farms outside. According to Danne et al. [34], there is favorable relationship between farm size and change practice behaviors. Indeed, they explained that the need for skilled labor and reliance on herbicides for timely weed control both rise with farm size.

In addition, Wilson et al. [35] and Llewellyn et al. [36] demonstrated that the short-term complexity of achieving direct effective results through the adoption of IWM, along with the associated cost in terms of time, learning, and money, have proven to be significant barriers to its adoption. Previous studies have confirmed that farmers’ knowledge, awareness, and on-farm management decisions are heavily influenced by their prior values, beliefs, and experiences, which are often based on personal needs rather than best agricultural practices [37,38,39,40]. Therefore, failing to understand and consider these factors when implementing agricultural outreach strategies is likely to hinder the effective promotion and adoption of IWM. Without addressing these underlying variables, the progress of IWM may be limited.

Consequently, to contribute to the development of an integrated and sustainable management program for controlling Orobanche and facilitate its adoption by farmers, this study aims to assess Tunisian farmers’ preferences and willingness to pay (WTP) for attributes of integrated Orobanche management (IOM) practices through a discrete choice experiment (DCE). Specifically, the primary objectives of this study are (i) to estimate farmers’ willingness to pay for the adoption of different IOM scenarios, including agronomic and chemical methods, as well as the use of resistant varieties, and to identify the key determinant attributes; (ii) to understand the decision-making process of farmers regarding on-farm management and how it influences their adoption of IOM; (iii) to identify the socio-demographic and socio-economic factors that affect farmers’ decisions to adopt IOM; (iv) ultimately, to provide policymakers with insights into farmer behavior that can be used to design more effective outreach strategies, encouraging wider adoption of the IOM approach.

This work contributes to the limited literature addressing the socio-economic factors influencing farmers’ behavior in adopting the IWM approach. It is the first study focusing on the decision-making process of farmers to increase the adoption of IOM, both globally and particularly in Tunisia. Furthermore, it is the first using a DCE to analyze the trade-off in farmers’ preferences regarding the attributes designing the IOM.

The DCE methodology is based on the economic theory of consumer choice, which has been extended to cover the non-market valuation and the medicine framework. Recently, this approach has been increasingly used to investigate farmers’ preferences and the determinant factors affecting their decision-making process [33,34,41,42,43,44,45,46]. In a DCE, respondents are presented with a series of choice situations representing different scenarios or alternatives. They are then asked to select the option they prefer in each case [47]. Therefore, the DCE is a suitable tool for presenting farmers with practically relevant decisions, where they must balance and choose between different components of an integrated management approach to control Orobanche.

2. The Tunisian Context Related to Orobanche Control

Over a century ago, Boeuf [48] first reported Orobanche infestation in Tunisia. The primary host crop for Orobanche, particularly O. crenata and O. foetida, is faba bean (Vicia faba), with yield losses in heavy infested area as much as 50–100% [3,49,50,51]. The areas infested by Orobanche and the severity of the infestation have been steadily increasing [5,52].

Consequently, a breeding program focused primarily on developing faba bean varieties tolerant/resistant to O. foetida was initiated several years ago at National Institute of Agricultural Research of Tunisia (INRAT). Since then, many faba bean varieties partially resistant to Orobanche have been registered in the Tunisian Official Catalogue of Plant Varieties such as “Najeh”, “Chorouk”, “Chams” and “Zaher” [5,50,53]. However, recent climate change has exacerbated Orobanche damages to other important crops, such as sunflower, chick-pea, and lentil, which has discouraged farmers from cultivating susceptible crops, particularly faba bean [5].

In response, the research program initiated by Field Crop Laboratory at INRAT has intensified its research activities in various areas of Orobanche control, including chemical treatments [3,54], agronomic management practices (e.g., the effect of sowing date and intercropping with trap crops like fenugreek) [19,55,56], and biological methods [8,57,58,59]. While several potential methods have been studied and disseminated to farmers, none of these approaches have proven completely effective in controlling Orobanche. The most successful strategy to combat this parasite is through an integrated management approach, primarily based on selecting genetic material with tolerance/resistance to Orobanche [2,4,5].

Unfortunately, Tunisian farmers adopting an IOM remain a minority and the behavioral factors that are crucial for understanding farmers’ decision-making are still insufficiently researched [60]. Additionally, it is puzzling that, to date, no studies have evaluated the benefit–cost of different IOM scenarios from farmers’ perspectives or examined which potential methods should be incorporated into these scenarios to increase adoption rates among farmers. For these reasons, our study aims to first examine faba bean farmers’ behaviors towards the adoption of different IOM scenarios.

3. Methodology

3.1. Integrated Orobanche Management Attributes and DCE Design

DCE is the ideal methodology for assessing the trade-offs between the different scenarios. In this study, DCE was used to present various scenarios of IOM to farmers, based in six key attributes, variety, chemical treatment, sowing date, intercropping (with two levels each attribute), technical package cost (with three levels), and output (with four levels) (

Table 1. The attributes and attribute levels of IOM scenarios.

Attributes		Levels		Status Quo
Variety		Non-tolerant		Local variety
		Tolerant
Chemical Treatment		TRT with Glyphosate		Without any TRT to control Orobanche
		TRT With Imazamox + Bentazone
Sowing Date		November		November
		December
Intercropping		Without fenugreek		Without intercropping
		with fenugreek
Package Cost (TND/ha)		[900; 1000]		[750; 850] TND/ha
		[1000; 1100]
		[1100; 1200]
Output		Yield increase	Decrease in Orobanche plant	[0–1.5] qx/ha
	Output 1	↑ 100–400%	↓ [50–75%]
	Output 2	↑ 100–400%	↓ [75–95%]
	Output 3	↑ 500–1000%	↓ [50–75%]
	Output 4	↑ 500–1000%	↓ [75–95%]

“↑” represents the increase of yields by 100–400% or by 500–1000%; “↓” represents the decrease of Orobanche plant by 50–75% or by 75–95%.

). The selection of these attributes was informed by the literature, discussions with experts involved in the Orobanche control program at the Field Crop Laboratory at INRAT, and the results of experimental studies aimed at identifying effective IOM strategies for controlling the parasite. The experimental studies were conducted over two faba bean-growing seasons and their main objective was to assess the agronomic impact of different scenarios for IOM, based on a combination of factors such as variety, chemical treatment, intercropping, and sowing date. Additionally, it is worth noting that the selected attributes are the most practical from farmers’ perspective, especially regarding preventive practices (which farmers tend to reject) and biological methods, which are unpopular and difficult to carry out.

Given the number of attributes and their levels, a full factorial design would generate 192 (3 × 2⁴ × 4 = 192) possible IOM scenarios, which would be too complex for farmers to assess. To reduce the number of combinations farmers needed to evaluate and avoid cognitive overload, we adopted the approach outlined by Street and Burgess [61] and generated an orthogonal fractional factorial design with 16 scenarios. These 16 IOM scenarios were then presented as the first option in each choice set for the respondents.

Each choice set included five alternatives: three IOM scenarios, one status quo, and one opt-out alternative. The status quo option represents the current situation of faba bean farmers in areas heavily infested by Orobanche. In this scenario, most of the farmers do not invest significant efforts in controlling Orobanche. They continue to use local susceptible faba bean varieties, typically sowing them at the beginning of November, or perhaps earlier, without applying any chemical treatments to control the parasite. Additionally, they do not use trap crops to manage Orobanche. The average cost of faba bean cropping season per hectare range from 750 to 850 Tunisian dinars (TND), and typically, their yields do not exceed 1.5 quintals per hectare (qx/ha). The other two IOM scenarios were generated using the generators (111110) and (011113) [61], resulting in a 100% efficient main-effect design. Figure 1 illustrates an example of a choice set. To make it easier for participants, we divided the 16 choice sets into two blocks. Each participant was presented with eight choice decisions, drawn randomly from one of the two blocks.

3.2. Theoretical Framework and Model Specification

The DCE modeling framework is based on both random utility theory (RUT) [62] and Lancaster consumer theory [63]. According to RUT, the value of a good is the sum of the values of its characteristics. In the context of DCE, the utility provided by alternative j (j = 1…J) from choice set s (s = 1…S) to individual i (i = 1…N) is given by the following:

$$U_{ijs}=V_{ijs}+{\epsilon }_{ijs}$$

(1)

where $V_{ijs }$ is a deterministic component and ${\epsilon }_{ijs}$ is the stochastic component. In a traditional model, the (indirect) utility can be described as a function of alternative attributes as follows:

$$V_{ijs}={\beta }_{ijks} \ast X_{kjs}$$

(2)

where $X_{kjs}$ is the vector of attributes related to alternative j; ${\beta }_{ikjs}$ is the vector of marginal utilities of the individual i related to the k attributes in alternatives j from the choice set s.

Assuming that the individuals are fully rational in their choices, the participants should be choosing the alternatives that provide them with the greatest utility. Therefore, the probability of respondents i choosing the alternative j out the total set of alternatives is:

$$P_{ij}=Prob\left[U_{ij}>U_{ik}\right]=Prob\left[V_{ij}+{\epsilon }_{ij}>V_{ik}+{\epsilon }_{ik}\right] \forall j\ne k \in s$$

(3)

By assuming that the stochastic component is distributed following type I extreme value, we obtain the familiar multinomial logit model where the probability of respondents i choosing option j from the specific choice set s is:

$$P_{ij}={\displaystyle \frac{e^{{\mu V}_{ij}}}{{\sum }_{k=1}^J{e}^{{\mu V}_{ik}}}} \forall j\in s$$

(4)

3.3. A Random Parameter Logit Model (RPL)

To estimate the marginal utility associated with each attribute, a random parameter logit (RPL) model was employed. Also known as mixed logit [64], the RPL model has the advantage of accounting for preference heterogeneity among farmers, unlike other multinomial logit models such as the conditional logit model. Specifically, the RPL model assumes that the parameter estimates associated with each attribute are not constant but instead vary across individuals, following a specific distribution, such as normal, triangular, lognormal, etc. [64].

In this model, the individual-specific parameter estimates, ${\beta }_{ij}$, are expressed as:

$${\beta }_{ij}={\beta }_j+{\sigma }_j{\vartheta }_{ij}$$

(5)

In this formulation ${\beta }_j$ is the sample mean for the alternative j, ${\vartheta }_{ij}$ is individual specific heterogeneity, with mean zero and standard deviation equal to one [65].

The probability that individual i chooses the alternative j in a particular choice set s is given by:

$${Prob}_i\left\{j is chosen\right\}=\int L_{ij}\left({\beta }_{ij}\right)f\left({\beta }_i/\theta \right)d{\beta }_i, with j\in s$$

(6)

where $f\left({\beta }_i/\theta \right)$ is the density function of the coefficients ${\beta }_i$ and $\theta$, referring to the moments of the parameter distributions, and $L_{ij}\left({\beta }_{ij}\right)$ is given by the follow equation:

$$L_{ij}\left({\beta }_{ij}\right)={\displaystyle \frac{e^{{\beta }_{ij}{X}_{ij}}}{{\sum }_{k=1}^J{e}^{{\beta }_{ik}{X}_{ik}}}} k\in s$$

(7)

3.4. Willingness to Pay (WTP)

The WTP for product or service attribute represents the monetary value that an individual is willing to pay for a unit increase in a given attribute. It can be calculated as the negative ratio of the partial derivative of the utility function with respect to the attribute of interest, divided by the derivative of the utility function with respect to the monetary variable:

$${WTP}_{Attribute}=-{\displaystyle \frac{\frac{{\partial U}_{ijs}}{\partial Attribute}}{\frac{{\partial U}_{ijs}}{\partial Monetary attribute}}}=-{\displaystyle \frac{{\beta }_{Attribute}}{{\beta }_{Monetary attribute}}}$$

(8)

3.5. Empirical Model

The deterministic component $V_{ijs}$ was commonly specified as linear in parameters including variables that represent the attributes of the IOM scenario concept and the characteristics of respondents. In the empirical specification, the deterministic component is given by:

$$\begin{gathered} V_{ijs}={\beta }_{NOP}NOP+{\beta }_{STQ}STQ+{\beta }_{VTOL}{VTOL}_{ijs}+{\beta }_{TGLY}{TGLY}_{ijs}+{\beta }_{DEC}{DEC}_{ijs}+ \\ {\beta }_{WFENG}{WFENG}_{ijs}+{\beta }_{output2}{OUTPUT2}_{ijs}+{\beta }_{output3}{OUTPUT3}_{ijs}+{\beta }_{output4}{OUTPUT4}_{ijs}+ \\ {\beta }_{tpcost}{TPCOST}_{ijs} \end{gathered}$$

(9)

In Equation (9), the attribute levels were coded as dummy variables, except the cost attribute is specified as continue variable (

Table 2. Definitions and categories of the different variables included in the empirical model.

Variable	Category	Definition
NOP	Dummy Variable	Takes the value 1 when the respondents select the opt-out option and do not prefer the other presented alternatives; 0 otherwise.
STQ	Dummy Variable	Takes the value 1 when the respondents select the status quo alternatives and do not have willingness to change their actual situation; 0 otherwise.
VTOL	Dummy Variable	Takes the value 1 when the tolerant variety attribute level is included in the presented IOM scenarios; 0 otherwise.
TGLY	Dummy Variable	Takes the value 1 when the chemical treatment with glyphosate attribute level is included in the presented IOM scenarios; 0 otherwise.
DEC	Dummy Variable	Takes the value 1 when the December sowing date attribute level is included in the presented IOM scenarios; 0 otherwise.
WFENG	Dummy Variable	Takes the value 1 when the intercropping practice using fenugreek attribute level is included in the presented IOM scenarios; 0 otherwise.
OUTPUT2	Dummy Variable	Takes the value 1 when the output of the corresponding IOM scenario consists of increasing yields by 100 to 400% and decreases in Orobanche plants m−2 to around [75–95%]; 0 otherwise.
OUTPUT3	Dummy Variable	Takes the value 1 when the output of the corresponding IOM scenario consists of increasing yields by 500 to 1000% and decreases in Orobanche plants m−2 to around [50–75%]; 0 otherwise.
OUTPUT4	Dummy Variable	Takes the value 1 when the output of the corresponding IOM scenario consists of increasing yields by 500 to 1000% and decreases in Orobanche plants m−2 to around [75–95%]; 0 otherwise.
TPCOST	Continue Variable	The cost of the corresponding IOM scenario could be one of the three alternatives: [900, 1000], [1000, 1100], and [1100, 1200] TND/ha.

).

In addition, an extended model was estimated to test the effect of two variables on farmers’ decision-making: (i) an economic variable “debt level”, this variable assesses the percentage of the farmers’ debts relative to their incomes for the current crop season. The debt level is measured on scale of five points (<10%; 10–25%; 26–50%; 50–70%; >70%); (ii) and an agronomic variable “Orobanche infestation rate”: this variable evaluates the infestation rate of Orobanche on the total area of infested crops. The infestation rate is measured on scale of five points (<5%; 6–25%; 26–50%; 51–75%; 76–100%). The goal is to understand whether a farmer’s financial situation (debt level) or the severity of Orobanche infestation plays a significant role in their willingness to adopt new practices, and if so, how these factors shape their preferences for different IOM scenarios.

3.6. Presentation of the Questionnaire and Data Collection

The questionnaire was divided into five sections. Section 1 and Section 2 focused on collecting socio-demographic and agro-economic data, respectively, from the sample. Information on crops infested by Orobanche and the level of infestation level was gathered in Section 3. Section 4, farmers’ risk preferences were assessed, with related questions included in this part of the questionnaire. Section 5 presented the choice set cards for the DCE. This section detailed 16 choice sets, which were organized into two blocks. The order of the choice sets was randomized to minimize hypothetical bias associated with learning and fatigue effect.

The survey, conducted using face-to-face interviews, took place from March to June 2024, with the majority of interviews held directly on the farms. At the beginning of each interview, the main objective of study was clearly explained. The DCE approach and IOM were then described in detail and simplified for the farmers. Next, the attributes and their levels were explained. To minimize hypothetical bias, we conducted a practices exercise with the farmers using an example of a choice card, which was different from the actual cards used in the experiment. The participants were asked to select their preferred option. Following this, the farmers were informed that the DCE section of the interview would begin. After completing the choice sets, the farmers were asked socio-demographic, agro-economic, and attitudinal questions from the remaining sections of the questionnaire.

4. Results and Discussion

4.1. Data Descriptive Statistics of the Sample

The data used in this study were obtained from a survey carried out on a sample of Tunisian’ farmers. A filter question was used to select the sample, specifying that the Agricultural Utilized Area (SAU) managed by farmers should be infested with Orobanche. As a result, 124 farmers (A small sample size is a common limitation in DCE studies targeting farmers, as this population is often more difficult to reach [33]. Several published DCE studies focusing on farmers have had small sample size. For example, Schulz et al. [66] surveyed 128 German farmers, Greiner [67] recruited 104 Australian farmers, Beharry-Borg et al. [68] collected data from 97 English farmers, Hudson and Lusk [69] included 49 American farmers, Jaeck and Lifran [70] studied 104 French farmers, and Chèze et al. [33] surveyed 90 French farmers) were recruited from nine regions to participate in this study. Over 57% of our sample is from the northwestern part of Tunisia, particularly from Beja, Jendouba, and Bizerte, which represent 27%, 19%, and 11% of the sample, respectively. This region is the main area for seed legume production, principally faba bean [71].

Table 3. Socio-demographic profile of the farmer respondents.

Variables and Levels	Frequency	Percentages
Gender
Male	121	97.58
Female	3	2.42
Age
Average (SD)	53.46 (11.71)
<40 years old	16	12.90
40 to 60 years old	63	50.81
≥60 years old	45	36.29
Education
Illiterate	16	13.11
Primary level	45	36.89
Secondary level + professional training	47	38.52
University level	14	11.48
Agriculture Training
No	93	75.61
Yes	30	24.39
Other Professional Activity
No	102	82.26
Yes	22	17.74

and

Table 4. Agro-economic profile of farmer respondents.

Variables and Levels	Frequency	Percentages
Useful Agriculture Area (UAA)
Average (SD)	33.94 (75.31)
<5 ha	35	28.23
5 to 10 ha	19	15.32
10 to 20 ha	27	21.77
20 to 50 ha	29	23.39
>50 ha	14	11.29
Land Tenure
Ownership	56	45.53
Rented	5	4.07
Sharecropping	4	3.25
Combined	58	47.15
Animal Husbandry
Yes	79	63.71
No	45	36.29
Financial Health
Highly unsatisfactory	11	8.87
Unsatisfactory	27	21.77
Not bad	63	50.81
Satisfactory	18	14.52
Higly satisfactory	5	4.03
Debt Level
[0–10%]	63	50,81
[10–25%]	17	13.71
[25–50%]	24	19.35
[50–70%]	16	12.90
>70%	4	3.23

present the descriptive statistics of the sample, including socio-demographic and agro-economic data, respectively. The majority of our sample are men, with only 2.42% of the sample being women farmers (Table 3). The average age of the participants is 53.46 years, with standard deviation of 11.71, reflecting the aging population of farmers. In fact, nearly 37% of the sample consists of farmers over 60, while just 13% are younger farmers. Half of the sample is either illiterate or has only a primary education, while just 11.41% have completed university education. Most participants have no formal agriculture training, and agriculture is their primary, if not sole, source of income. However, from Riemens et al. [72] and Sharma et al. [73] perspectives, this profile of farmers is not the good one to enhance the adoption of IWM. The authors display that full time young farmers, or farmers with limited experience, are more likely to adopt IPM strategy. This indicates that the other profile of farmers, experienced aged farmers based on their own experiences, can slow down the adoption of this strategy.

Table 4 shows that the average farm size of the participants is approximately 34 hectares, with a higher standard deviation value is around of 75, reflecting significant variability in farms size and, consequently, in farming system among participants. All farmers’ categories—from very small to large farmers—are represented in our sample. Meanwhile, it is important to note that the group most representative of the real Tunisian context is relatively tiny farmers represents 28% of the sample. However, it is important to note that between the factors that complicate the adoption of IWM was the restricted capacity to trial it on a small scale [74]. An important factor influencing farmers’ adaptation decisions and their vulnerability is their land tenure system. Table 4 also displays that the two main land tenure system adopting by our sample are the ownership and the combined system (i.e., ownership with rented land, ownership with sharecropping, or combination such as sharecropping with rented land), accounting for 45.53% and 47.15% of the sample, respectively.

Additionally, one of the main preventive practices for limiting the spread of Orobanche on farms is controlling animal grazing movement. Therefore, it is important to understand the extent to which farmers engage in animal husbandry activities. Indeed, 63% of the sample practice mixed farming, combining cropping system with livestock farming. The results shown in Table 4 reflect the heterogeneity of the farming systems adopted by Tunisian farmers. However, this variability could be a limiting factor for adopting IWM, as it may have lower compatibility with existing farming systems [75]. Furthermore, it is noteworthy that over 60% of participants perceive their financial health positively and report little or no debt relative to their income levels. This suggest that most of the sample view their financial situation as satisfactory. Consequently, financial constraints are unlikely to prevent farmers from adopting Orobanche control practices or IOM approach.

Table 5. Comparison of sample socio-economics profile with Tunisian farmers statistics.

Variables and Levels	Sample Statistics (%)	Tunisian Farmers’ Statistics (%)
Age
Average (SD)	53.46	50–60
<40 years old	12.90	13
40 to 60 years old	50.81	44
≥60 years old	36.29	43
UAA: 0 to 20 ha	65.32	83
Education
Proportion that finished or not the secondary level	88.52	Around to 80%
Proportion having a higher education level	11.48	Around to 10%

compares the key socio-economic characteristics of our sample with those of the broader population of Tunisian farmers. The table shows that our sample is representative in terms of age, useful agriculture area, and education level.

Additional descriptive statistics related to faba bean cropping and Orobanche infestation rate are displayed in

Table 6. Current faba bean farming, Orobanche infestation, and farmer respondents’ knowledge level of Orobanche management practices.

Variables and Levels	Frequency	Percentages
Faba bean cropping
Yes	86	69.35
No	13	10.48
I stopped	25	20.16
Orobanche infestation level
<5%	21	16.94
6 to 25%	18	14.52
26 to 50%	24	19.35
51 to 75%	37	29.84
76 to 100%	24	19.35
Knowledge level
Not at all informed	64	52.46
More and less informed	29	23.77
informed	19	15.57
Fairly informed	6	4.92
Well informed	4	3.28

. The data revealed that 20% of our sample has stopped growing faba bean, due mainly to Orobanche infestation. Moreover, approximately 70% of the participants reported that the infestation level in their area is moderate to heavy. However, it is surprising to observe that more than 76% of the participants have little to no information about Orobanche management practices or how to control this plant parasite.

4.2. Farmers’ Preferences Towards IOM Attributes

In the DCE data analysis, the RPL model is used to estimate farmers’ marginal utilities associated with attribute and attribute levels. Moreover, the model allows for correlation between alternatives by estimating the full covariance matrix structure. Following Burton [76], when using an RPL model with dummy variables, such as in our case, it is important to account for the correlation between random parameters to avoid an increase in Type-I errors.

Table 7. Estimated parameters results of the three RPL models, standard RPL, RPL with correlation, and RPL with correlation and interactions factors.

	Model 1	Model 2	Model 3
Variables	Coefficients (SE)
Opt-Out (NOP)	−3.121 *** (0.955)	−4.029 *** (0.975)	−4.535 *** (1.198)
Status Quo (STQ)	−4.280 *** (0.643)	−5.202 *** (0.677)	−5.139 *** (0.699)
Variety Tolerant (VTOL)	0.522 ** (0.221)	0.319 * (0.189)	1.113 *** (0.304)
Chemical Treatment—Glyphosate (TGLY)	−0.003 (0.135)	0.053 (0.126)	0.118 (0.151)
Sowing date (mid-December) (DEC)	−0.876 *** (0.174)	−1.141 *** (0.203)	−1.089 *** (0.183)
Intercropping–Fenugreek (WFENG)	−0.062 (0.132)	−0.493 ** (0.167)	−0.097 (0.164)
Output 2	1.465 *** (0.194)	1.837 *** (0.303)	2.790 *** (0.464)
Output 3	0.287 (0.255)	0.938 ** (0.323)	1.569 *** (0.484)
Output 4	2.514 *** (0.235)	2.784 *** (0.300)	3.406 *** (0.475)
Technical Package Cost (TPCOST)	−0.001 ** (0.0008)	−0.001 ** (0.0008)	−0.002 ** (0.001)
Variables	Standard deviations (SE)
VTOL	1.886 *** (0.181)	2.474 *** (0.233)	2.449 *** (0.229)
TGLY	0.751 (0.529)	0.432 ** (0.187)	1.145 *** (0.215)
DEC	1.412 *** (0.186)	1.732 *** (0.213)	1.953 *** (0.241)
WFENG	0.961 *** (0.171)	1.661 *** (0.225)	2.113 *** (0.253)
Output 2	0.190 (0.950)	2.424 *** (0.399)	1.956 *** (0.360)
Output 3	1.675 *** (0.390)	2.229 *** (0.349)	2.491 *** (0.331)
Output 4	1.504 *** (0.381)	2.532 *** (0.326)	2.921 *** (0.420)
Interaction Factors	Coefficients (SE)
Debt Level (Debt) × VTOL	--	--	−0.905 ** (0.392)
Debt × Output2	--	--	−1.004 ** (0.489)
Debt × Output3	--	--	−0.100 (0.534)
Debt × Output4	--	--	−0.175 (0.502)
Debt × TPCOST	--	--	0.001 ** (0.0005)
Infestation Level (TINFEST) × VTOL	--	--	0.741 ** (0.361)
TINFEST × Output2	--	--	−0.089 (0.471)
TINFEST × Output3	--	--	−0.996 * (0.531)
TINFEST × Output4	--	--	−0.298 (0.479)
TINFEST × TPCOST	--	--	−0.001 ** (0.0005)
Goodness of fit
Number of Observations	4960
Degree of Freedom (Df)	17	38	48
Akaike Information Criterion (AIC)	1893.163	1805.713	1769.802
Log-likelihood Ll (null)	−1070.215	−1070.215	−1056.648
Log-likelihood Ll (model)	−929.5813	−864.713	−836.9009

***, ** and * indicate that the corresponding parameter is statistically significant at the 1%, 5% or 10% level, respectively.

presents the parameter estimates for three estimated models. The first model (model 1) estimates a standard RPL, the second model (model 2) accounts for the correlation between alternatives but excludes interaction factors, and the third model (model 3) estimates a RPL that includes both correlation between alternatives and interaction factors.

As commonly assumed in the literature [65], the marginal utility associated with the cost attribute of the technical package is considered constant, while the other attributes are treated as random parameters following a normal distribution (Table 7).

The goodness-of-fit indicators show that the model 3 provides the best to the data. This model has the lowest values for both Akaike Information Criterion (AIC) and log-likelihood criteria (Ll), in absolute values. Therefore, for the remainder of the paper, we will focus on the results from this model.

Model 3 provides estimates of the mean of marginal utilities associated with the attributes’ levels included in the DCE, as well as the standard deviations for the random parameters considered (VTOL, TGLY, DEC, WFENG, OUTPUT2, OUTPUT3, and OUTPUT4). To further explore sources of heterogeneity and better understand farmers’ choices, we analyzed potentially influencing factors by examining the interactions between “debt level (Debt)” and “infestation rate (TINFEST)” with the attribute levels (VTOL, OUTPUT2, OUTPUT3, OUTPUT4 and TPCOST).

The results show that the coefficients for both STQ (Status Quo) and NOP (Opt-Out) choices are negative and highly significant. This suggests that the farmers in the sample tend to reject the status-quo or for the no-option alternative when presented with IOM scenarios. In general, participants seem willing to accept the various IOM scenarios and their associated characteristics.

The statistically significant positive coefficients for the variables relative to faba bean tolerant varieties to Orobanche (VTOL), as well as they associated with the outputs of the different IOM scenarios (output 2, output 3, and output 4), indicate that farmers are more likely to accept IOM scenarios that include, primarily, tolerant varieties for Orobanche. These results confirm findings of previous studies, which emphasize that host plant resistance is a key element in the fight against Orobanche, although it is not an effective standalone control strategy [2,4,5]. Lamichane et al. [77] confirm that a possible way to enhance effectiveness IWM is by incorporating resistant crop varieties, along with cultural practices, physical and mechanical tactics, chemical control, and others methods.

Additionally, the results reveal that farmers do not present preferences for using Glyphosate as a chemical treatment to control Orobanche. The coefficient associated with the attribute level (TGLY) is not statistically significant. Danne et al. [34] highlight that farmers have no clear preferences for glyphosate use. Intensive research has identified herbicides with good potential for controlling Orobanche [11], including Glyphosate. However, several obstacles limit the successful use of herbicide, such as its limited persistence, high costs, lack of it approval as herbicide for Orobanche control, like our case in Tunisia [2]. Despite promising results from research conducted by Kharrat and Halila [3], which demonstrated the effectiveness of very low doses of Glyphosate in controlling Orobanche, this selective herbicide has not yet been registered for use against the parasite in Tunisia. Doole and James [75] highlighted that the drivers of herbicides broad scale use by farmers are their high observable efficacy, low cost, low complexity, and significant flexibility. However, Tunisian farmers seem to be wary of using this selective herbicide, as applying incorrect doses could lead to serious crop yield losses. This underscores the critical role of extension services in providing farmers with the necessary, accurate information and effective practices for managing Orobanche [2]. Abang et al. [2] add that after 15 years promoting the use of Glyphosate to control O. Crenata in faba bean, in Morocco, only 15% of the interviewed extension agents were able to give a correct description of its application technology.

Similarly, the coefficient associated with the attribute level for intercropping with fenugreek (WFENG) is not statistically significant. This result indicates that the majority of the participants do not appear to be convinced of using fenugreek as a trap crop to control Orobanche. Despite the effectiveness of growing trap crops, particularly fenugreek, in reducing Orobanche shoot counts [2,13,19], Tunisian farmers seem to be unaware of this practice. This lack of awareness may be due to the challenge of managing both crops simultaneously, faba bean and fenugreek, during the sowing and harvest phases.

Moreover, the results reveal negative marginal utilities for sowing date change attribute in faba bean cropping. The estimated parameters for this attribute level (DEC) is negative and highly statistically significant. In line with studies on chickpea cropping, Rubiales et al. [14], Van Hezewijk [78], and Kebreab and Murdoch [79] have demonstrated that Orobanche infection is favored by early sowing dates (October-November), mild winters, and rainy autumns and springs. Furthermore, Mesa-Garcia and Garcia-Torres [80] highlighted that the number of Orobanche plants successfully attaching to the host, as well as the duration of the underground stage of Orobanche, decreases as broad bean planting is delayed. However, it appears that the recruited farmers do not consider this practice to manage Orobanche and may not be fully informed about the optimal conditions for delaying the sowing dates.

Upon examining the estimated parameters associated with the output attributes presented in Table 7, it is clear that participants value the outputs levels of different IOM scenarios. All the marginal utilities for OUTPUT2, OUTPUT3, and OUTPUT4 are positive and highly significant. Additionally, it is noteworthy that the marginal utilities for OUTPUT2 and OUTPUT4 are more important than the marginal utility associated with OUTPUT3. These results suggest that, given the serious damage suffered by faba bean farmers, increasing yields is a very important goal. However, reducing the Orobanche shoot count appears to be even greater significance.

In the same vein, the results indicate that the cost of technical package is not a serious barrier for adopting IOM scenarios for Orobanche control. Table 7 shows that the coefficient associated with the attribute “technical package cost (TPCOST)” is negative and significant, but very low, at approximately −0.002. This suggests that, although participants show a slight negative utility towards the cost of technical package, it is not a major deterrent, likely due to the serious problems they face with Orobanche.

Furthermore, Section 2 of Table 7 includes the standards deviation of all random parameters considered in the model: VTOL, TGLY, DEC, WFENUG, OUTPUT2, OUTPUT3, and OUTPUT4. All the standard deviation coefficients are important and highly significant. This indicates that preferences for the IOM scenarios attributes and their respective levels vary across the participants. To better understand the source of this heterogeneity, two potentially influencing factors were introduced in the model interacting with various attribute levels that define the IOM scenarios.

4.3. Influence of Farmers’ Debt Level

The primary concern of this study regarding the adoption the IOM scenarios by farmers is to redesign cropping systems in a way that minimizes the impact of the Orobanche parasite, ensuring that its presence does not negatively affect yields and farmers’ profitability. Nonetheless, IWM, and particularly IOM, has been reported to be expensive due to its combination of multiple methods, including chemical, biological, mechanical, and specific crop management techniques [25,72,81]. Swanton et al. [82] demonstrated that IWM systems can be perceived as unreliable, increasing management risk. Furthermore, there is no clear, direct economic benefit from IWM, and sustained support for its adoption has been limited.

Given this, it can hypothesize that farmers’ financial conditions might act as a constraint to adopting such an approach. The literature suggests that the annual income level of farmers is a key factor governing the decision to adopt IWM [83,84]. Specifically, higher income levels increase the likelihood of farmers adopting IWM. However, Houngbo et al. [45] present a contradictory finding, showing that farmers who are more likely to apply integrated pest management (IPM) tend to have relatively low incomes. Increased debt level raise farming costs, reduce profitability, and negatively affect annual incomes. Therefore, consistent with the finding of Houngbo et al. [45] it can be argued that farmers with higher debt levels (and relatively low incomes) are more likely to adopt IOM strategies.

The results presented in Table 7 confirm the findings of Houngbo et al. [45]. Specifically, the interaction terms “DEBTxVTOL” and DEBTxOUTPUT2” are negative and statistically significant. This indicates that farmers with very low levels of debt are less likely to choose IOM scenarios involving Orobanche-tolerant varieties. Several factors may explain this result, including the higher cost of tolerant varieties compared to non-tolerant and local ones, limited access to these varieties in the market, and the weak dissemination efforts aimed at promoting them [2].

For these farmers, increasing faba bean yields is a top priority. They are less inclined to choose alternatives that offer only modest improvements in yields, even if such alternatives reduce Orobanche shoot count. Furthermore, this group of farmers appears to be more tolerant to the technical package cost than those with higher levels of debt. The coefficient of the interaction term “DEBTxTPCOST” is positive and statistically significant (Table 7), suggesting that these farmers do not face significant financial constraints in adopting IOM strategy, unlike the majority of farmers, for whom cost is a major obstacle.

Therefore, a targeted dissemination strategy aimed at educating this group of farmers about the principles and practices of IOM, particularly the role of Orobacnhe-tolerant varieties, could enhance their acceptance and adoption of IOM practices.

4.4. Influence of Infestation Rate

Orobanche infestation continues to rise, a significant threat to the livelihoods of millions of farmers. In Severe cases, infestation can lead to 100% crop failure. Therefore, the objective of this section is to understand how Orobanche infestation rates influence farmers’ preferences for IOM attributes and affect their decision-making process. Horowitz and Lichtenberg [85] emphasized that the level of infestation is one of the key sources of risk in making informed decisions about pest/parasite control practices. Similarly, Lopez-Granados and Garcia-Torres [86] highlighted Orobanche infestation as a critical factor in determining weed management strategies.

The results presented in Table 7 underscore the significant influence of Orobanche infestation rates on farmers’ perception. Specifically, the coefficient for the interaction term “TINFEST × VTOL” is positive and statistically significant. This finding suggests that farmers currently experiencing high Orobanche infestation rates have positive preferences for the inclusion of tolerant varieties in IOM scenarios. This outcome aligns with the conclusions of Horowitz and Lichtenberg [85] and Lopez-Granados and Garcia-Torres [86], confirming that higher infestation rates increase farmers’ willingness to adopt strategies that incorporate Orobanche-tolerant varieties.

Nonetheless, the estimated parameter for the interaction term “TINFEST × TPCOST” is negative and statistically significant. This indicates that farmers with higher Orobanche infestation rate are more sensitive to the technical package cost of IOM scenarios. In contrast, the interaction terms between infestation rate and the output attribute levels “TINFEST × OUTPUT2”, TINFEST × OUTPUT3”, and “TINFEST × OUTPUT4” are not statistically significant. This suggests that farmers experiencing significant Orobanche infestation do not have strong preferences for any particular output level of IOM scenarios. Instead, their primary concern is to adopt effective practices that can immediately reduce Orobanche shoot plants.

Consequently, this section allow us to conclude that it is essential to provide special attention to the farmer segment that is suffering high Orobanche infestation rates in order to successfully disseminate the many faba bean varieties that are tolerant or resistant to Orobanche. It is important to emphasize how growing these varieties can reduce Orobanche infestation when used as a determinant factor in an IOM strategy.

4.5. Willingness to Pay

Table 8. Farmers’ willingness to pay (WTP) for IOM attributes.

	VTOL	DEC	Output 2	Output 3	Output 4
WTP	418.009	−409.149	1047.665	589.457	1278.833
IC	[10.13; 825.88]	[−748.85; −69.43]	[128.62; 1966.70]	[2.67; 1176.24]	[197.35; 2360.30]

IC: Confidence Interval.

presents the farmers’ willingness to pay (WTP) estimates derived from the considered estimated model (model 3), along with their confidence intervals (IC). The results indicate that Tunisian faba bean growers are willing to pay 418 Tunisian Dinars (TND) per hectare (ha) for Orobanche-tolerant faba bean varieties over sensitive varieties. Furthermore, growers are willing to pay a premium ranging from 590 TND/ha to 1280 TND/ha to adopt IOM scenarios with outputs, OUTPUT2, OUTPUT3 and OUTPUT4 compared to the base IOM scenario with OUTPUT1.

It is important to note that the IOM output attribute is a composite measure combining both yield improvements and Orobanche plant shoot reduction (see Table 1). Accordingly, the results demonstrate that farmers are willing to pay a premium or around 590 TND/ha for an IOM scenario that significantly increases faba bean yields from 500% to 1000% while maintaining the same reduction level in Orobanche plant shoot count (up to 75%) as the base level (OUTPUT1). However, the premium that farmers are willing to pay for the IOM scenario with OUTPUT2, which offers a 95% reduction in Orobanche shoot plant count while maintaining the same level of yield improvement, is approximately 1050 TND/ha. This premium is almost double the amount they are willing to pay for OUTPUT3. These findings suggest that farmers are highly concerned with Orobanche infestation rates.

Moreover, the highest premium farmers are willing to pay is for the IOM scenario with OUTPUT4, at approximately 1279 TND/ha. This premium is more than double the amount for OUTPUT3 and slightly higher than the premium for OUTPUT2. Accordingly, farmers are concerned about crop yield; however, their primary concern is the infestation rate of Orobanche. In conclusion, the farmers in our sample are willing to pay a substantial premium for IOM scenarios that focus first on reducing Orobanche plant shoot count, and then, improving crop yields.

Finally, it is important to note that our sample perceives being compensated for changing the sowing date from November to December. Table 8 displays that farmers’ WTP for DEC attribute is negative and significant, with an amount of −409 TND/ha. This confirms our results, indicating that recruited farmers are unaware about the good effects of this agronomic practice and the appropriate conditions needed for success.

5. Conclusions

Several factors play a crucial role in farmers’ decision-making when selecting an Orobanche management approach. The present study examines Tunisian farmers’ preferences and their WTP for attributes of integrated approach to control this parasitic that continues to threaten their incomes. The findings show that our sample of faba bean growers consists mostly of older men with low educational level, many of whom are illiterate. While they are familiar with Orobanche and its potential damage, they have limited knowledge about parasite life cycle and the practices that could be used to control it. Approximately, 20% of the farmers in our sample have ceased growing faba beans due to the damage caused by Orobanche, and half of the sample suffers from moderate to severe infestations. These results suggest that the number of farmers abandoning faba bean cultivation is likely to increase unless an effective control strategy is implemented. For this strategy to succeed, it must involve all relevant stakeholders, particularly the farmers and extension services.

Tunisian farmers appear to be interested in the integrated scenarios for controlling Orobanche presented in this study, however, there is significant preferences heterogeneity across the sample regarding the different attributes and attribute levels defining the IOM. The majority of farmers expressed positive preferences for the inclusion of Orobanche-tolerant faba bean varieties in the IOM scenarios. In contrast, they were more reluctant to combine control methods with a delayed sowing date. Farmers did not show significant preferences for the use of chemical treatments or intercropping practices with trap crops like fenugreek to control Orobanche. This reluctance could primarily be attributed to the lack of technical knowledge and the gap in their understanding of these two practices. Most farmers have difficulty to manage properly the chemical products recommended for Orobanche control, as they lack knowledge about the correct doses, timing, and number of applications required. Similarly, they are unfamiliar with the proper sowing and harvesting techniques for intercropping practices with fenugreek. As a result, farmers appear to be hesitant to adopt these control methods. Furthermore, the study found that farmers were more willing to pay an important premium ranging from 1050 to 1280 TND/ha, for IOM scenarios that focused primarily on reducing Orobanche plant shoot counts, rather than improving faba bean yields.

The findings suggest that, to facilitate the adoption of IOM approach by farmers, a target promotion strategy is necessary, with a focus on two key attributes: the availability of Orobanche tolerant faba bean varieties, and the effectiveness of IOM scenarios in controlling Orobanche and reducing crop losses in infested areas. Furthermore, following Gonzalez-Andujar [25], there is a still need to strengthen the efforts of all stakeholders involved, such as research institutions, extension services, and interprofessional groups, to disseminate research findings. It is crucial to simplify information and clarify the technical aspects of control methods like intercropping with trap crops, delaying sowing date, and applying chemical treatments, to ensure farmers can effectively implement them. Additionally, decision-makers should prioritize accelerating the herbicide registration process for Orobanche control, which could help overcome the barriers that have hindered the successful use of some herbicides like Glyphosate and other containing efficient molecules from Imidazolinone families.

To conclude, it is noteworthy to emphasize that a successful communication strategy targeting farmers should be developed using a participatory approach. This strategy must address several key pillars: (i) understanding the vulnerability of farmers, which can manifest in various forms like financial insecurity and relatively low incomes, low education levels, high Orobanche infestation, diverse land tenure systems, and local farming practices; (ii) identifying distinct farmers groups, understanding their concerns, selecting appropriate communication tools, and tailoring messages to each group’s specific needs; (iii) crafting clear and relevant messages that raise farmers’ awareness about IOM scenarios, while considering their local conditions, concerns and priorities; (iv) diversifying knowledge transfer methods and communication tools (roundtable, conferences, information days, scientific coffee, simplified brochures, recurring short messages, applications, social media, farmer-to-farmer network, etc.); and (v) ensuring that adequate resources (financial, technical, and logistical) are allocated to effectively implement and sustain the strategy.

Future research should focus on defining the characteristics of the appropriate communication strategy taking into account the priorities, context, socio-economics characteristics, and farmers’ psychological behaviors. This shift towards a more comprehensive approach to control Orobanche will play a crucial role in ensuring the long-term viability of farming systems and sustain farmers’ resilience. However, additional future research focusing on assessing the long-term cost-effectiveness and financial sustainability of IOM for farmers is crucial. Ultimately, various factors are determinant in farmer’s process decisions; however, little attention has been paid to the effect of farmers’ risk perception or risk aversion.