Adoption of Roguing to Contain Banana Bunchy Top Disease in South-East Bénin: Role of Farmers’ Knowledge and Perception

ABSTRACT

Banana bunchy top disease (BBTD) caused by Banana Bunchy Top Virus (BBTV) has emerged as a major constraint of banana and plantain production in south-East Bénin. Roguing method is used in this area to destroy the BBTV-infected plants and to contain its spread. This study was conducted to assess the level of adoption of the method in relation to farmers’ perception. The research was conducted in eight communities and data collected from 186 randomly selected banana growers. Farmers’ perception on the roguing method as compared to traditional practices was analyzed using an aggregated perception index. The counterfactual method based on Average Treatment Effect was used to estimate adoption rates and determine the factors influencing adoption. Results showed that roguing is perceived to provide more advantages than traditional practices, but it is often time consuming and very labor demanding. The actual adoption rate is 36%. It would have been 56% if all the growers sampled were sensitized. As a matter of fact, the majority (73%) of sensitized growers adopted the method. More efforts and resources should be put into sensitizing and mobilizing farmers on the adoption of this technology while improving its efficiency and taking account the shortcomings reported by farmers. This, coupled with an increase in easy access to healthy planting materials and use of good production practices could improve the adoption of roguing and consequently ensure the effective containment of BBTD.

KEYWORDS:

• Adoption
• BBTD
• roguing
• farmers’ perception
• healthy planting materials

Introduction

Banana and plantain are staple food in Africa and produced year-round, providing an extremely valuable source of food during the lean season (Sodédji et al., 2016: 314). It procures more food per area than any other staple crop, generates lower production costs, and is not subject to price fluctuations unlike rice, maize, and wheat (Atkinson et al., 2015: 1). Mainly traditional and cropped on a small scale, banana is an important source of income for farmers even as a “backyard crop” (Adigbonon and Allagba, 2015: 14) indicating the crucial role of the crop in food security for thousands of households (Zandjanakou-Tachin, 2009: 1). In the main growing areas of the crop in Bénin, high yield losses and production of unmarketable fruit have been reported (Sodédji et al., 2016: 314). Several factors cause these losses. The most important are pests and diseases, which cause significant losses and a reduction of area allocated to banana affecting gross margins of the different actors involved in the crop’s value chains (Pemsl and Staver, 2014: 1). Caused by Banana Bunchy Top Virus (BBTV), Banana Bunchy Top Disease (BBTD) is the most economically important viral disease of banana (Wang and Hooks, 2011:1; Ramathani and Beed, 2013:110; Mukwa et al., 2014: 648). It causes yield losses up to 100% and a significant reduction of cropped areas (Sodédji et al., 2016: 314). There are two means of spread of the BBTV: through aphid vector (Pentalonia nigronervosa) and use of infected planting materials moved from infected areas to non-infected areas (Mukwa et al., 2014: 648). BBTD has been reported in 14 countries in Africa including Bénin (CGIAR, 2014: 7), where the first report of the disease was made in 2012 (Lokossou et al., 2012: 13). This is an evidence that BBTD is a threat to food security affecting the income and food of millions of farmers, traders, and consumers (Ramathani and Beed, 2013: 109) and increasing their vulnerability.

To control the disease several strategies are being triggered to limit the impact of its spread. These include quarantine measures, restriction of the movement of banana planting material from BBTD-affected areas, regular inspection of plantations, education and awareness programs for farmers, and the rapid destruction of infected plants (Iskra-Caruana, 2003:14; Kumar et al., 2011: 180). Two methods of eradication of infected plants are available: the chemical method by injection of glyphosate (Wang and Hooks, 2011: 1) and the roguing method (mechanical method) (Iskra-Caruana, 2003: 14). In the framework of its actions, the BBTD project sensitized farmers on the use of roguing in South-East Bénin (Sodédji et al., 2016:319; Zandjanakou-Tachin et al., 2017: 1). Associated with many other measures (registration of every banana plantations, establishment of quarantine zones, restriction of planting materials movement and use, regular inspections, implementation of trainings and extension programme), roguing has contributed to the complete rehabilitation of banana industry after a slump due to BBTV in Australia (Iskra-Caruana, 2003: 14). It has helped reducing the disease incidence and the risks of new outbreaks (Allen, 1978: 535). Adopting roguing could thus help to effectively control BBTD in Bénin and boost banana productivity countrywide. However, the technical performance of the innovation is unfortunately not sufficient to justify its relevance and guarantee its adoption. It is the perceptions of adopters on the attributes of a given innovation that affect its rate of adoption (Rogers, 1983: 212). As highlighted by the same author, innovation, like beauty, exists only in the eye of the beholder and it is its perception that influences behavior. Thus, the present study aligns well with this assumption as it focuses on the rates and determinants of adoption of the roguing method taking into account farmers’ perceptions.

The theoretical framework of this study is built on the counterfactual approach based on the calculation of the Average Treatment Effect (ATE) developed by Diagne and Demont (2007: 202) to estimate, without bias, the rates and the factors determining technologies’ adoption. This parameter corrects “non-exposure” or selection bias, which common adoption estimators suffer from, assuming one cannot adopt a technology which they do not know exists (Mwangi et al., 2014: 227). This approach was combined with the perception index developed by Rahman (2003: 34) and adapted by Vidogbéna et al. (2015: 9) to assess the perception of technology in comparison with endogenous strategies, a very important element influencing adoption. This study therefore adds on to the existing literature on adoption of technologies in relation to farmers’ perceptions. It paves a way for policy development to guide future project interventions to improve technology and awareness campaigns for a better adoption and a more successful disease control, so contributing to improvement of banana productivity in Bénin.

The containment and eradication activities are emergency measures applied to a pest outbreak. The purpose of an eradication is to arrive at a zero percent infection leading to the establishment of a Pest Free Area (PFA). As for the containment, the purpose is to prevent economic damage by the pest or to limit the area in which it already causes damage. A successful containment of an outbreak is an opened-door for a possible eradication and re-establishment of a PFA (Black, 2004). Within the framework of this study, eradication is used in the context of diseased plants destruction at a small-scale (field) level. Thus, roguing as method of eradication is used to destroy the infected plants after detection. It is a key step in the process to limit the spread of BBTD. This method coupled with others preventing measures as the use of BBTV-free planting materials, the monitoring, and the control of movement of infected planting materials to control the spread can help to contain the disease and in the long-run lead to the large-scale (national) eradication.

Methodology

Data Collection

A survey was conducted from September to October 2016 in the Communes of Adjarra and Akpro-Missérété in South-East Bénin (Figure 1).

Figure 1. Map of South-East Bénin presenting the communes of Adjarra and Akpro-Missérété.

The study used a criteria-based approach to select the villages. These criteria include the economic importance of banana in households (Adigbonon and Allagba, 2015: 14), the major constraint to banana production (Banana Bunchy Top Disease) (Adigbonon and Allagba, 2015:45; Adognon and Medenou, 2015: 34), and the sensitization on the use of the roguing method (Sodédji et al., 2016:316; Zandjanakou-Tachin et al., 2017: 1). The roguing method was disseminated in Adjarra and Akpro-Missérété by the BBTD project funded by the International Institute of Tropical Agriculture, Ibadan (IITA-Ibadan), Nigeria in association with the CGIAR, Bioversity International, and the Center for International Cooperation in Agricultural Research for Development (CIRAD).

Primary data (quantitative and qualitative) were collected in two phases. First, focus groups of seven to eight people were organized to i) identify the characteristics or criteria farmers consider in choosing any pest and disease control method and ii) identify the endogenous practices or methods farmers used to control BBTD before the sensitization on the use of roguing. This information was used to refine the structured questionnaire administered to individual farmers sampled during the second phase of the investigation. The two-level stratified sampling method was used with the village and household representing the first and second stratification level, respectively. In each commune, two target villages of the project and two villages where the project did not intervene were selected. The villages of intervention chosen are those in which sensitization on the use of the method was conducted. The approach also took into account that the two categories of villages chosen are far from each other to avoid any bias due to the exposition to the technology. The survey was then carried out in eight selected villages. A total of 186 banana growers were randomly chosen from a list of farmers obtained from the BBTD project which was completed by a preliminary survey. Data were collected on socio-demographic and economic characteristics of the respondents, the knowledge and use of the method, and farmers’ perception of the method in comparison with the endogenous strategies used.

Empirical Framework

Perception of Roguing Compared to Endogenous Practices

The Aggregated Perception Index of the ith respondent was estimated to compare farmers’ perceptions on roguing and endogeneous practices. The empirical model used was (Rahman, 2003:34; Vidogbéna et al., 2015: 9):

(1) $I{P}_i=\sum _{j=1}^J\sum _{m=0}^M\sum _{Q=0}^Q{E}_j{R}_m{W}_q$(1)

With, j = 1 … J, m = 0 … M, q = 0 … Q,

E_j: Attributes of the technology, R_m: Ranking score assigned to each attribute considering its relative importance on a five-point scale W_q: Weights assigned to each attributed after conversion of scores. A weight of 0.2 is assigned for the lowest rank of 1 and a weight of 1 is assigned for the highest rank of 5.

The Aggregated Perception Index of the sample is estimated as (Vidogbéna et al., 2015: 9):

(2) $\mathrm{G}\mathrm{P}\mathrm{i}=\mathrm{I}\mathrm{P}\mathrm{i}/\mathrm{N}$(2)

N = Size of the sample.

According to Vidogbéna et al. (2015: 9), an overall perception index of at least 0.70 indicated that roguing was preferred over endogenous practices. For an overall perception index between 0.60 and 0.69, we assumed that the two methods would provide the same satisfaction to the farmers.

Rates and Factors Determining Adoption of Mechanical Roguing

We used the counterfactual method based on Average Treatment Effect (ATE) to estimate actual and potential adoption rates and the factors determining the adoption of roguing. This approach is used to correct the “non-exposure” or ”selection” bias (Diagne and Demont, 2007: 201), and is used where every farmer in the population has two potential outcomes: with and without exposure to the technology. The exposure is the awareness of a technology’s existence and is critical in determining adoption of a technology. The exposure variable in this study accounts only for mere awareness of the technology’s existence. Assuming that $y_1$ is the potential outcome of a farmer when he is exposed to the technology and $y_0$ the potential outcome when he is not exposed, the potential outcome of adoption may either be adoption status (dichotomous 0‒1 variable) or a measure of intensity of adoption (affected crop area on which the method is used). The treatment effect for a farmer i is measured by the difference$y_{i1}-y_{i0}$. Therefore, the expected population adoption impact of the technology given by $E(y_1-y_0)$, is referred to as ATE. But, the difficulty to observe both an outcome and its counterfactual makes it impossible to measure $y_1-y_0$for any farmer. However, since exposure to a technology is a condition for its adoption, $y_0=0$ for any given farmer. Therefore, the impact of the exposure of a given farmer to the method is $y_{i1}$ and the average adoption impact is given by $ATE=E{y}_1$. Unfortunately, $y_1$ is only observed for farmers exposed to the method. Hence, the expected value of $y_1$ cannot be estimated with the average of a randomly selected sample, since there will be some missing values of $y_1$ (Diagne and Demont, 2007: 202). The potential adoption rate gives an estimation of adoption rate when all the population is aware of the existence of a technology. It does not have to be necessarily 100% since, not all the farmers for instance adopt once there are exposed to a technology. Some farmers choose not to adopt the technology even after being aware of its existence (Mwangi et al., 2014: 228). The explanatory variables in the model and their expected signs are presented (Table 1).

Table 1. Descriptive statistics of variables used in knowledge and adoption models.

Variables	Description	Expected signs	Mean	Standard deviation	Min	Max
Age	Age of the respondent (years)	±	48.110	15.960	19	80
Sexe	Gender of the respondent (0 = Female; 1 = Male)	+	0.661	0.475	0	1
Tailm	Size of the household	+	7.065	3.331	1	17
Instru	Level of education of the respondent (0 = No, 1 = Primary, 2 = Secondary, 3 = High school, 4 = University)	+	1.522	0.890	1	5
supban	Banana farm size (ha)	±	0.040	0.040	0.01	0.20
vulgar	Access to extension services (0 = No; 1 = Yes)	+	0.317	0.467	0	1
Form	Participation to at least one training on banana production (0 = No; 1 = Yes)	+	0.199	0.400	0	1
Rapaff	Annual banana income (FCFA)	+	104 089	76 964	8 000	500 000
Zone	Location (0 = control; 1 = pilot)	+	0.527	0.501	0	1
pergbon	Overall perception on the technology compared to endogenous practices (0 = Endogenous practices are perceived more advantageous than the technology/The two methods provide the same advantages; 1 = The technology is more advantageous than endogenous practices)	+	0. 285	0.453	0	1

Results and Discussion

BBTD Control Options Used in the Area of Study

Several strategies were developed by the surveyed farmers to control BBTD. These include: simple cutting (75.14%), simple uprooting (17.51%), uprooting plus manure (1.69%), cutting with chopping of the remaining pseudostem (2.26%), cutting plus manure (0.56%), cutting plus ash (1.13%), shifting of mats (0.56%), cutting plus hot water over the remaining pseudostem (0.56%), and salty water on the roots (0.56%). Only 4.84% of the farmers had not developed any endogenous practices. These endogenous practices identified in the study area therefore follow a well-defined logic to control the disease. Aside these strategies, the BBTD project sensitized banana growers to use the roguing as eradication method to control BBTD and restore devastated plantations in Bénin, through capacity building and a learning alliance (mobilization and training) (Sodédji et al., 2016: 314). The application of the method requires: i) Identifying the infected plant; ii) digging and clear the plant and all its secondary roots; iii) Spraying systemic insecticides to kill aphids and cut the infected plants, corms, rhizomes into small pieces; and iv) burying left-overs off the farm to avoid the risk of re-sprouting of the infected plant (Sodédji et al., 2016: 316). Of the farmers sampled, 49.46% are aware of the existence of the roguing/mechanical method. Farmer-to-farmer is the main means (45.65%) of information dissemination, followed by extension agent-to-farmer (CARDER) (30.43%) and project-farmer (23.91%) (Table 2). A higher optimization of these means of dissemination, an increasing farmer’s awareness and knowledge on the technology.

Table 2. Knowledge of roguing and sources of information.

Knowledge of the method	BBTD project	CARDER	Farmers of the community	Farmers of another community	Total
Frequency	22	28	34	8	92
Proportion (%)	23.91	30.43	36.96	8.7	100

Perception on Roguing Compared to Endogeneous Practices

To assess the perception of roguing compared to endogenous practices, the overall perception index was estimated (Table 3). Based on the most important criteria in the choice of pest and disease control methods, it appears that criteria such as the efficiency and quality of the harvest are the first criteria on which farmers have judged the roguing to be more advantageous than endogenous strategies. For criteria such as simplicity of use, simplicity of learning, and access to technical support, roguing was also perceived to offer higher advantages than endogenous practices. As for time for implementation, labor mobilized, and access to planting material support, endogenous practices were perceived to be more advantageous than roguing. In fact, when there are a high number of infected plants to eradicate, the method is time and labor consuming. However, farmers recognized the effectiveness of the technology and the simplicity of its use, learning, and transfer. Blomme et al. (2014: 4) reported similar findings about a method of controlling Banana Xanthomonas Wilt (BXW), which according to farmers is time and labor consuming, and becomes very cumbersome when the number of infected plants is high.

Table 3. Perception of farmers on the technology compared to endogenous practices.

Criteria	Perception Index
	Mean	stdev	Rank
Efficiency	0.93	0.13	1
Quality of the harvest	0.88	0.15	2
Total cost of carrying out the method	0.65	0.20	11
Time taken to carry out the method	0.54	0.22	15
Labor to be mobilized	0.58	0.20	14
Health expenses	0.86	0.13	3
Simplicity of use	0.81	0.18	5
Simplicity of learning	0.80	0.19	6
Simplicity of transfer	0.79	0.18	7
Barriers to use	0.85	0.12	4
Key people influence	0.74	0.15	8
Level of knowledge required	0.70	0.19	9
Influence of the community	0.62	0.18	12
Access to technical support (trainings/demonstration)	0.70	0.23	10
Access to healthy planting materials	0.59	0.23	13

The aggregate selection criteria of control methods against pests and diseases in relation to their respective attributes (Table 4) revealed that for all attributes (performance, ease of use, social pressure) except external support, farmers perceive mechanical eradication to be more advantageous than endogenous practices. In addition, the overall perception index for the sample indicates that the technology is generally perceived as more advantageous than endogenous practices.

Table 4. Overall perception of farmers on technology compared to endogenous practices.

Overall Perception	Mean	stdev
Performance	0.74	0.17
Ease of use	0.80	0.10
Social pressure	0.73	0.11
External support	0.65	0.21
Overall Perception Index	0.74	0.08

Adoption of Roguing of BBTD

The impact of the exposure of the roguing method on its adoption is estimated (Table 5). Only 49% of the farmers are aware of the existence of the method. This low number of knowledge rate limits the actual adoption rate to 36%. The potential adoption rate (ATE), which provides information on the unmet demand for the technology, is estimated at 56% and is significant at 1% threshold. This shows that if all farmers were exposed to the technology, the adoption rate would have been higher. This is consistent with the findings of Mwangi et al. (2014: 233) which show that if farmers were fully exposed, Push-Pull Technology of Striga eradication would have been expectedly more adopted implying additional efforts to close the adoption gap. The results also showed that the adoption rate within the sub-sample of farmers who were exposed to the technology was 73%. The adoption rate within the sub-sample of non-exposed is estimated at 40% and was significant at 1%. It is worth noting that the ATE1 and ATE 0 results cannot be used as true estimates of adoption rates due to “the selection bias”. This results from the fact that progressive farmers (innovators) are particularly targeted by promoters and technology developers, or farmers self-engage in exposure. As a result, the rate of adoption within the exposed sub-sample leads to an overestimation of the actual rate (Diagne and Demont, 2007: 202). Hence, ATE was then used to estimate the potential adoption rate. We accounted for the expected selection bias (PSB) by subtracting the potential adoption rate (ATE) from the adoption rate within the sub-sample of those exposed (ATE 1). The estimated PSB was very high (17%) and significant at 5%. This means that if ATE 1 had been used as a potential adoption rate, without correcting for this bias (PSB), the results would have been biased upward by 17%.

Table 5. Adoption rates of roguing.

Parameters	Values	Robust standard error
Adoption rate in the full population (ATE)	0.56***	0.08
Adoption rate within the subpopulation exposed (ATE1)	0.73***	0.04
Adoption rate within the subpopulation not exposed (ATE0)	0.40**	0.13
Joint Exposure and Adoption rate (JEA)	0.36***	0.02
Estimated population adoption gap (GAP = ATE-JEA)	−0.20**	0.07
Expected population selection bias (PBS = ATE1-ATE)	0.17**	0.06
Knowledge rate (Ne/N)	0.49***	0.04
Total of observations (N)	186
Number of farmers exposed (Ne)	92
Number of adopters	65

*p < .05; **p < .01; ***p < .001.

Furthermore, we estimated the Population Adoption Gap (GAP) by the difference between the observed joint exposure and adoption rate and the population adoption rate estimated. This “gap” exists solely because of the incomplete diffusion of the technology in the population (the unrealized potential) (Diagne and Demont, 2007: 209). In other words, the difference gives the adoption gap for the technology or the unmet demand if the population from which the sub-samples were drawn, were to be exposed (Mwangi et al., 2014: 234). This was estimated at 20% in the current study. These values may explain that if the whole population was exposed to the roguing method, the adoption rate could have been 56% instead of the 36% observed during the study period. Continuous exposure to the technology is therefore likely to enhance adoption rate. Mwangi et al. (2014: 234) in their study on adoption of Striga control technologies found similar results, respectively, for the actual adoption rate (36%) and the adoption rate within the sub-sample of those exposed (72.3%). The actual adoption rate (23%) and the adoption rate within the exposed sub-sample (59%) of improved rice varieties were relatively lower, unlike the potential adoption rate (58%) and the adoption rate within the non-exposed subsample (57%) that were higher than those found in this study (Diagne et al., 2007: 402). Improved yam varieties, meanwhile, have relatively low adoption rates, whether it is the adoption rate in the sub-sample of exposed (62%) or the potential adoption rate (37%) (Adégbola and Adékambi, 2008: 23). It is clear from these results that more effort needs to be put in, in terms of information, to enhance the adoption rate of the technology. Farmers are conservative and need time and information to be persuaded to adopt a technology (Shinohara 2000: 32). Adoption is a process that follows certain steps and there are four categories of adopters (Rogers, 1983: 36). Based on the categorization of adopters according to Rogers (1983: 22), the proportion of farmers who have adopted roguing is likely to be composed of innovators, early adopters, and some open-minded farmers.

Several reasons may explain the adoption rate obtained in the present study. One of them is the use of certified planting materials or planting materials from safe sources (BBTV-free planting materials) to restore affected plantations to prevent the spread of the disease. There is evidence that lack of healthy planting materials is one of the factors limiting banana production (Adigbonon and Allagba, 2015: 41; Sodédji et al., 2016: 320). However, these planting materials were not available in the study area before the arrival of the BBTD project, which trained farmers on banana production schedules and macropropagation techniques, and established pilot sites in the communes of Adjarra and Akpro-Missérété. These actions aim to make available and accessible healthy planting materials for farmers in those communes (Sodédji et al., 2016: 317) and has somehow impacted the adoption rate of the technology. Given the lengthy process and the aspect of risk minimization, some farmers required project support for provision of healthy planting materials before eradicating the diseased plants from their field. Another reason is the fear of losing some local varieties such as “Glinsi” (dessert banana) that is very appreciated but scarce. This situation leads farmers to use “simple cutting” as a control method, to avoid the total destruction of the plant hoping for a re-sprouting of a healthy plant (not affected by BBTD); however, this is never the case because BBTD is a systemic disease. Moreover, some farmers prefer to leave the infected plants hoping that it will produce fruits while marketing the leaves during the production cycle. Other factors of non-adoption of the roguing method as perceived by farmers include tediousness of the method and high time, labor, and resources costs when many infected plants need to be eradicated. Also, the need to eradicate this important number of diseased plants may also coincide with the peak period of labor requirement for other crops or activities in the seasonal calendar affecting the time and the labor available for roguing (Allen, 1978: 543). These reasons are consistent with those reported by households affected by Banana Xanthomonas Wilt (BXW), using a similar method of control. In fact, according to the study, when a large number of infected rejects have to be removed, this strategy is often highly demanding for resource-poor households. In addition, even after the destruction of infected plants, recurrent field monitoring is required to ensure complete destruction of the mats to prevent re-sprouting (Blomme et al., 2014: 7).

For more effective control of BBTD, roguing should be accompanied by good practices such as monitoring of plantations and use of healthy planting materials. The study showed that 52.15% and 39.25% of farmers, respectively, are aware of the importance of monitoring and the use of healthy planting materials from safe sources in the control of BBTD. A total of 46.24% of the farmers monitor the presence of symptoms on the plant and 10.39% of them established new fields using planting materials from safe or certified sources. The main sources of planting materials for the other farmers are own-field (46.38%), neighboring fields (34.78%), community farms (15.94%), and farmers in another community (2.90%). The availability of clean planting materials and the upturn of production practices increases the likelihood of BBTV containment success (Pemsl and Staver, 2014: 16). Regular monitoring (inspection) of the symptoms of the disease in the field is deemed important in the prevention and containment of BBTD in case of detection. In fact, as soon as the mother plant is infected, the virus quickly colonizes the host plant and is distributed in all the suckers (Iskra-Caruana, 2003: 15). Therefore, rapid eradication responses can improve efforts to mitigate the spread of the disease (Kumar et al., 2011: 181). Besides, early studies revealed that the effectiveness of roguing is function of the incubation period of BBTV, the detection efficiency, the relative infection rate, the eradication efficiency and the inspection interval (Allen, 1978: 542). According to the same author, the determination of the relationship between the infection rate and the monitoring interval is crucial in predicting any trend in future infection and roguing models to be developed. It is worth noting that the exclusive use of planting materials from a safe or certified source is very important in the control of this disease. The use of healthy planting materials combined with good management practices has made BBTV infection less likely in some countries with commercial production (Cameroon, Ghana, Mozambique, Philippines) (Pemsl and Staver, 2014: 16).

Factors Influencing Access to Information and Adoption of Roguing/mechanical Method of Eradication

The adoption model of roguing/mechanical method of eradication is estimated (Table 6). It appears that the model is globally significant at 1% threshold. This result shows the overall validity of the model. Six factors influencing the adoption of mechanical eradication were identified. These are: location, participation in training on banana production, access to extension services, banana farm size, annual banana income, and general perception of the technology.

Table 6. Determinants of access to information and adoption of the method.

		Access to information		Adoption
Variables	Description	Coefficients	Error-types	Coefficients	Error-types
zone	Location	1.358***	0.237	1.102***	0.311
form	Participation to at least one training on banana production	0.985**	0.427	1.083***	0.400
age	Age	−0.011	0.008	0.007	0.010
sexe	Gender	0.092	0.264	0.089	0.297
instru	Level of education	−0.028	0.137	0.124	0.189
tailm	Size of the household	0.040	0.036	0.019	0.038
vulgar	Contact with extension services	1.059***	0.291	0.683**	0.296
pergbon	Overall perception on the technology compared to endogenous practices			2.268***	0.401
rapaff	Annual banana income			−4.35E^−06*	2.35E^−06
supban	Banana farm size			6.52E^−04*	3.95E^−04
_cons		−0.922	0.583	−2.588***	0.818
Log pseudo likelihood		−79.168344		−48.411103
Wald chi2(9)		99.49***		75.98***
Pseudo R^2		0.3859		0.5934

*p < 0.10; **p < 0.05; ***p < 0.01.

The location positively influences access to information and the adoption of roguing. This implies that farmers who live in the pilot zones of the project are more likely to be acquainted with the method and adopt it than those in the control areas. In addition, the coefficient of this variable is significant at 1% threshold, which indicates that this variable is highly determinant in the knowledge and adoption of the technology. The pilot zones consist of villages in which the method has been disseminated. In addition, demonstration fields have been established within these areas, which also benefit from the action of public extension services. These are very useful because farmers often need practical information about the technology to decide whether to adopt it or not. These demonstration sites can provide this type of information to guide farmers in adoption while allowing them to minimize the risks of implementing the technology (Shinohara 2000: 32). Then, they have access to accurate information about the benefits of the technology, which allows them to evaluate it and make a positive decision about its adoption.

Participation in training on banana production also positively influences access to information and adoption of the method. In fact, farmers who have received training on banana production are more likely to be aware of the method and can easily adopt it. Training increases the ability of farmers to seek out and appreciate innovation by improving their access to knowledge and information (Shinohara 2000: 43). It strengthens and enhances the knowledge and skills of individuals and leads them to adopt a positive attitude toward innovations (Adékambi et al., 2010: 15). Several studies have emphasized the crucial role of training in the knowledge and adoption of innovation (Diagne et al., 2007: 401; Adékambi et al., 2010: 15; Adétonah et al., 2011: 8; Meijer et al., 2015: 44). Diagne et al. (2007: 401) showed in an adoption study of improved rice varieties that participation in training is the most important factor contributing to the knowledge and adoption of a technology.

Access to extension services has a positive influence on the access to information and probability of adoption of roguing. Being an indicator of exposure, access to extension services allows farmers to obtain information which positively influences adoption and has been demonstrated in several studies (Adégbola and Adékambi, 2008: 10; Adékambi et al., 2010: 9). Indeed, it is through this contact that they access information and knowledge about innovation including benefits of the technology (Shinohara 2000: 44, Rogers, 1983: 168). This contact, through visits and discussions to sensitize farmers on new technology, helps to promote and increase the adoption of innovations (Ochienno, 2014: 35). The paucity of information makes much longer the period within a given adoption intensity is expected to be reached (Pemsl and Staver, 2014: 17).

The larger the banana farm size, the greater the probability of adoption of the method. This shows the positive influence of farm size on the probability of adoption. Farmers with a relatively large farm size have made a significant investment in production. An uncontrolled BBTV infestation exposes them to financial risks, so, they take all appropriate measures to control the disease. According to Daku (2002: 170), the farm size can have a positive, negative, or neutral effect on the adoption of the innovation. In this context, Gabre-Madhin and Haggblade (2001: 20) in their study on the adoption of high yielding maize varieties showed the positive effect of farm size on adoption. According to this study, large-scale farmers have a greater tendency to quickly adopt a technology than small-scale farmers. Adétonah et al. (2011: 8), in their study on adoption of botanical extracts, reported contrary results. On the other hand, Ntsama Etoundi and Kamgnia Dia (2007: 18) concluded from the study on the adoption of improved varieties that it is not the fact of having a large farm size that leads individuals to adopt an innovation. This implies a neutral effect of farm size on adoption of innovation.

The annual income from banana production negatively influences the probability of adoption of the method. Farmers with a high income are more likely to reject the technology. This is not what is expected because this income represents the gross margin after BBTD infestation. When income is high, the farmer does not see the need to adopt the roguing/mechanical method of eradication, since it does not affect overall income. In fact, for a technology to be adopted, it must be seen as advantageous (Rogers, 1983: 218).

The overall perception of roguing compared to endogenous strategies positively influences its probability of adoption. This implies that a better perception of mechanical eradication compared to endogenous strategies leads to a high probability of adoption. This is not atypical, because according to the paradigm of adopters’ perception, the attributes of the technology as perceived by farmers, condition their adoption behavior (Ochienno, 2014: 19). In this regard, Adégbola and Adékambi (2008: 10) highlighted that the perception of the attributes of technology undoubtedly influences the probability of adoption of the technology.

Conclusions

This study assesses the degree of adoption of the roguing/mechanical method. The method was perceived by farmers as more advantageous than the endogenous strategies commonly used. Nevertheless, it is time consuming, labor demanding, and consumes financial resources when there is high number of infected plants to be eradicated. In addition, according to some farmers, there is a lack of healthy planting materials to support the adoption of the technology. Thirty-six percent (36%) of farmers actually adopted the technology. However, this rate would have been 56% if all the farmers had been sensitized since the majority (73%) of those who were sensitized adopted. The results also showed that the adoption of the method is determined by location, participation in training on banana production, banana farm size, access to extension services, annual income of banana, and the overall perception of the method versus endogenous strategies. The containment of BBTD and the restoration of infected fields is a major issue for food security given the status of this crop in agricultural households in the production areas. It is thus important to multiply sensitization programs to improve the level of adoption of this method, which is one of the options to contain this disease, improve production, and consequently increase the income of these farmers. Audio and video support, mobile phones, and media can be used as extension and dissemination tools. The development and extension of other methods, taking into account the shortcomings identified by farmers, the easy access to healthy planting materials and the use of good production practices would help to effectively contain BBTD outbreak.

Acknowledgments

Authors gratefully acknowledge the CGIAR Research Program on Roots, Tubers and Bananas (CRP-RTB) program for financial support for this research work.

Additional information

Funding

This work was supported by the CGIAR Research Program on Roots, Tubers and Bananas (CRP-RTB) [Task order No. 04-12-RTB-TO].