Economic Performance of Soil and Water Conservation Practicesin Burkina Faso

The continuous degradation of agroecosystems is a major concern for Sub-Saharan African countries, particularly Burkina Faso. To fight against this problem, various research projects and programs have implemented Soil and Water Conservation practices (SWC) in Northern Burkina Faso. The objective of this study was to assess the economic performance of stone rows, grass strips, zaï, filtering dikes, half-moons and agroforestry on agricultural production in this part of Burkina Faso. Stochastic Frontier Analysis was used to estimate SWC’s technical efficiency. Results indicated that the cost for SWC construction did not influence white sorghum and pearl millet yield. However, an increase of 1% in the investment for SWC implementation results in a 0.42% increase in groundnut yield and 0.19% in cowpea yield. Although, the half-moon technique had a positive effect on the farmer’s technical efficiency, the effects of stone rows, filtering dikes, zaï and grass strips were not significant. Given the tremendous efforts that farmers develop to implement these anti-erosion practices, one recommendation is that policy makers strengthen the technical, financial and equipment supports to farmers for efficient implementation of SWC techniques to ensure sustainability of agricultural production systems in Northern Burkina Faso. www.scholink.org/ojs/index.php/ra Research in Agriculture Vol. 4, No. 1, 2019 39 Published by SCHOLINK INC.


Introduction
Burkina Faso, like most countries in Sub-Saharan Africa, is facing a continuous degradation of its ecosystem. In most situations, this degradation of ecosystems is cause stagnation or even a decrease in crop yields, biomass availability for livestock and the availability of ligneous and harvesting products (Palé et al., 2019a(Palé et al., , 2019b. In 2006, results from a study conducted by the Permanent Secretariat of the National Council for the Environment and Sustainable Development (SP/CONEDD) indicated that about 11% of the country's land was greatly degraded and 34% considered moderately degraded.
Factors affecting the land degradation are mainly the very strong human population grow and economic pressure, the use of unsustainable production practices such as no application of organic as well as mineral fertilizers, overexploitation of natural resources around cities and villages and low use of SWC practices. This leads to a shortening of fallow duration, a decrease in crop yields and a degradation of soil properties (Roose et al., 2017).
To fight against this degraded situation, SWC such as stone rows, grass strips, zaï, filtering dikes, half-moons and agroforestry ( Figure 1) had been introduced in the Yatenga Province several decades ago in Northern Burkina Faso. After decades of SWC practice, the impact of these different techniques in terms of the level of adoption by farmers and the effects on crop yields and livestock production merit study. Literature indicated that little work has been done to assess the economic performance of the SWC techniques in the country (Da, 2008). Therefore, a study on the influence of these anti-erosion techniques on crop yields in Northern Burkina Faso particularly in the Yatenga Province was needed.
Through an econometric approach based on the estimation of production frontiers and the analysis of technical efficiency, the present study attempts to respond to the question of how the use of SWC techniques could improve the agricultural performance of farmers' productions in the northern Region of Burkina Faso. More specifically, the study aims to assess the effects of SWC techniques on farmers' efficiency.

Study Area
The survey was conducted in four villages in Yatenga Province with the capital Ouahigouya

Data Collection
The data were collected through a questionnaire and from farmers in the four villages covered by the survey. The unit of observation for this was each household and the respondent was the farmer, responsible of all decisions to be taken in farm management.
Due to the absence of a population data basis giving the list of all farmers practicing SWC techniques in the 4 villages, the sample size was set at thirty (30)  farmers.

Data Analysis
Three categories of variables were selected for the model analyses: (1) output or production, (2) inputs, and (3) variables likely influencing farmer's efficiency. The inputs considered were the three basic factors of production: land, labor and capital (or cost of implementation). Fifteen (15) explanatory variables to technical efficiency were selected (Table 1). one's knowledge about the technology of the sector studied (case of the agricultural sector for example), the estimation of production frontiers is preferable (Bosman & Frecher, 1992). Thus, based on literature (Bosman & Frecher, 1992) and the random nature of agricultural sector, the SFA approach being more appropriate for estimating production frontiers was used. The estimation of the production function parameters was based on the maximum likelihood method. Once the coefficients of the production frontier are estimated, the variances of the errors σ2 and γ are calculated according to equation 1.
where γ, an important estimate in the analysis of technical efficiency, represents the part of the deviation between the observed production and the potential production which is explained by the inefficiency of the farmer. A value of γ equals zero means that the deviation from the production frontier is entirely due to random factors (not dependent on the farmer); a value of γ equals 1means that the deviation from the production frontier is entirely due to the efficacity of the farmer.
The technical efficiency index (Tei) used to dissociate the contribution due to technical inefficiency from the purely random contribution in the term i in the case of stochastic production frontiers (Jondow et al.,

1982) is obtained from Eq. 2.
Eq. 2: The estimates of ui and σ* vary with the distribution of the term inefficiency: If ui follows a semi-normal law then u * i If ui follows an exponential law then u * i εi σv 2 σu 2 et σ * σv

Choice of Production Technology
The literature distinguishes two production frontier specifications: Cobb Douglas type function and Translog type function. The first one only gives the main effect of each single input on the output, while the second gives both main effect of each input and interaction effect of combination of inputs on the output. That's why, in the case of Cobb Douglas function, coefficients estimated can be interpreted directly. But, in the case of Translog function, partial elasticities (eki) obtained by Eq. 3 give better interpretations.

Eq. 3:
Thus, interpreting the effect of the production factor xk on production y is done by interpreting the average effect of its partial elasticities obtained by equation 4 (Eq. 4).
Eq. 4: The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) were used to make a choice between Cobb Douglas and Translog production functions. The model using Cobb Douglas function was the one that minimized the AIC and BIC criteria for white sorghum and pearl millet. So, Cobb Douglass production function was used for the white sorghum and pearl millet productions and Translog function which minimized the selection criteria and being more appropriate for estimating the production function of groundnut and cowpea was used for these two crops (Table 2). Note. AIC: Akaike information criterion; BIC: Bayesian information criterion.
Source: Estimates from Authors.

Empirical Model Presentation and Estimation Procedure
Based on the Battesse and Coelli model (1995), the empirical model described by Eq. 5. Eq.5:

White Sorghum
For the white sorghum production, the value of γ estimated by the Cobb Douglas type function was 0.69 (Table 3) indicating a technical inefficiency for white sorghum farmer. This γ value also shows that the difference between the observed production and the potential production related to the frontier is explained by 69% of farmer's inefficiency. The fact that a white sorghum farmer does not produce maximum yield from the total amount of inputs used, is explained at 69% by his inefficiency and 31% by uncontrolled random effects. It is important to indicate that all the estimated coefficients of the production function are significant at the critical p-value of less of equal to 5%, except for the β3 coefficient associated with the total cost of SWC construction. In fact, 1% increase in field size leads to 0.61% increase in white sorghum production. Similarly, an increase of 1% in labor size led to an increase of 0.29% in white sorghum production. Thus, to increase their productions, white sorghum farmers have to increase their labor size and field size rather than increase the cost of SWC construction.

Pearl Millet
For pearl millet production, the value of γ estimated according to Cobb Douglas type function is 0.33 (Table 3). In other words, the fact that a white sorghum farmer is not located on the production frontier is explained at 33% by his inefficiency. As found in white sorghum production, the estimated coefficients for pearl millet are all significant at the critical p-value of less of equal to 5% with the except for the one associated with the cost of SWC construction. Thus, an increase of 1% in pearl millet field size results in 1.19% increase in pearl millet production; an increase of 1% in farm labor size leads to an increase of 0.59% in pearl millet production. Therefore, to increase their production, pearl millet farmers have to increase their field size or labor size.

Groundnut
For groundnut production, the value of γ estimated is 0.20 depending on the Translog function used for the estimation (Table 4). In other words, the fact that a groundnut farmer is not located on the production frontier is explained at 20% by his inefficiency and 80% by uncontrolled random effects.

Cowpea
For cowpea production, the value of γ estimated is 0.55 depending on the Translog function used for the estimation (Table 4). In other words, the fact that a cowpea farmer is not located on the production frontier is explained at 55% by his inefficiency and 45% by uncontrolled random effects. For better interpretations, in the case of Translog function, it's necessary to calculate the partial elasticities for each farmer. The average partial elasticities for groundnut are shown in Table 5. Table 5 indicates that an increase of 1% in the groundnut field size results in 5.46% increase in groundnut production; an increase of 1% in the farm labor size leads to 2.03% increase in groundnut production and that 0.42% increase in groundnut production occurs when the total cost of SWC construction was increased by 1%. The average partial elasticities for cowpea are shown in Table 5. Table 5 indicates that an increase of 1% in the cowpea field size results in 5.70% increase in cowpea production; an increase of 1% in the farm labor size leads to 0.98% increase in cowpea production and that 0.19% increase in cowpea production occurs when the total cost of SWC construction is increased by 1%.

Technical Efficiency Scores
All farmers therefore produce below the production frontier, which corroborates the existence of inefficiencies. The results from the calculation of technical efficiencies (Table 6) indicate that on average white sorghum farmers have a technical efficiency of 66%. In other words, these farmers produce on average 66% of the maximum possible production and lose 34% of their potential production due to their inefficiency and uncontrolled random effects. For pearl millet, results show an average technical efficiency scores of 61%. In other words, pearl millet farmers lose on average 39% of their potential production due to inefficiency. For groundnut and cowpea farmers, have an average technical efficiency of 74% and 49.5%, respectively. These average technical efficiencies indicate that these groundnut and cowpea farmers produce on average 74% and 50% of the maximum possible production, respectively. After completion of the efficiency score analysis, it is important that the factors affecting the different technical inefficiencies be analyzed.

Factors Affecting Technical Efficiencies
Results from the analysis of factors that explain the level of technical inefficiency of white sorghum farmers (Table 3) indicate that only the coefficients associated with the farmer's education level, the use of half-moon technique, the financial and technical support received, and the breeding of cattle are significant at the critical p-value of less of equal to 5%. In other words, only these variables significantly influence the levels of inefficiency. Indeed, the fact of having received equipment by technical or financial support in combination of livestock production in the farm have a positive effect at the critical p-value of less or equal to 5%. In contrast, the education level of farmers and the practice of half-moons have a negative effect on the efficiency level of white sorghum farmers. For pearl millet farmers, only the coefficient associated with farmer's age significantly influenced farmers' technical efficiencies (Table 3).
The positive sign of this coefficient indicates farmer's age has a negative effect on his technical efficiency. For groundnut and cowpea, only the coefficients associated with farmer's sex and age have a significant effect on technical efficiency (Table 4). Since the coefficients have the same sign, this indicates that being a man has a positive effect on the efficiency of groundnut and cowpea farmers.

Discussion
Results from the study show that no farmer is on the production frontier thus confirming the existence of  Vlaar and Wesselink (1990) who found that filtering dikes increased white sorghum yields between 0.5 and 1.5 t ha -1 , with a larger increase in relatively dry years. Zougmoré (2003) also found that in good rainy season, the use of stone rows does not significantly improve yields. For pearl millet, groundnut and cowpea farmers, it is clear that none of the SWC techniques had a significant effect on the farmer's technical efficiency. SWC techniques were therefore not efficient since they did not have a significant positive effect on the farmer's efficiency. This result seems to contradict farmers' perception. However, it should be noted that farmers find the techniques more efficient compared to the fields where no SWC technique is applied, and argue that crop productions are improved with application of those techniques (Coulibaly, 2018 millet, groundnut and cowpea farmers results indicate a negative effect of the farmer's age on his technical efficiency at the critical p-value of less of equal to 5%. One could think that older farmers who have more experience in farming be more efficient than youngers but unfortunately, they are not. Nevertheless, these results could be attributed to the continuous decrease in power for these old farmers whereas the construction of SWC techniques such as stone rows and particularly zaï, are very physical effort demanding. Therefore, in a situation of highly degraded soils, young farmers would be more efficient because they still have physical capacity to better implement SWC techniques. The level of education has a positive effect on the farmers' technical efficiency. However, for white sorghum farmers, the level of education affects negatively these farmers' efficiencies. White sorghum farmers with very low levels of education are more efficient than those with higher levels. In contrast, results indicate that pearl millet, groundnut and cowpea farmers' technical efficiencies are rather not influenced by the farmer's level of education. This result for white sorghum farmers seem a bit paradoxical and contrary to that of Coelli and Flemming (2004), as educated farmers are supposed to be those who have got high technical skills for using improved production techniques. However, Audibert et al. (1999) came to the same conclusion for farmers in Ivory Coast. According to these scientists, the higher educated farmers are more attracted by prestigious jobs and activities that provide rapid and high incomes than farming, thus they do not devote great time on agricultural activities (Nuama, 2010).

Conclusion and Involvement of Political Authorities
The objective of this research was to assess the economic performance of the SWC techniques in the Northern Region of Burkina Faso. To achieve this, the SFA approach was used to estimate production frontiers. Results show that no farmer is located on the production frontier. The analysis of production functions indicated that the cost of SWC implementation has a mixed effect on agricultural production. A 1% increase in the total cost of SWC construction leads to an increase of 0.42% in groundnut production and 0.19% in cowpea production. However, no significant effects are observed on white sorghum and pearl millet productions. The results also show that the combination of different SWC techniques by farmers does not have the same effects on the farmers' technical efficiency. Indeed, except for the half-moon practice that has a significant and positive effect on the technical efficiency of farmers, none of the other techniques including stone rows, filtering dikes, zaï and grass strips has influenced farmers' technical efficiencies. On the basis of these results, the authorities in charge of the making and implementation of agricultural policies should define appropriate training programs to enhance farmers' technical capacities, provide financial and equipment supports to farmers for an efficient use implementation of SWC techniques to ensure sustainability of their production systems.