Is contract arrangement source of income gain among parboiled rice stakeholders in Benin? A doubly robust analysis

Abstract

Rice supply remains insufficient for the world's consumption despite the agro-ecological potential, including the sub-Saharan African countries. The organization of the rice sector in Benin aims to a better profitability and above all in the increase in rice income of producers and processors. With this in mind, the objective of this study is to assess the impact of the contractual system on the earnings of parboiled rice stakeholders in the hill departments of Benin. For this reason, a random sample of 650 rice farmers spread over 400 producers and 250 processors made it possible to estimate of the Endogenous Switching Regression (ESR). The results of the estimates revealed that adherence to the agricultural contract as a function of socio-demographic factors such as human capital; gender; membership of an agricultural cooperative; have access to agricultural extension innovations and economic-institutional factors such as free entry into the market; access to quality agricultural products and access to credit. These results also confirm the positive effects of the acceptance of contracts on the parboiled rice income of the two actors. Adherence to agricultural contracts remains an effective agricultural policy likely to increase rice income in developing countries and in particular in the department of Collines which has a very high potential for arable land. The effective exploitation of these lowlands is a real source of increasing the rice supply in the perspective of insurance and guaranteeing better levels of food and nutritional security in Benin.

1. Introduction

Benin's agricultural sector has made progress in terms of production. 2020 posted low rainfall of 980 mm in 53 days compared to 1225 mm in 78 days in 2019. This negatively impacted the efforts made by all actors in the agricultural sector, causing variable production levels depending on the crops compared to 2019. Thus, cereal production increased from 2,177,787 tons in 2019 to 2,203,105 tons in 2020. As for industrial crops, cotton production reached a new African record, increasing from 714,714 tons in 2019 to 731,073 tons in 2020. Rice production is constantly increasing; it increased from 32 tons in 2016, to 201 tons in 2017 and 303 tons in 2018 and should reach 800 tons in 2019. Rice cultivation only really started in Benin after independence. From 1961 to 1978, production experienced rapid growth with the development of irrigated areas by national companies.

Benin has significant potential for the promotion of rice cultivation in terms of irrigable land, ground and surface water resources and proven technologies developed through research. Today, Benin is at more than 400,000 tons of rice production and aims to reach 1 million tons of rice in 2022 [1]. Rice production is constantly increasing, it rose from 32 tons in 2016, to 201 tons in 2017 and 303 tons in 2018 and should reach 800 tons in 2019. Despite strategic, political and technical support, yields for both Nérica rice and lowland rice remain below potential yields (2400 kg/ha for Nérica and 3500 kg/ha for lowland rice [2] against 5000 kg/ha and 7000 kg/ha respectively for Nérica and the lowland variety IR 841 [3] In that context, the adherence to contract might be of great importance for boosting more the rice production in Benin.

Adherence to contracts is primarily based on an agreement between producers and processors. The two actors agree beforehand on the terms and conditions of the production and processing of the agricultural product. These conditions usually specify the price to be paid to the operator, the quantity and quality of the product required by the processor and the date of delivery to the processor. The contract may also include more detailed information on how production will be conducted or, if applicable, whether inputs such as seeds, fertilizers and technical advice will be provided by the processor. “Contract farming” has given rise to many writings and debates over the past decades, with varying positions depending on the actors: institutional authors, from public or private research, or from civil society.

In this perspective, several empirical studies have evaluated the effects of the adhesion to agricultural contracts by producers and processors. Among these works [4], explain in their study that contract arrangement agreements are associated with welfare gains associated. Likewise, the work of [5] reports that contract arrangement is widely seen as a means of improving well-being in developing countries. But this improvement remains mixed due to the self-selection of smallholders in contract farming, it is not clear if contract arrangement really improves the welfare of farmers. They use data across multiple regions, businesses and crops in Madagascar to show that a one percent increase of contract arrangement adoption is leads to a 0.5% increase in household income. The work of [6] et al. (1990) focuses more on the politico-institutional aspects of agricultural contracts. Indeed, their work emphasizes that the participation of small producers can be linked to the national incentive policies of the 1990s, to situations where entry/exit from contracts is frequent, or to the values and image that the company wishes to have. They insist on the inclusion of small producers who can be changed over time. They raise the unsustainability of some small producers in agricultural contracts due to quality requirements and sustainability among retreating observers.

[7] underlined that non-adherence to agricultural contracts results from the risk of indebtedness of small producers, loss of autonomy, but also the poor conditions of agricultural workers within companies or on farms, and the risk of increasing discrimination against women, who are very rarely signatories to contracts despite their role on farms. Despite its advantages, contract arrangement is still not very widespread in Africa, except for traditionally exported products (cotton, coffee, cocoa, etc.). It is struggling to penetrate the domestic market (barely 10% of rice production in the Senegal River valley) and raises several questions. On the one hand, even if the economic literature credits it, on average, with a significant increase in agricultural income, this effect is not always observed. The advantages of contracting vary greatly depending on the production and depending on the context. Contract farming carries risks. Smaller producers may be excluded, because of the additional transaction costs they generate for businesses. They also have stronger market power and are often able to impose unfavorable purchasing conditions on farmers. Finally, contracting is likely to favor unsustainable production methods, if buyers define specifications that are potentially harmful to soil fertility, human health and the environment, or if they do not control the appropriate use.

Contract production can lead farmers to improve their practices, if companies promote them to consumers sensitive to these issues or if stricter standards oblige them to do so. Thus [8], have evidenced the positive effects of adherence to agricultural contracts on well-being, profit and income producers, then facilitates the reduction of transaction costs and the uncertainty that surrounds prices and marketing options, thereby facilitating planning and investments and increasing productivity by improving farmers' access to extension, services financial and agricultural inputs. These authors confirm that contract farming aims to improve coordination between actors in the sectors, offering enormous opportunities. It can strengthen the efficiency of value chains by reducing transaction costs, ensuring a better supply-demand balance of agricultural markets (in quantity, quality and regularity of flows), reducing post-harvest losses and improving food safety. In West Africa, rice occupies an increasingly determining place in the agricultural and food economy of the region. Rice has in fact gained shares of the regional cereal supply, due to the rate of production growth estimated at more than 5% per year.

In sub–Saharan African countries, adherence to agricultural contracts reduces uncertainties that emerge when two individuals want to cooperate. Agricultural contracts improve the quantitative performance of production by encouraging farmers to make better use of factors and by remunerating them according to a performance index. However, not all contracts, through the establishment of their own compensation scheme, achieve the same degree of efficiency. Several barriers to adhering to agricultural contracts weaken the marketing and increase in agricultural incomes for stakeholders with a view to improving their well-being. Agricultural contracts are part of the institutional mechanisms that help resolve situations of high transaction costs and agricultural losses.

The main research question of this study is as follows: Is the contractual arrangement a source of gain among parboiled rice stakeholders in the hill departments of Benin? This article therefore sets itself the objective of evaluating the contractual arrangement on the gain of parboiled rice stakeholders in the departments of collines in Benin. It is a specific first step to analyze the technical and economic determinants of the adoption of contractualization of parboiled rice in the collines departments of Benin. In a second step, it will be a question of measuring the gain in productivity resulting from the contract of farmers and supervisors in the departments of the collines in Benin.

By doing this, the study contributes to the relevant literature on contract farming and income gain as follows. First, to our knowledge, this is the first study in Benin that focuses on both producers and processors by investigating jointly the driving factors of participation in contract and their effects on their income. Most of the previous studies focused separately on either the producers or the processors and mainly investigated the determinants of the participation in contract farming [9,10,11,12]. Second, to our knowledge very few studies have assessed the income effect of participation in contract farming using rigorous econometric methods. In this study, we aim to fill these gaps in the literature by taking advantage of a rich dataset containing information on producers and processors to investigate jointly the gender-based driving factors of farmers participation in contract and its effect on their income. We specifically employ the endogenous Switching Regression (ESR) model. These methods address the issues of endogeneity and selection bias emanating from observed and unobserved heterogeneity. We check the robustness of the switching regression results using Propensity Score Matching (PSM). Last but not the least, this article therefore contributes enormously to further promote the production of parboiled rice in increasing agricultural income in an area with high agro-ecological potential. The choice of parboiled rice for agricultural contracting results mainly from its nutritional values better than those of white rice, is an interesting alternative to curb Benin's dependence on imported white rice. Likewise, several authors demonstrate that Beninese's consumers have a preference for imported white rice, and explain that local white rice cannot compete with imported white rice because of its quality considered “inferior” by urban consumers [13,14].

The rest of the article is structured as follows. Section 2 provides the related literature. Section 3 presents the methodology as well as the data used in the study. Section 4 analyzes the estimations results followed by the conclusion.

2. Related literature

[15,16]stated the theory of agency and related models [17]. present a review of the work relating to the economics of transaction costs. The difference between transaction cost theory and agency theory lies in the motivation of the contracting parties. The main reasons for using contracts, according to agency theory, are risk transfer (insurance) and alignment of incentives. On the other hand, transaction cost theory considers contracts as efficiency-enhancing availabilities that allow ex-post adjustments to be structured and discourage rent-reducing efforts that seek to influence the distribution of gains, including ex post negotiations. Post, hold-up issues and the costs of ex ante research and evaluation. In this context, according to the theory of transaction costs, contracts should be determined by: (i) the need for specific investments that create interdependencies so that partners seek protective devices (which should determine the type of contracts and their clauses; (ii) the need to improve the efficiency of the supply chain by reducing costs; (iii) the need to establish close coordination in a context where quality, variety and safety of products are essential aspects. Efficiency is at the heart of the arguments of the theory of transaction costs and is one of the main reasons for contracting, given the productivity gains that the improvement of material skills promotes improved management skills, technology transfer and coordination [18]. argue that while Cash incentives effectively encourage cost reduction, they do not allow effective control to reduce costs, they do not effectively control opportunistic behavior of farmers, processors or distributors.

Contract farming has given rise to many writings and debates over the past decades, with varying positions depending on the actors: institutional authors, from public or private research, or from civil society. Three major themes in particular are recurrent in recent work: first, the work of [19,20] note the positive effects of contract farming on household income, food security, market integration and technical change. Secondly, the work of [8] notified the conditions for the execution of contracts and the organizational effectiveness of contract farming. The results of these authors are less convergent, between the reduction of transaction costs and the risks of opportunistic behavior of companies or farmers in certain situations [21]. Finally, the work of [22,23] highlighted the inclusion of small producers in contract farming. Here too the findings are contradictory, between the effective inclusion of small producers. Critics of transaction cost economics argue that market power is perhaps the fundamental motivation for the development of vertical integration and contractual arrangements. The central argument is that dominant firms would use contracts to extend or exercise their market power.

[24] highlighted the role of agricultural contacts in improving rice value chains They model the participation and intensity of contract farming in four different ways along the vertical coordination continuum and highlight the role that both policies can play in fostering the inclusion of contract farming in the modernization of the rice value chain in Vietnam. The work of [25] has highlighted the major role of rice control in Togo. They pointed out that consumption has been increasing rapidly due to urbanization. Using cross-sectional data based on the multistage sampling technique, they showed how to value local rice at the expense of imported rice. The results of the endogenous switching regression model estimation revealed that participation in contract farming improved the quality of paddy rice. They reported that certain factors such as: the number of extensions visits to the farmer, the threshing method used by the farmer and the agro-ecological zone of the farmer were decisive for the improvement of paddy rice in the Togo.

In other spatial settings, for example in Kenya, the work of [26] has also shown that contract farming is considered as a tool for creating new market opportunities. They used data that was collected from 100 small avocado growers. With an instrumental variable (Probit-2SLS) model to control for the endogeneity of contract participation and examine the effect of contract farming on household, farm and farm income. The results revealed that this agriculture has favoured the increase in the income of smallholder farmers. However, this contract farming has been dangerous for small farmers to the detriment of large farms. Similarly, their results indicated that participation in contract farming was not sufficient to improve household well-being, farm and avocado income.

Recent work by Ref. [11] has opted for improving access to resources for smallholder farmers in Ghana. They have made efforts to promote contract farming in Ghana. This strategy, which was proposed with the aim of increasing the agricultural productivity of farmers, giving better access to the market and guaranteeing an adequate supply of raw materials to agro-industries. In their objective, it was a question of knowing how this form of agriculture led to an improvement in the food security of farmers. Their article therefore sought to explore the extent to which farmers' food security status was influenced by their participation in contract farming activities. They used a Cragg double-barrier model to analyze contract farming participation. Their results showed that the paddy yield and the farmer's wealth are the main factors that influenced the amount of paddy rice to be contracted in the contract farming agreements. This article also revealed that participation in contract farming increased food security by 109%.

The work of [27] on the inclusion of smallholders in contract farming relates to support for the integration of the industrial milk and tomato sectors which are subject to premiums. In this contract farming, the premiums played a decisive role in increasing the price to producers and motivating them to deliver their production to the company, thus securing supplies. The strong participation in the marketing contract prompted the largest tomato cannery in Algeria (CAB) to offer farmers production contracts. Small producers, although in the majority, participate very little, for reasons highlighted in the article [28]. report that contract farming facilitates the participation of smallholders in the market. This agriculture improves household well-being and promotes rural development. The results show that contract farmers achieve higher incomes than their counterparts without a contract than in some countries. Contract farmers in most countries have an increased demand for hired labor, suggesting that contract farming is boosting employment, but we find no evidence of spillover effects at community level. They concluded that contract farming unambiguously improves welfare. The implications for policy and research are relevant beyond contract farming [9]. report that contract farming has become a popular mechanism to encourage vertical coordination in agriculture in developing countries. This type of agriculture is limited by its ability to stimulate structural transformation in rural economies. They show that all contracts have positive effects on measures of well-being and productivity. This agricultural contract regulates agricultural risk thanks to a fixed-price offer and induces the behavior of farmers allowing them to face other constraints themselves.

Similarly, the work of [29] enumerates an alternative typology of contract farming agreements (CFA) based on the theory of transaction costs. Their results based on a two-by-two matrix of contracts based on the interaction of transaction attributes show that four types of contracts can be distinguished: total, group, light and market contracts. Additionally, CFAs that do not match the transaction attributes have side-selling and inefficiency issues. They concluded that this new method of evaluating agricultural contracts is a tool to help managers and policy makers design CFAs that match the underlying transaction attributes, thereby improving the stability and efficiency of CFAs. The work of [12] explains that contract farming is a practice aimed at reducing the risk and uncertainty commonly done on agriculture. They notify that this form of agriculture is divided into two types which are contract production and market contract. Contract farming is generally practiced strategically and at high risk. One of the contractual agricultures practiced on potatoes is a contract. Interlaced contract farming was the type of contract farming market. The comparison of the potato; the incomes of farmers following contract farming were lower than those of independent potato farmers.

Moreover [30], obtain an essential result of contract theory from their work. They show that contracts can help transfer risk from one party to another, with the latter assuring first. Their work reveals that participation in contract farming is associated with a 0.20 standard deviation decrease in income variability. They rely on mediation analysis to examine the mechanism behind this conclusion, they find support for the hypothesis that fixed price contracts; that transfer the entire price risk from the producer to the processor; explain the reduction in income variability associated with contract farming. They suggest that contract farming would likely be more beneficial for households that do not participate than for those that do. They therefore support the idea that, in a context where formal insurance markets fail, contracts can serve as partial insurance mechanisms. However [24], empirically analyzes the influence of contract farming on income and agriculture. He identifies the mechanisms of contract farming on income, sustainability and well-being using the qualitative method. These results show that contract farming has a negligible impact on farm income while it can facilitate agricultural activities and reduce difficulties. Factors such as the education of the head of the household, the gender of the head, type of crop, and technology all contribute to increased farm income.

In general, contract farming can have positive impacts on income, sustainability and well-being in the medium and long term. In the short term, the result is not significant due to similar or lower price compared to the spot market price, increasing production cost, decreasing productivity and poor contract execution.

3. Methodology

3.1. Estimation strategy

Participation in a contract arrangement can be analyzed in the context of utility maximization theory. Let $U_{i1}$ be the utility derived from the participation in contract arrangement and $U_{i0}$ be the utility derived from not participating. Let's denote $U_i\text{,}$ the difference in utility between participation and non-participation. The parboiled rice farmer i decided to participate in contract farming when it gave him more utility than in the case of non-participation. Mathematically, we will have equation (1).

$$U_i=U_{i1}-U_{i0}>0\text{.}$$ (1)

Since $U_i$ is not observable, the preference of choice of the farmer can be represented by the latent variable ${Contract}_i$* for the participation in contract farming:

$${Contract}_i{}^{*}=\partial Z_i+{\mu }_i$$ (2)

$${Contract}_i=\{\begin{array}{c}1if{Contract}_i^{*}>0\\ 0if{Contract}_i^{*}<0\end{array}$$

where ${Contract}_i$ denotes participation in contract arrangement which takes the value 1 if the producer or the processor participated in the contract scheme and 0 otherwise, $Z_i$ is a vector of characteristics of the producers and the processors that are supposed to influence the decision to participate in contract farming and ${\mu }_i$ the error term.

The impact model of the participation in contract arrangement on parboiled rice farmer's income is presented as follows equation (3):

$$Y_i={\alpha }_i{X}_i+{{\beta }_i{Contract}_i+v}_i$$ (3)

with $Y_i$ denoting the income of producers or processors, α and β the parameters to be estimated and $v_i$ the error term.

Several techniques are used in adoption or participation regression. These include the Heckman selection model, the Propensity Score Matching model (PSM), the Instrumental Variable models (IV) and the Endogenous Switching Regression (ESR). These models are based on strong hypotheses and address the selection bias issues. Indeed, the decision to adopt contract farming or participate in contractualization in Benin is not random. Each producer or transformer decides for himself. In this case, there may be observable or unobservable factors that can influence the adoption decision, which creates selection biases. When the biases are due to unobservable factors (motivation, skills, etc.), the PSM and Heckman models cannot address the endogeneity bias. In this case, instrumental variables (IV) model and endogenous Switching Regression (ESR) are commonly used. The use of instrumental variables model is based on the quality of the instrument used, which must be strongly correlated with the adoption variable and not correlated with the impact (outcome) variable.

In fact, the advantage of the ESR model is that it simultaneously estimates the adoption or participation decision and the impact model. In addition, this model can be used to do counter factual analyzes, that is, to determine what would have been the performance of the producer or transformer who participated in contract farming if she had not participated in. Conversely, what would have been the income of non-adopters or non-participants if they participate in contract farming. ESR model addresses selection biases due to observable and unobservable factors [31,32].

The ESR model involves separate estimates for the two groups of both producers and processors. Consequently, the adoption of contract farming becomes the selection criterion indicating the regime (adoption or non-adoption) to which the producers or the processors belong. A producer or a transformer is considered to adopt contract farming if she participated in the contract farming exclusively to produce or transform parboiled rice.

Following equation (2), incomes are observed for the two producers’ groups depending on production or transformation [31,33].

$$\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{m}\mathrm{e}1:Y_{1i}={\alpha }_1{X}_{1i}+v_{1i}\mathrm{P}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{p}\mathrm{a}\mathrm{n}$$ (4)

$$\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{m}\mathrm{e}2:Y_{2i}={\alpha }_2{X}_{2i}+v_{2i}<\mathrm{s}\mathrm{p}\mathrm{a}\mathrm{n}\mathrm{i}\mathrm{d}{=}^{″}\mathrm{z}\mathrm{o}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{o}-\mathrm{d}\mathrm{r}\mathrm{a}\mathrm{g}″\backslash />\mathrm{N}\mathrm{o}\mathrm{n}-\mathrm{P}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{p}\mathrm{a}\mathrm{n}\mathrm{t}$$ (5)

where $Y_{1i}$ is the income from either production or transformation of parboiled rice, $X_i$ a vector of exogenous variables affecting the income of the producer or the processor, and $v_i$ the error term.

The error term in the participation equation (2) and the error terms in equations (4), (5) may be correlated. To address this problem, Equations (2), (4) and (5) will be simultaneously estimated using the Full information maximum likelihood method which appear as the most effective approach used with the “movestay” command under Stata [32]. However, to have a more robust identification, we included at least one exclusion restriction. Hence, following the literature, we use a set of binary variables that captures technical assistance and a member of a cooperative as instruments in the contract farming model that are excluded from X in the outcome model. The rationale behind using those variables as instruments is that farmers that benefit from a technical assistance and are member of a cooperative are more likely to have access to contract farming than those who do not.

Following previous studies [1,26,34,35], the ESR model is used to compare the income of participants in contract farming to non-participants and to estimate the expected returns in the case of counterfactuals where the participants have not participated in contract farming, and where the non-participant have participated. The effect of participation in contract farming on participants is expressed by equation (6). It is the “treatment effect on treaties” (TT) which is the difference between cases (a) and (c) [31,36].

$$\mathrm{T}\mathrm{T}=E\left(y_{1i}\right|{Contract}_i=1)-E\left(y_{2i}\right|{Contract}_i=1)$$ (6)

Similarly, the effect of the treatment on non-treated (TU) equation (7):

$$\mathrm{T}\mathrm{U}=E\left(y_{1i}\right|{Contract}_i=0)-E\left(y_{2i}\right|{Contract}_i=0)$$ (7)

In the PSM method, a counterfactual is constructed so that reduce from the issue of selection biases. The main purpose of using this method is to find a group of non-treated farmers that is similar in all relevant observable characteristics to the treated. Using propensity scores, the average treatment effect for the treatment group (ATT) can be estimated [17]. Let $Y_{1i}$ and $Y_{0i}$ denote the potential income for parboiled rice actors i that has participated in contract arrangement and those who did not, respectively. In reality, $Y_{1i}$ or $Y_{0i}$ cannot be observed at the same time. Let Contract represent a binary treatment variable that equals one if the producer or the processor has participated in contract farming and zero otherwise, equation (8).

$$Y_{ATT}=E\left(\Delta |Contract=1\right)=E\left[(\tau \left(1\right)|Contract=1\right]-E\left[(\tau \left(0\right)|Contract=1\right]\text{,}$$ (8)

where ${\Delta }_{ATT}$ is the average treatment effect on the treated; $E\left[(\tau \left(1\right)|Contract=1\right]$ is the expected outcome variable of parboiled rice farmers participant in contract arrangement; and $E\left[(\tau \left(0\right)|Contract=1\right]$ is the outcome variable of non-participant in contract arrangement that would be expected if they had not adopted the variety. PSM requires imposing a conditional independence assumption and common support assumption for identification [36]. If these two assumptions are met, the PSM estimator for ${\Delta }_{ATT}$ is given as:

$$\Delta \frac{PSM}{ATT}=E_{p\left(X\right)|Contract=1}\left\{E\left[(\tau \left(1\right)|Contract=1,p\left(X\right)\right]-E\left[(\tau \left(0\right)|Contract=1,p\left(X\right)\right]\right\}\text{.}$$ (9)

To check whether the PSM results are sensitive to hidden bias due to unobserved factors, we apply the bounding approach proposed by Ref. [37], which determines how strongly an unobserved factor may influence the selection process in order to invalidate the results of PSM analysis [17].

3.2. Data and variable specifications

We used a survey data collected in the commune of Glazoué in the department of Collines in central Benin. At the commune level, 13 villages were selected for the survey by mutual agreement with the local authorities and the commune's agricultural development agents. The criteria for choosing the villages included the presence of lowlands, the presence of a significant number of rice producers, and the presence of forms of contract farming. At the village level, producers were selected at random from lists of producers provided by the farmers' organizations. We interviewed 50 producers per village for reasons of convenience, and a total of 650 farmers including 275 actors adopting contract farming and 375 who have not adopted contract farming. Among the adopters, 175 farmers are parboiled rice producers and 100 are processors. Among non-adopters, 225 are producers against 150 processors. The departments of Collines in Benin have many lowlands characterized by hydromorphic soils that have not yet been exploited ([4]. Similarly, the lowlands in the collines departments of Benin are recognized as a potential agro-pastoral, fertilizing zone that supports permanent and intensive crops ([38]. The departments of Collines in Benin include the communes of Savé, Ouéssé, Dassa, Glazoué, Savalou and Banté. This department covers an area of approximately 13,899 km², or a portion of 12.35% of the national territory. In terms of rice-growing potential, more than half of the lowlands of the country are in this department [13], and these very wide lowlands are used by the populations for agricultural purposes, especially rice growing. The colliness department is the area of Benin where lowland development activities have taken on greater importance in recent years and [39,34]. The lowlands are used in the colliness department for two main purposes: self-consumption and marketing.

4. Results and discussion

4.1. Descriptive results

Table 1 presents the characteristics contract farming adopter and non-adopter. Significant differences were observed between adopter and non-adopter for both producer and transformer groups for the following variables: income, age, education, gender, access to quality product, supervising contract, technical assistance and member of a cooperative.

Table 1.

Mean difference for outcome and socio-economic characteristic variables.

Variables	Measurement	Contract for producer		t-test/chi-2	Contract for processors		t-test/chi-2
		Yes	No		Yes	No
Income	FCFA1	788391.77 (267655.45)	624500.38 (345672.34)	5.67***	888291.77 (367655.45)	724510.28 (376672.34)	3.67***
Age	year	45.93	45.63	−1.30*	42.83	47.62	−2.30**
Credit	FCFA	325025 (163887)	311458 (226578)	0.56	425035 (143587)	411358 (216478)	0.56
Type of contract
Tenant	1 = Yes 0 = No	35.90	33.50	0.22	45.90	49.50	0.82
Sharecropping	1 = Yes 0 = No	22.90	23.50	0.72	35.90	59.50	1.90*
Wage labor	1 = Yes 0 = No	29.90	37.50	0.92	15.90	19.50	0.62
Education (1 = Yes; 0 = No)	1 = Yes 0 = No	25.17	22.88	3.67***	27.56	25.84	4.57***
Gender	1 = Male; 0 = Female	83.75	81.25	1.41*	87.15	73.74	2.41**
Ln household Size	Number of people	5.63	5.42	−0.21	4.67	5.67	−0.21
Experience	Years	33.55	23.78	0.56	32.56	30.18	0.108
Access to quality product	1 = Yes 0 = No	56.54	49.31	−1.23*	59.78	55.98	−1.43*
Price level of parboiled rice	1 = Yes 0 = No	50.52	57.44	−0.34	52.62	53.24	−0.54
support to contracted farmers	1 = Yes 0 = No	38.90	36.50	0.12	48.90	46.50	0.72
Supervisor contract	1 = Yes 0 = No	61.10	48.50	3.45***	71.10	38.50	2.45**
Technical assistance	1 = Yes 0 = No	59.27	59.73	3.71***	53.25	50.43	3.71***
Member of a cooperative	1 = Yes 0 = No	60.18	59.50	4.01***	70.18	69.80	5.01***
Distance to market	Km	5.58 (2.03)	6.56 (1.89)	0.72	6.88 (2.03)	6.96 (1.89)	0.32

Note ***p < 0.01; **p < 0.05 et * p < 0.1. Standard error within parenthesis.

Among producers, those who participate in a contract farming are more productive in terms of revenue than the non-participants. Participating in a contract farming yields 788391.77 FCFA (1433 USD1) against 624500.38 FCFA (1136 USD) yearly. Similarly, the participants' processors get 888291.77 FCFA (1615 USD) against 724510.28FCFA (1318 USD) yearly for non-participants in a contract farming. Although the significant level is lower for producer than processors, the age of the farmers is key in respect to the participation in a contract farming. The participant producers are a little bit older than non-participants. Unlikely to producers, participants’ processors are younger than non-participants. In regard to education, there is a significant difference between participants and non-participants among the groups of producers and processors. Coming to gender, men are more involved in contract farming than women among producers and processors. In addition, having access to quality product, being supervising, benefiting of technical assistance as well as being a member of a cooperative significantly differ participants from non-participants in the two groups.

4.2. Empirical results

4.2.1. ESR estimates of contracts farming in parboiled rice farming

The full information maximum likelihood estimates of the driving factors of contract farming participation (selection equations) and impacts of the contract farming on the parboiled rice income (outcome equations) in ESR for producers and processors models are presented in Table 2. Column (1) and column (4) in Table 2, present estimates of the selection equations for producers and processors, respectively. Most of the variables in the model have hypothesized signs. The model has a good fit with explanatory variables as shown by the significance of the likelihood ratio test (p-value <0.01) of independence between the equations. The estimated coefficient of correlation between participation in contract farming and parboiled rice producers and processors ‘income is positive and significantly different from zero suggesting that self-selection occurred in the adoption of contract farming. The difference in the coefficient estimates between producers and processors’ participation and non-participation in contract farming indicates the superiority of the switching regression to a simple treatment effect model.

Table 2.

Determinants of participation in contract farming and parboiled rice income.

Dependent variable: Income	Endogenous Switching Regression
	Producers			Processors
	Selection	Parboiled rice income		Selection	Parboiled rice income
		(1)	(2)		(3)	(4)
		Contract	No contract		Contract	No contract
Age (year)	0.004 (0.030)	0.008 (0.006)	0.146*** (0.045)	0.021* (0.012)	0.060** (0.035)	0.216*** (0.025)
Education (1 = Yes; 0 = No)	0.396** (0.326)	0.267** (0.126)	0.207 (11.29)	0.196** (0.101)	0.280 (0.610)	−0.407 (11.08)
Gender (1 = Male; 0 = Female)	0.029** (0.005)	0.050*** (0.015)	−0.054 (2.264)	0.024* (0.015)	0.028*** (0.008)	−0.054 (1.964)
Ln household Size	0.034** (0.017)	−0.319** (0.150)	−1.435 (10.29)	0.021** (0.014)	−0.319** (0.15)	−1.435 (5.990)
Experience	0.052* (0.025)	0.002* (0.008)	0.043 (6.082)	0.062* (0.015)	−0.0118** (0.014)	0.023 (3.062)
Access to quality product	0.043 (6.012)	−0.041 (1.637)	−1.238 (−0.700)	0.013 (3.082)	−0.0413 (1.637)	−1.231 (−0.510)
Level of parboiled rice	−0.328 (0.700)	−0.055 (0.122)	−0.035 (0.252)	−1.238 (−0.700)	0.040*** (0.075)	−0.034 (3.964)
Government support to contracted farmers	0.074** (0.035)	0.043 (0.133)	0.122 (0.113)	−0.024 (0.352)	−0.119 (0.15)	−1.064 (7.79)
Supervisor contract	0.111 (0.103)	−0.194 (0.133)	−0.194 (0.400)	0.172 (0.183)	0.017* (0.01)	0.121 (0.042)
Distance to market (km)	0.139 (0.241)	0.206*** (0.055)	1.578*** (0.014)	0.286 (0.176)	0.036 (0.055)	0.178*** (0.022)
Agricultural credit (1 = Yes 0 = No)	0.130*** (0.023)	0.064*** (0.021)	0.500*** (0.072)	−0.280 (0.733)	0.052*** (0.010)	0.700*** (0.035)
Technical assistance	0.075*** (0.017)			0.055* (0.032)
Member of a cooperative	0.291* (0.160)			0.090*** (0.012)
Constante	−0.1012 (1.092)	−0.031 (5.512)	0.048 (16.99)	−0.121 (1.092)	−0.021 (4.412)	0.038 (13.92)
Rho_0		7.71*** (0.210)			4.71*** (0.160)
Rho_1		14.21*** (0.172)			11.11*** (0.162)
		Loglikelihood = −504.35			Log likelihood = −4187.082
		Wald chi2 (15) = 40.41			Wald chi2 (15) = 633.49
		Prob > chi2 = 0.0000			Prob > chi2 = 0.0000
		LR test = 6.63			LR test = 5.33
		Prob > chi2 = 0.04			Prob > chi2 = 0.04
		Obs: 400			Obs: 250

Note ***p < 0.01; **p < 0.05 et * p < 0.1. Standard error within parenthesis.

The main drivers of parboiled rice producers' decision to participate in contract farming were education, gender, experience, household size, and member of a cooperative, Government support to contracted farmers, technical assistance and agricultural credit for parboiled rice producers (Columns 1). Similarly, the driving factors of parboiled rice processors’ decision to participate in contract farming were age, education, gender, household size, experience, technical assistance and member of a cooperative (Column 4). Education plays a significant role in participating in contract farming. Producers that are educated are more likely to participate in contract than the non-educated. Indeed, the educated producers of parboiled rice are more likely to participate in contract farming than the non-educated producer. In addition, men are more likely to participate than women in contract farming. In the same vein, the size of the household, being a member of a cooperative as well as getting an agricultural credit and having technical assistance increase the probability of participating in the contract farming respectively.

Likely to parboiled rice producers, there are several factors that drive the decision of parboiled rice processors to participate in a contract farming in central Benin. From column 4 of Tables 2 and it raises that the older processors are more likely to participate in contract farming. Similarly, education has a significant effect on processors’ decision to participate in a contract farming. Furthermore, men processor are more likely to participate in a contract farming than women. In addition, a higher household size augments the probability of participating in contract farming among parboiled rice processors. Similarly, benefiting a technical assistance as well as being a member of a cooperative increase the probability for a parboiled rice transformer to participate in contract farming.

All these results are in line with empirical literature that posits that socioeconomic characteristics including age, education, gender, household size, experience are the main drivers of participation decision in contract farming [40,35,29]. In addition, a number of studies find that variables relate to institutional and informational contexts including technical assistance and member of a cooperative are key in participating in a contract farming [4]. Indeed, a producer or a transformer with institutional support would be more favorable to change decision toward participation in contract farming than the others [22].

Columns 2, 3, 5 & 6 of Table 2 presents the driving factors of parboiled rice producers and processors’ income for participants and non-participants in contract farming. The results indicate that education, gender, household size, experience, member of a cooperative, distance, agricultural credit were variables that significantly influence income of parboiled rice producers that participated to contract farming in central Benin while only age, distance and agricultural credit affect the income of producers that did not participate to contract farming. Concerning the processors, ages, gender, household size, experience, member of a cooperative, distance, agricultural credit were variables that significantly influence income of parboiled rice processors that participated in contract farming while only age, distance and agricultural credit affect the income of processors that did not participate to contract farming.

The level of education was positively associated with the income of parboiled rice producers and processors in Benin. This implies that among the participant to the contract farming, being educated increases the income of parboiled rice producers and processors. Gender also has a positive effect on the income of parboiled rice producer, implying consequently that men are more productive than women in terms of revenue generated by the production and transformation of parboiled rice in central Benin. Furthermore, experience in rice producing and transforming was positively associated with parboiled rice producer's income. Indeed, an increase in rice farmers experience in rice production and transformation contributes to improve income among participant in contract farming in Benin. The distance to market as well as agricultural credit were all positively associated to parboiled rice producer's income. Another interesting result is the negative association between the household size and the income of the parboiled rice producer. Although striking, this finding implies that it is challenging for households with high size to invest in inputs to increase their parboiled production and therefore rip benefit from rice income. Besides, the level of parboiled rice price and the contact with supervisors have no effect on parboiled rice producers' income but positively affect the processors ‘income. This implies that knowing information about the price of parboiled rice as well as being in regular contact with supervisor are key to increase performance of parboiled rice processors in Benin. However, very striking is the effect of the government support to contracted parboiled rice farmers on processors’ income in Benin. This result may be explained by the fact that those supports are generally mismanaged even when it is a financial support.

The expected producer's income under actual and counterfactual conditions is reported in Table 3. Cells (a) and (b) represent the expected producer income observed in the sample. The expected income for parboiled rice producer that participated in contract farming is higher than the one of the groups of parboiled rice producer that did not adopt. Based on this simple comparison, it can be misleading to attribute the difference in parboiled rice income to the participation in contract farming. The treatment effect for contract farming participants is 0.015 (ATT = a – c). This is equivalent to 01.51% [exp (ATT)-1] difference in the average income. In other words, when parboiled rice producers did not participate in contract farming, their income would have been decreased by 01.51%. If non-participants participated in contract farming, their income would have been increased by 09.41%. However, the transitional heterogeneity effect is positive, implying that the effect of participation in contract farming on parboiled rice producer's income is significantly greater for rice producers who actually participated compared to those that did not participate.

Table 3.

Impact of contract arrangement on parboiled rice producers’ income.

Bub group	Contract status		Effects of contract on producers' income
	Yes (N = 175)	No (N = 225)
Participated Didn't participate	(a) 0.605 (d) 0.555 EH1 = 0.050***	(c) 0.590 (b) 0.465 EH0 = 0.125***	ATT = 0.015*** ATU = 0.090*** TH = 0.075***

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

Similarly, the expected processor's incomes under actual and counterfactual conditions are reported in Table 4. Cells (a) and (b) represent the expected transformer's income observed in the sample. The expected income for parboiled rice processors that participate in contract farming is higher than the group of parboiled rice processors that did not adopt. Based on this simple comparison, it can be misleading to attribute the difference in parboiled rice income to the participation in contract farming. The treatment effect for contract farming participants is 0.035 (ATT = a – c). This is equivalent to 03.56% difference in the average income. In other words, when parboiled rice processors did not participate to contract farming, their income would have been decreased by 03.56%. If non-participants participated in contract farming; their income would have been increased by 05.12%. However, the transitional heterogeneity effect is positive, implying that the effect of participation in contract farming on parboiled rice producer's income is significantly greater for rice processors who actually participated compared to those that did not participate.

Table 4.

Impact of contract arrangement on parboiled rice processors’ income.

Sub group	Contract status		Effects of contract on processors income
	Yes (N = 100)	No (N = 150)
Participated Didn't participate	(a) 0.815 (d) 0.765 EH1 = 0.050***	(c) 0.780 (b) 0.715 EH0 = 0.065***	ATT = 0.035*** ATU = 0.050*** TH = 0.015***

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

The last rows of Table 3, Table 4 which account for potential heterogeneity effect in the sample, reveals that parboiled rice producers and processors who actually participated in contract farming would have higher income than parboiled rice farmers that did not participate in the counterfactual cases (c) and (d). This highlights that there are some important heterogeneity factors that makes the participation in contract farming better off than the non-participants.

4.2.2. Robustness check: propensity score model

The results from the ESR models above may be sensitive to the exclusion restriction assumption; hence we also used the PSM approach to check the robustness of the estimated effects. We compare our ESR results with results from standard propensity score matching (PSM) that are presented in Table A1 and Table A2 in annex. The same variables were used in the estimation of propensity scores. We followed the rule of [41,38], for quality implementation of propensity score estimation. Fig. 1 displays the common support region that show that for each treated parboiled rice stakeholder there is a matched untreated stakeholder. Table A1 provides the ATT estimates from the PSM approach. Similar to ESR results, the PSM estimates show that the participation in contract arrangement increases the income of producers and processors by 0.45 and 0.62 respectively. In addition, we tested hidden bias with the bounding approach proposed by Ref. [37] and found that the PSM estimates were robust to hidden unobserved characteristics (see Table A2). We therefore conclude that the results PSM are robust to unobserved characteristics and our estimates are doubly robust.

Fig. 1.

Common support.

5. Concluding remarks

In this article, we have investigated whether contract arrangement among parboiled rice stakeholders in central Benin can increase their income. Findings shown that contract arrangement affect positively the income of producers and processors in the parboiled rice farming in Benin. Besides, several factors drive the participation in contract farming including the socio-economic characteristics of the stakeholders that also influence their income. Being educated, member of a cooperative group is also driving the adoption of contract arrangement in Benin. These findings are in line with adoption theory that is very key in improving productivity in agricultural sector. This may be accompanied by adequate institutional support including access to credit and education facilities for rice farmers.

Author contribution statement

Mounirou: Initiated research project, participated in literature search, supervised data collection, participated in drafting article. Jérémie Yebou: Participated in literature search, participated in field data collection, analyzed and interpreted data, drafted the article and made all revisions.

Data availability statement

Data will be made available on request.

Declaration of competing interest

I am pleased to submit a final revised version of my research article entitled Is Contract arrangement Source of Income gain among Parboiled Rice stakeholders in Benin? A Doubly Robust Analysis, for publication in this journal.

I believe that this manuscript is appropriate for publication by your journal because it addresses issues related to agriculture and development, which are part of the journal's aims & scope.

This manuscript has not been published and is not under consideration for publication elsewhere. There is no conflicts of interest to disclose as well as no direct funding associated to this research.

Appendices.

Table A1.

Nearest neighbor estimates of ATT

Stakeholder	Algorith	Outcome	ATT	Number of treated	Number of control
Producer	NNM	Ln income	0.45*** (0.078)	175	225
	KBM		0.36*** (0.051)	175	225
Processor	NNM	Ln income	0.52*** (0.038)	100	150
	KBM		0.47*** (0.028)	100	150

Note ***p < 0.01; **p < 0.05 et * p < 0.1. Standard error within parenthesis.

Table A2.

Impact of contract arrangement on parboiled rice processors' income

Sub group	Contract status		Effects of contract on processors income
	Yes (N = 100)	No (N = 150)
Had Contracted Not had contracted	(a) 0.815 (d) 0.765 EH1 = 0.050***	(c) 0.780 (b) 0.715 EH0 = 0.065***	ATT = 0.035*** ATU = 0.050*** TH = 0.015***

Note ***p < 0.01; **p < 0.05 et * p < 0.1. Standard error within parenthesis.