The study was conducted in the Sanmatenga province, in the Centre-North region of Burkina Faso. The Centre-North region is located between the parallels 12˚40'1; 14˚ North (N) and the meridians 0˚15; 25˚ West longitude (W). The sample covers ten villages from three municipalities, including six villages from the Boussouma municipality (Damiougou, Foutirgui, Goaragui, Guilla, Tanhoko), three villages from the Korsimoro municipality (Boalin, Sabouri-Tansobdogo, Wara), and two villages from the Ziga municipality (Niongtenga, Pissiga) (Figure 1).

Figure 1.Map of study area.

This study was conducted among recipients of food assistance for the creation of productive assets. It includes host communities and internally displaced persons. The equipment used for conducting the study mainly consists of a smartphone equipped with the KoboCollect application, which is used for administering the questionnaire.

2.3.1. Sampling

For the determination of the factors influencing the efficiency of the beneficiaries of the FFA program, four criteria guided the selection of villages to be surveyed: security accessibility, the presence of IDPs benefiting from the FFA program, the duration of WFP intervention, and the creation and use of productive assets. Indeed, regarding the accessibility criterion, the villages chosen were those that had not experienced terrorist attacks and were accessible from the town of Kaya. Additionally, the selected villages hosted IDPs who had benefited from the program for at least two years. With respect to the duration of intervention, the villages considered were those that had benefited from the FFA program interventions since 2022 or earlier. As for the creation of productive assets, it mainly concerns the distributed inputs and the work of recovering or developing land intended for agricultural production by host communities and internally displaced persons. The recovered or developed land was utilized during the 2022-2023 agricultural season.

In the identified villages, the surveyed individuals were chosen from among the beneficiaries of the FFA program. These beneficiaries include host communities and IDPs who participate in FFA activities for resilience. The selection of respondents was done randomly, taking into account their status as food assistance recipients in terms of livelihoods, as well as their involvement in carrying out agricultural production activities during the 2022-2023 wet agricultural season on land recovered or developed under the FFA program. Given that there are often multiple beneficiaries within the same household, only one beneficiary was chosen. Thus, emphasis was placed on household diversity to account for the variability of different characteristics of the respondents.

2.3.2. Data Collection

The collection of primary data focused on both quantitative and qualitative data. It was conducted through surveys carried out with the beneficiaries of the FFA program. In addition to the surveys, semi-structured interviews were conducted with the organizations responsible for the implementation and monitoring of the FFA program activities, namely ten representatives from the Village Development Committee (CVD) offices of the identified villages, as well as WFP partner organizations, particularly the Regional Directorate of Agriculture, Animal Resources, and Fisheries of the Central North (DRARAH-CN), including the WFP Focal Point and seven officers from the Technical Agricultural Support Units (UAT) in the study area, and the AVAD Association through the Resilience Project Officer. These interviews allowed for the preliminary collection of qualitative data on the management of the FFA program.

The primary data collected from the Kobocollect application were transferred to the KoboToolbox platform. These data were then retrieved in the Microsoft Excel 2016 spreadsheet for the creation of graphs and the conduction of various statistical analyses. Three types of statistical analyses were performed on the collected data, namely descriptive statistics, efficiency score measurement, and econometric modeling. Descriptive statistics were used to summarize the sociodemographic, economic, and institutional characteristics of the beneficiaries, the characteristics of the agricultural extension services (FFA), as well as the contribution of the FFA to the empowerment of beneficiaries in the Sanmatenga province. IBM SPSS software was used to determine the values of statistical parameters, including means, standard deviations, percentages, etc.

In this study, the DEA method in variable returns to scale mode was used. Thus, each beneficiary was considered a decision-making unit (DMU) that transforms inputs into outputs. This method provides a composite assessment of household efficiency by simultaneously synthesizing several partial efficiency measures. It allows identifying which households have the best practices among the studied sample based on each household’s distance from the efficiency frontier [9]. The distance separating inefficient households from the efficiency frontier (where the best practices are located) is measured using the efficiency score.

The calculation of efficiency scores in this study is based on the input-oriented model. In an input orientation, the DEA model minimizes inputs for a given level of outputs; in other words, it indicates by how much an organization can reduce its inputs while producing the same level of outputs. The rationale for choosing this method is that it aligns with farmers having control over inputs rather than outputs, which are mostly related to climatic hazards. In an output orientation, the DEA model maximizes outputs for a given level of inputs. In other words, it indicates by how much an organization can increase its outputs with the same level of inputs [10]. According to [11] (or the Charnes, Cooper, and Rhodes (CCR) method), the efficiency of a decision-making unit “k” is the solution to the following problem:

Efficiency scores were determined using the Win4DEAP2 software. The model variables are divided into inputs and outputs. The inputs selected are those mainly used for production on land reclaimed or developed under the FFA program. These include cultivated areas, seeds used, fertilizers (both organic and mineral), and labor in terms of agricultural workers. As for the outputs, these are the products and crop residues obtained to meet the needs of the beneficiaries.

In this part of the study, the aim is to establish the relationship between the level of efficiency (efficiency score) in the use of FFA and the variables related to the sociodemographic, economic, and institutional characteristics of the FFA program beneficiaries. The predominant method in the literature for finding the determinants of efficiency gaps among farms is Tobit regression analysis [12][13], because efficiency scores are censored at the maximum value of efficiency scores, which range between 0 and 1. The Tobit regression model takes the form of the equation below:

with:

TE isthe efficiency score or the dependent variable;

α is a constant that represents the value of the y-intercept;

β is the vector of coefficients affecting the explanatory variables;

X i refers to the set of explanatory variables;

ε i constitutes the model error term, which differs from one observation to another.

The Stata software was used for modeling. The variables selected to explain the efficiency of FFA use are described in Table 1.

Table 1. Variables used in the Tobit regression model.

The characteristics of the FFA beneficiaries are summarized in Table 2. The average age of the farmers was 47.67 years, with an average household size of 13 people. The membership rate in a Farmers’ Organization (OPA) was 33.15%. All farmers (100%) had access to land, with an average farm size of 2.76 hectares. Access to agricultural credit concerned 7.90% of farmers, while 79.02% had access to training. Access to agricultural extension services reached 94.27%, and 63.76% of farmers had access to agricultural information.

Table 2.Sociodemographic and economic characteristics of FFA beneficiaries.

The cultivated area is on average 0.04 ha. The average yield in onion production is estimated at 14,323 kg/ha for farmers in the market gardening perimeter (Table 3).

Table 3.Descriptive statistics of onion production inputs and outputs.

The average efficiency score of the operators within the Foutirgui vegetable perimeter is estimated at 60.83% with a median score of 56.70%. Efficiency scores range from 28 to 100%. The majority of producers (83.33%) have an efficiency score between 0.25 and 0.75 (Figure 2).

Figure2. Distribution of market garden operators within the perimeter according to efficiency score.

In the lowland rice-growing area of Goaragui, each producer has one to three plots, each measuring 625 m². For an average farm size of 0.09 ha, producers use 70.44 kg/ha ± 33.33 kg/ha of rice seeds for the nursery, 7.56 cartloads/ha ± 12 cartloads/ha of organic fertilizers, and 454.89 kg/ha ± 305.78 kg/ha of mineral fertilizers. The fertilizers used include manure and compost as organic fertilizers, and NPK and urea as mineral fertilizers. The labor used averages about 5 workers per farm (Table 4).

Table 4. Descriptive statistics of inputs and outputs in the lowland rice field.

The efficiency score of lowland rice farmers in Goaragui averages 77.06% with a median score of 78.50%. Efficiency scores range from 27.70% to 100%. More than half (58.06%) of the lowland rice farmers have an efficiency level with a score between 75% and 100% (Figure 3).

Figure 3. Distribution of lowland rice farmers according to the efficiency score.

The results of the Tobit censored regression model estimation applied to the vegetable perimeter operators are presented in Table 5. These results show that the model is overall significant at the 1% level (Prob > chi² = 0.0000 < 0.01). Thus, the model is able to explain the efficiency of use through the various explanatory variables considered.

Table 5.Tobit regression model on the efficiency of market garden perimeter use.

Note: **p < 0.05; ***p < 0.01.

The equation of the functional form of the Tobit regression model on the use of the vegetable garden is written as follows:

The results of the censored Tobit regression model estimation applied to lowland rice farmers are presented in Table 6. These results show that the model is overall significant at the 1% level (Prob > chi² = 0.0010 < 0.01). Thus, the model is suitable for explaining the efficiency of lowland rice use in Goaragui.

Table 6. Tobit regression model on the efficiency of lowland rice utilization.

Note: *p < 0.1; **p < 0.05; ***p < 0.01.

The equation of the functional form of the regression model on the efficiency of lowland rice utilization is written as follows:

The average efficiency is estimated at 77.06% among lowland rice farmers and 60.83% at the market gardening perimeter for onion production. It is generally well appreciated but reflects non-optimal use of production factors, with an average overall level of waste among rice farmers and market garden producers estimated at 22.94% and 39.17%, respectively. In other words, the overall waste level indicates the possible reduction in inputs if the explanatory factors of the regression model are controlled. In this regard, producers can increase their production without raising the level of inputs. Similarly high efficiency scores were obtained by [14] in Morocco (67%), [15] in Burkina Faso on millet farms in the Sahel region (71.23%), and [16] on the rice plains of Bagré (80%). On the other hand, it was relatively low (44%) on cereal farms in Burkina Faso according to the work of [17].

The negative coefficient of the age of farmers (Age) implies a decrease in the efficiency level of older producers in the Foutirgui market gardening area. This can be explained by the fact that older individuals are attached to traditional production techniques and remain reluctant to adopt technological innovations. In other words, a farmer who ages by one year will see their efficiency level decrease by 0.2% in onion production. Meanwhile, young people are generally more able to collaborate with extension services and seek out information. Similar results have been obtained in Senegal and Burkina Faso [18][19]. The household size (Size) of the beneficiaries shows an inverse relationship with the efficiency score of lowland rice and market garden operators because the variable has a negative coefficient. There is a strong relationship between household size and the number of agricultural workers. The correlation coefficient is 0.87 for onion producers and 0.91 for rice producers. This translates into an intensification of agricultural labor. Thus, the larger the household size, the more agricultural workers there will be on the same cultivated areas. Studies on farm efficiency in Morocco have reached similar results [14].

Experience (Exp) is positively correlated with the efficiency level of onion producers and is highly significant (1%). This indicates that farmers with at least 5 years of experience in market gardening are technically more efficient than those with fewer years of experience. The producer corrects past mistakes and thus adheres to the principle of learning by doing or learning through practice. These results are consistent with those of [20] on rice farmers in Mali. Access to training (Forma) positively influences the efficiency of onion producers in the Foutirgui market gardening area. Mastery of cultivation techniques is an essential factor not only for optimizing the use of inputs but also for ensuring maximum onion production while avoiding damage during the production process. Having benefited from training on nursery management and the production of biofertilizers under the FFA program, the farmers were able to optimize the quantities of seeds used and combat crop pests. These results are consistent with those of [12] on French agricultural farms.

The most educated producers (Educ) were likely to be more efficient because they would be able to make better technical decisions. However, only literate producers showed a significant level of efficiency at the 5% threshold in rice production with a positive coefficient. This could be explained by the attachment of individuals who had attended at least primary school to non-agricultural activities in order to better utilize their school-acquired knowledge. This result is similar to that of [13] on the technical efficiency of family farms in Mauritius. On the other hand, other studies have shown that educated individuals tend to be more efficient in agricultural production because they are open to innovations and have the capacity to manage resources rationally [15][19]. The positive coefficient of the variable (Orga) reflects the interest of peasant organizations in agricultural producers in relation to technical efficiency. Indeed, rice farmers who are members of a peasant organization are more efficient compared to those who have not joined. Membership in a peasant organization constitutes a significant leverage for improving efficiency. It is an indicator that captures the producer’s openness to benefiting from the experience of others and from innovation, which helps minimize the use of inputs in farming operations. These results support the findings of [14][20] and [21] in analyzing the determinants of technical efficiency of family farms in Mali, Cameroon, and Morocco, respectively.

Access to agricultural information (Info) positively influences the technical efficiency of rice farmers. Through information channels, mainly interpersonal exchanges and radio, producers listen to programs that facilitate learning good production practices and making better decisions in the use of inputs. This result supports the work of [22] on the effects of social services on the technical efficiency of agricultural holdings in Burkina Faso. These results show that owning a radio and being close to a rural dirt road can increase the technical efficiency of small farms.

However, the empirical analysis did not allow for the identification of the factors affecting the efficiency level of farms that have benefited from the construction of CES/DRS structures, which is due to the specificity of CES/DRS structures that need several years to have an impact on agricultural productivity. These structures are particularly used for food crop production (sorghum, millet, and cowpea). At this level, producers note that several pedoclimatic factors affect the efficient use of inputs. Pedoclimatic conditions encompass the combined influence of soil and climate. These conditions are critical in determining soil processes, properties, and SOC storage potential, impacting carbon sequestration. Pedoclimatic factors also significantly affect plant traits, particularly in organic farming. Favorable pedoclimatic conditions are essential for agriculture, attracting investments, and supporting agricultural activities in specific regions. Indeed, the planned quantities of seeds are multiplied due to repeated sowing following dry spells. Additionally, some soils are not suitable for the practice of half-moons and zaï pits because the low water permeability of these soils leads to crop flooding, especially for sorghum. The early end of the rainy season, specifically the 2023-2024 agricultural campaign, also constitutes a major constraint as it significantly affects yields despite farmers’ control over other production factors. The work of [23] confirms the effects of drought and flooding on the inefficiency of agricultural production in Burkina Faso.