Land Access Modes and Agricultural Productivity in Benin

Abstract

Improving productivity is an important channel for satisfying household food needs through food availability. Increasing the productivity of small-scale farmers is linked to a number of factors, including land access, labor, chemicals, fertilizers and so on. Most farmers resort to expanding their cultivated areas to increase production because of insufficient financial resources available for purchasing inputs during agricultural seasons. This situation, combined with increasing pressure on land, does not favor agricultural production and, by extension, food security. This study therefore assesses the impact of different modes of access to land on agricultural productivity. Regime-switching multinomial regression techniques were used to account for endogeneity bias due to observed and unobserved factors. The results of the study show that access through free loans, tenancy and sharecropping negatively affect agricultural yields. The counterfactual analysis reveals a positive gain estimated at 509.9 kg/ha from land access by landowners. If the lenders were landowners, their gain would be 396.6 kg/ha, whereas the farmers would gain 485.1 kg/ha if they were landowners, and similarly the sharecroppers would gain 389.8 kg/ha if they were landowners. It is clear from these results that improving agricultural yields depends on securing land and requires the establishment of an effective land ownership system. This research impacts land ownership policies, which need to be revised to address customary rights and reduce inequalities in access to secure land. It highlights the way land tenure security drives agricultural advancements and offers actionable recommendations for policy improvements on food security.

1. Introduction

Land is a determining factor in the economic development of the agricultural production sectors. This development depends essentially on improving productivity and is imperative for the growth and sustainability of agriculture in the future [1]. Land is also an important source of wealth accumulation and transmission from one generation to another, as well as a vehicle for well-living and investment [2,3]. Policymakers have regularly prioritized securing land ownership in order to guarantee and develop more productive agriculture [4]. This priority given to securing land ownership is part of the implementation of the Sustainable Development Goals (SDGs) through the reduction in poverty and hunger, responsible consumption and production, urban development, human settlements and sustainable infrastructure. Considering this, a land policy focused on agricultural areas serves as a guarantee against illegal claims to customary land rights [5].

Benin, like most African countries, is characterized by low incomes and low productivity [6]. For example, for maize, which is the main food crop in Benin, the yield was 1075 kg/ha in 2019, whereas the world yield is 4 Tns/ha [7]. Agricultural productivity is determined by factors like production, land, labor and agricultural inputs. Labor is not fully exploited and most farmers use few inputs because of the inadequate financial resources often available during the farming seasons [8]. In order to increase their productivity, farmers resort to extending the areas under cultivation, leading to forest degradation. Although access to and use of land resources have become important factors in identifying the causes of food production problems, they are also sources of inequality [9].

The farming practices in Benin are characterized by smallholdings, and, in 2018, the country had around 35.03% arable land [6]. Despite this, the majority of land-poor households use leases to acquire more land and reduce access inequality, although the size of the leased land is often small [10]. This pressure on the arable land, resulting from extensive farming using inefficient tools, increases the difficulties regarding access to land, particularly in the south of Benin (Atlantic, Mono/Couffo and Ouémé/Plateau), where the soil has already exceeded the threshold for good farming productivity without fertilizers [8]. According to Joel and Bergaly [11], the pressure on productive farming land is becoming a potential source of land ownership insecurity for households. This leads to land-use conflicts, low investment in land improvement and, lastly, low productivity. Insecurity and limited access to land, together with the adverse effects of climate change on land ownership, have contributed to low agricultural production [12,13].

The land tenure system in Benin is characterized by insecurity of tenure due to institutional weaknesses and poor coordination between the land tenure players, and an institutional structure that has been overtaken by the demands of the modern economy. Thus, the Beninese government has taken a number of measures to facilitate conditions of access to secure land tenure, with a view to creating a favorable framework for the mortgage credit necessary for economic growth. To this end, a number of pilot operations were set up under the law governing rural land in Benin, passed in 2007, adopted in 2013 and revised in 2017, followed by the creation of the Agence Nationale du Domaine et du Foncier (ANDF) in 2015. The aim of this new legislation was to unify and modernize land law and combat speculation. This new land code has several important features that constitute major innovations. For this reason, the major challenges to be met for good land regulation include ensuring equitable access to land, secure investment and effective management of land conflicts in order to contribute to poverty reduction, the consolidation of social peace and the achievement of integrated and sustainable development. The main modes of access to land in Benin are primitive or free occupation, community ownership, succession, donation, purchase, rental, pledge, loan and sharecropping. The farm management systems often encountered in the field involve two types of tenure: direct and indirect. In the case of direct tenant farming, the farmer himself uses the land he owns under a right of ownership. This type of farming is the responsibility of farmers who have inherited and/or purchased their land. Indirect tenancy covers all the land ceded under sharecropping or tenant farming arrangements, generally for a limited period. In this context, the purpose of this study is to analyze the effects of different modes of land access on agricultural productivity, considering the socio-economic and legislative context of Benin.

According to economic theories, property rights encourage long-term investment in land and the adoption of new technologies [14,15]. Hanstad has argued that secure individual land rights are essential elements of a productive system in the agricultural sector, vital for poverty reduction and economic growth [16,17]. Farmers’ ability to return their investments in soil productivity is less important when they collectively own the land than when they rent it [18]. Land registration increases the use of credit by providing a greater incentive to invest in agriculture and reducing the incidence of land disputes [19,20]. Improving farming productivity requires appropriate investment conditions and land ownership security [21]. The customary land ownership systems, under which farmers often do not hold ownership on the land they cultivate, do not provide farmers with adequate incentives to adopt new technologies that can improve production [12,22]. This emphasis on secure property rights is not empirically proven.

The research on the relationship between access to land and farming productivity has produced mixed results. The work of Chand and Yala, Newman et al., Linkow and Li and Zhang has shown that secure land ownership is associated with high productivity [23,24,25,26]. They all agree that secure land ownership improves agricultural productivity. But, Michler and Shively and Lawin and Tamini find no evidence between access to land and agricultural productivity [27,28]. They show that non-landowners have consistently higher levels of technical efficiency and productivity. The difference observed in the empirical results may be linked to the land policy and customary management practices in each local context. In addition, the majority of these studies were based on land titles as a measure of access to land. The observed conflicts in the results of the previous studies could be attributed to the land tenure systems and socio-economic contexts, as well as differences in the research methodologies. For example, the land tenure system in Benin is characterized by a mix of ownership, rental and sharecropping, thus affecting productivity differently compared to other contexts. Also, access to credit and modern agricultural inputs in Benin are limited, resulting in decreased productivity compared to other socio-economic frameworks where such resources are more widely available.

The research question is the following: what are the effects of different land access modes on farmers’ agricultural productivity? The aim of this article is to study the effect of different land access modes on the agricultural productivity in Benin. To this end, we formulated the following hypothesis: households with plots acquired through sharecropping or tenancy have a lower level of agricultural output. Throughout the literature, land rights are endogenous [14]. Regime-switching multinomial regression techniques were used to account for endogeneity biases due to observed and unobserved factors.

In our study, we adapt and build upon the existing conceptual models linking land ownership, agricultural productivity and land security. While previous models identified in the literature have explored the effects of land ownership security on agricultural production through investment as well as input use, the contribution of the current study lies in a novel perspective that refines the previous models with a specific focus on unaddressed key factors such as the interaction between secured land ownership and agricultural equipment investment; the use of variable inputs (fertilizers, seeds and pesticides); the connection between land market effect and technical efficiency and the indirect impact of land security through credit markets (in this respect, we argue that land security enables farmers to use their land as collateral, improving their access to credit and facilitating investments in both inputs and agricultural equipment). In addition, the novelty of our study is represented by the integration of the effects of various land access modes with detailed productivity measures, focusing mainly on agricultural yield, for a deeper understanding of how land tenure arrangements influence productivity and investment, particularly in the context of the intensive farming systems in Africa. Moreover, the use of the Multinomial Endogenous Switching Regression (MESR) method enabled us to obtain more accurate and detailed estimations of the effects of land access modes on agricultural productivity.

In various African areas, including Benin, one of the striking barriers to socio-economic development (food security being at the top) is represented by low agricultural productivity due to limited financial resources, inefficient farming practices, insecure land tenure systems and land access modes. In this context, our study was required to fill the literature gap and recommend possible solutions to the challenges related to agriculture, affecting a significant portion of the population. The policy level addressed included farmers (landowners, tenants, sharecroppers and borrowers), local and national decision makers as well as credit and funds providers.

To provide a clear understanding of how this research has been conducted, the structure of the article is organized into four distinct sections. The first presents the theoretical and empirical review, the second the research methodology, the third the results and discussion, including descriptive statistics, and, finally, the implications of the study for public policy, the economy and society.

2. Theoretical and Empirical Link between Land Access and Farming Productivity

Since the time of the physiocrats, farming land has been the foundation of agriculture, food and economic life despite the rise of soilless agriculture [29]. The Ricardian hypothesis of land rent theory states that agricultural production is subject to diminishing returns because cultivation requires land and the area of arable land is limited [30]. According to Ricardo, it is the cultivation of marginal, less fertile land in response to population growth that explains the creation of land rent. Land rent is the remuneration for the private appropriation of land [30]. This economic concept is related to the notions of scale economy and diseconomy and the law of diminishing returns, which plays a central role in the theory of rent developed by Ricardo [31]. The new concept of land renting takes into account three factors of production: labor, land and capital, whereas the neoclassical approach only takes into account the capital factor and labor [32]. Two fundamental assumptions for the modern concept of land renting have been formulated: the hypothesis of intrinsic land utilities, which takes on the appearance of a financial product, and the hypothesis of the informational efficiency of the agricultural land market, which makes it possible to value public services in market land prices [32]. This second hypothesis gave rise to the theory of contracts applied to agricultural land in order to understand an incentive framework for tenants, providing them with security and encouraging them to invest in the fertility of their land [32].

Contract theory is a relevant framework for understanding the nature of the trade-off for a landowner between a tenant farming contract (fixed payment) and a sharecropping contract inducing a payment proportional to the harvest [31]. According to the standard assumptions of neoclassical theory, the sharecropping contract is an institution that leads to the suboptimal use of inputs [31,33]. The reasoning stems from the fact that the quantity of labor used by the sharecropper is not that which equalizes the marginal product of the labor with the wage, which is nonetheless a first-order condition necessary for the maximization of profit from production [34,35]. All the other explanations for sharecropping can be grouped into three families: those involving the notion of risk, those stressing the problems of moral hazard and control (incentives) and those highlighting the incompleteness of the markets in developing countries [31].

2.1. Channels Linking Land Access and Agricultural Productivity

Feder identifies three important economic relationships concerning land rights and productivity [36]. First, title facilitates the use of land as collateral to obtain loans to finance agricultural investments [37,38]. Secondly, land titles could increase farmers’ security of ownership and strengthen their willingness to make medium- and long-term investments in their land. Finally, titling can stimulate land markets, facilitating the transfer of land resources to the most productive farmers. Moreover, guaranteed property rights also influence agricultural productivity because they encourage more efficient use of resources, known as the “factor intensity effect” [21].

The codification of private property rights should in theory increase investment and agricultural productivity and stimulate economic development [14,15,39]. In the absence of land titles, the threat of eviction on returns from investments in production implies that tenant farmers will make less effort in production [23]. This is a version of Marshallian inefficiency whereby an agent shows less than optimal effort in response to not having a full residual claim on the production [39]. Some authors question this idea about property rights, pointing out that poor credit and land markets weaken or even nullify certain hypothetical linkages, especially in Africa. It is widely accepted that many customary land ownership systems offer sufficient ownership security to stimulate investment and production growth [40,41,42]. Empirical work has focused on these theoretical debates, producing inconclusive results, especially in Africa.

2.2. Empirical Links between Land Access and Agricultural Productivity

The empirical evidence is contradictory throughout the literature, especially in Africa, where there is visible complexity regarding customary land relations and a limited capacity of central land administrations to issue land titles [43]. In Europe, the link between land ownership and agricultural productivity has focused much more on the macroeconomic level than the microeconomic level. In Europe, Gardi et al., showed that the impact of land seizure led to an equivalent loss of more than six million tons of wheat [44]. Subsequently, Popescu et al., showed that 15 countries have the highest economic efficiency in land use out of the 28 states of the European Union [45]. Only Italy, Spain, Greece, Cyprus, Croatia, Romania and Malta exceed the 40.95% average share of gross value added per ha of used land in the value of the EU-28 agricultural production. According to these authors, to make land use in European agriculture more efficient, farmers need to optimize the combination of production factors in order to increase their agricultural output. Also, Europe’s intense urbanization has a direct impact on its ability to produce food [45].

In Asia and Africa, we note the existence of a link between access to land and agricultural productivity by some work at the micro level, while others do not find such a link. Newman et al., in Vietnam, using a fixed-effects approach, showed that obtaining land titles is associated with higher yields, both for individual and joint titles [23]. Chand and Yala noted, in the context of Papua (New Guinea), that the higher productivity obtained is due to the benefits of economies of scale and the absence of the revenue sharing enjoyed by farms with more secure land tenure [24]. Joel and Bergaly and Rashid, using the conditional mixed process to capture the impact of land ownership security on agricultural productivity through access to credit, arrived at the same results in both Cameroon and Tanzania [11,46]. They showed that land ownership security improves agricultural productivity through access to credit. Chen et al., showed in Ethiopia that certification facilitates tenancy and improves agricultural productivity [47]. The effect of a counterfactual reallocation from no tenancy to efficient tenancy increases the agricultural productivity at the zonal level by 43% on average. The results of Keudem and Savadogo indicate that improved land ownership security leads to improved agricultural labor efficiency and agricultural production [42]. Rogito et al., showed in Kenya that limited land access is a major obstacle to young people’s involvement in agriculture, having an impact on all the stages of the value chain, with the exception of consumption [48].

However, other studies have found different results regarding the link between land access and agricultural productivity. In the Philippines, Michler and Shively examined the relationship between formalized property rights, land contracts and productive efficiency in agriculture using a stochastic production frontier model [27]. They found no link between land reform and land titling and the rental efficiency of land markets. The work of Ma et al., on the land security and technical efficiency in China using the stochastic production frontier shows that land certificates negatively affect technical efficiency [49]. Lawin and Tamini analyzed the impact of land ownership security on farmers’ technical efficiency in Benin based on a production-oriented stochastic distance function [28]. They used propensity score matching to correct for selection bias in the observed variables [28]. Their results show that non-owners have consistently higher levels of technical efficiency and productivity. The empirical analysis conducted at the micro level in Asia and Africa was based more on land security, but the other dimensions of access to land are also important, especially because of customary systems. This study differs from other studies identified in the literature since it takes into account the different land access modes.

3. Materials and Methods

This section deals with the geographical context of the study, conceptual model, operational concepts and the analysis model. Two softwares were used to analyze the dataset: SPSS Version 20.0 for Windows and Stata 16.

3.1. Geographical Context of the Study

Benin is a West African country located in the tropical zone between the equator and the Tropic of Cancer. It borders Niger to the north, Burkina-Faso to the northwest, Togo to the west, Nigeria to the east and the Atlantic Ocean to the south. Benin can be divided into five main regions: the coastal region, the terre barre plateau, the silica clay plateau, the Atacora mountains and the Niger plains. Benin has a surface area of 112,622 km², of which 38% is arable land. In Benin, as in most West African countries, political debates on land tenure often pit two modes of access to and relations with land against each other. On the one hand, there is custom and, on the other, colonial law, which defines the allocation of full private property to individuals. The significant increase in the urban population due to the rural exodus and the immigration of foreigners in search of better living conditions has led to strong pressure on urban land. This explains the reasons for the increase in land values in urban areas, with the consequent development of mainly informal land markets. In rural areas, the most common land conflicts can be categorized as boundary disputes, disputes related to contested property rights or inheritance disputes. These conflicts hinder investment by blocking land development. Benin’s land tenure system does not allow for secure industrial or agricultural investments. Failure to secure investment can jeopardize food sovereignty.

3.2. The Conceptual Model

The fundamental conceptual model linking land ownership and agricultural production growth was developed by Feder [36]. It establishes a unidirectional link between land ownership and agricultural productivity, from land ownership to agricultural productivity through investment. This model has been used by many authors to test the links between land ownership security, investment and productivity [43]. Based on Feder’s model, other authors have established that investment in land can have an impact on improving land security, leading to a bidirectional link [14,50]. The model of Feder [36] does not take into account the fact that greater ownership security may increase the use of variable inputs (e.g., fertilizer or labor). Another conceptual economic model linking land ownership and agricultural output growth has been developed by Place [43]. He states that the privatization of land rights can increase investment in agriculture and the purchase of variable inputs, thereby increasing productivity. But, his model does not incorporate the factor equalization effect of land markets. The model by Ma et al., on the other hand, takes these factors into account, as well as the potential impact of land security on rural–urban migration and its consequences for productivity and technical efficiency [49]. For these authors, production growth is the result of investment in land (investment effect), increased use of variable inputs (input effect), the transfer of land to more productive farmers (land market effect) and less efficient farm management practices resulting from rural–urban migration (migration effect).

The model developed in this study does not take into account the migration effects of the conceptual model of Ma et al. but adapts their model by focusing on inputs and investments in agricultural equipment resulting from the land market and secure access to agricultural land [49]. The investment effect and the input effect directly impact agricultural production, while the land market effect has an indirect impact through technical efficiency. It should also be noted that increased land security can indirectly affect agricultural investment through the increased availability of capital resulting from better access to the credit market. The input effect refers to the increased use of variable inputs, such as seeds, chemical fertilizers, pesticides and herbicides. Secured land can be used as collateral to improve access to credit where land sales markets exist. The increased availability of credit can stimulate the purchase of variable inputs [36]. The land market effect refers to the transfer of land to more efficient farmers. Figure 1 presents the conceptual model that explores the channels of land access and their effects on agricultural productivity.

Measuring Agricultural Productivity and Land Access Modes

Agricultural productivity is the level of the amount of agricultural output produced for a given amount of input or set of inputs [51]. There are several measures of productivity, for example, the amount of output per unit of input, or an index of many outputs divided by an index of many inputs [52]. Land yield or productivity, defined as total output divided by area sown [24,43], is generally used to assess the success of a new technology. It is also useful for determining the amount of land needed to meet future food demand [52]. It is a partial measure of productivity, focusing on land input and ignoring other factors of production and inputs. As this study focuses on land access modes, it uses agricultural yield as a measure of productivity, which is relevant for examining intensive farming systems.

The main mode of land acquisition in the African context is inheritance, and distribution among family members is conditioned by membership of the community through descent [53]. The different modes of land acquisition are land ownership, land rental, sharecropping, borrowed land or loans. Land owners are those who acquired the land by purchase or inheritance. Inheritance is the access mode by which a property is transmitted from a real or adoptive parent to an individual or group of individuals after the former’s death [53]. Land purchase is the method allowing the beneficiary to obtain ownership of a portion of land in return for payment. It enables ownership to be transferred and confers a lasting right on the purchaser. Land rental refers to a situation in which a family farms an area belonging to an owner to whom the farmer pays rent [54]. Renting is still very prevalent in the southern departments, where pressure on land is high. Sharecropping is a form of tenancy in which the farmer shares his crops with the owner [54]. In other words, it is a type of land lease in which an owner (lessor) entrusts another person, called a sharecropper, with the task of cultivating land in exchange for a harvest share. Borrowed land or loan is a situation in which a household lends land for cultivation and the owner expects it to show gratitude [53]. It is a land ownership arrangement whereby an individual temporarily uses a plot of land without paying a fee directly to the owner. A loan is a form of access that provides the beneficiary, also known as the borrower, temporary usufructuary rights without any formal compensation in cash or kind.

3.3. Analysis Method

The variables linked to land access are endogenous by definition as there are unobserved factors that can affect the productivity result via the error term. For example, it is possible that a household that does not have secure access to land (land borrowers) produces more than those with secure access (landlords). This may be due to the agricultural production capacity and the quality of the land available to this household. However, these characteristics are not observed and are therefore relegated to the term error. However, they are linked to productivity, which is why a simple OLS regression between access to land and productivity can lead to biased estimates.

Several techniques have been used in recent studies. These include propensity score matching (PSM) [55] and regime-switching multinomial regression (MESR) techniques [56,57,58]. Propensity score matching does not correct for selection bias from unobserved factors [59]. Unlike PSM, MESR can be used to correct for observable and unobservable bias using a selection bias correction method. This bias correction is conducted from the calculation of the inverse Mills ratio using truncated normal distribution theory and latent factor structure [56].

The MESR method involves fitting models in which the outcome variable depends on a dummy variable and is only observed if a particular selection condition is met. By modeling both the selection and outcome equations, MESR has the advantage of controlling for factors that affect the treatment itself and disentangling factors that influence outcome among treatment and control groups [60]. In addition to controlling for selection bias resulting from unobserved factors that may affect both land access and productivity variables, the MESR model controls for structural differences between individuals who have access to secure land and those who do not. The MESR model proceeds in two stages, and estimation is carried out simultaneously. The first stage is modeled using a multinomial logit choice model because the selection variable is multinomial. Next, the effect of access to land on productivity is estimated using the ordinary least squares (OLS) method with selectivity correction terms.

3.4. Multinomial Model of Land Access Selection

Access to land can be modeled in a random utility framework. Following Khonje et al. and Zhang et al., the probability of accessing land is a multinomial choice that leads to a positive difference between the utility of secure and insecure access to land [56,57]. Thus, the probability of an individual i choosing land tenure j can be specified by a multinomial logit model. Let ${{\mathit{F}}^{\mathbf{*}}}_{\mathit{i}\mathit{j}}$ be the latent variable of access to land for the individual i with several modalities j (j = 1...J): 1 = landowner, 2 = land tenant, 3 = sharecropper and 4 = borrowed land.

$${{\mathit{F}}^{\mathbf{*}}}_{\mathit{i}\mathit{j}}={\mathit{\alpha }}_{\mathit{j}}{\mathit{x}}_{\mathit{i}\mathit{j}}+{\mathit{\epsilon }}_{\mathit{i}\mathit{j}}$$

(1)

$${\mathit{F}}_{\mathit{i}\mathit{j}}=\left\{\begin{array}{c}\mathbf{1} \mathit{s}\mathit{i} {\mathit{F}}_{\mathit{i}\mathit{j} >}^{ \mathbf{*}}\underset{\mathit{m}\ne \mathbf{1}}{\mathbf{m}\mathbf{a}\mathbf{x}} \left({\mathit{F}}_{\mathit{m}\mathit{i}}^{\mathbf{*}}\right) \mathit{o}\mathit{r} {\mathit{\eta }}_{\mathbf{1}\mathit{i}}<\mathbf{0}\\ \mathit{J} \mathit{s}\mathit{i} {\mathit{F}}_{\mathit{i}\mathit{j} > }^{\mathbf{*}}\underset{\mathit{m}\ne \mathit{J}}{\mathbf{m}\mathbf{a}\mathbf{x}} \left({\mathit{F}}_{\mathit{m}\mathit{i}}^{\mathbf{*}}\right) \mathit{o}\mathit{r} {\mathit{\eta }}_{\mathit{j}\mathit{i}}<\mathbf{0}\end{array} \mathit{f}\mathit{o}\mathit{r} \mathit{a}\mathit{l}\mathit{l} \mathit{m}\ne \mathit{j}\right.$$

(2)

${\mathit{F}}_{\mathit{m}\mathit{i}}^{\mathbf{*}}$ denotes the latent variable for a mode m. Further, ${\mathit{x}}_{\mathit{i}\mathit{j}}$ denotes the deterministic components such as household and land characteristics and other determinants such as gender, marital status, level of education, income, access to credit and labor and the adoption of different technologies (improved seeds, herbicides, chemical fertilizers, manure and organic fertilizers, insecticides and fungicides). Finally, there is the random component ${\mathit{\epsilon }}_{\mathit{i}\mathit{j}}$.

Assuming that the error terms are independent and identically distributed, we have

$${\mathit{\eta }}_{\mathit{j}\mathit{i}}\underset{\mathit{m}\ne \mathit{J}}{=\mathbf{m}\mathbf{a}\mathbf{x}} \left({\mathit{F}}_{\mathit{m}\mathit{i}}^{\mathbf{*}}-{\mathit{F}}_{\mathit{j}\mathit{i}}^{\mathbf{*}}\right)<\mathbf{0}$$

(3)

Bourguignon et al., [61]

Thus, the probability of individual i choosing ownership system j is as follows:

$${\mathit{P}}_{\mathit{j}\mathit{i}}={\mathit{P}}_{\mathit{r}}\left({\mathit{\eta }}_{\mathit{j}\mathit{i}}<\mathbf{0}|{\mathit{x}}_{\mathit{j}\mathit{i}}\right)={\displaystyle \frac{\mathbf{e}\mathbf{x}\mathbf{p} \left({\mathit{\alpha }}_{\mathit{j}}{\mathit{x}}_{\mathit{j}\mathit{i}}\right)}{{\mathbf{\sum }}_{\mathit{m}\ne \mathbf{1}}^{\mathit{j}}\mathbf{e}\mathbf{x}\mathbf{p} \left({\mathit{\alpha }}_{\mathit{m}}{\mathit{x}}_{\mathit{j}\mathit{i}}\right)}}$$

(4)

McFadden, [62]

On the basis of this model, the endogenous treatment with regime change is established in order to study the treatment of access to land on agricultural productivity.

3.5. Multinomial Regime-Switching Model (MESR)

In the second stage of the MESR, we describe the strategy used to estimate the impact of access to land on agricultural productivity, which is measured by the yield of the first four main crops in the database used. The relationship between yield outcome variables and a set of explanatory variables Z (household and land characteristics, inputs and socio-economic factors) is estimated for each land access mode, following the multinomial selection bias correction framework approach of Bourguignon et al. [61]. The reference mode is j = 1 (owners) and the other access modes are renting, sharecropping and free loan (j = 2, 3, 4). The yield function Yij used to assess the implications of access in terms of productivity for each possible regime is provided as follows:

$$\left\{.\begin{array}{c}\mathit{R}\mathit{e}\mathit{g}\mathit{i}\mathit{m}\mathit{e} \mathbf{1}: {\mathit{Y}}_{\mathbf{1}\mathit{i}}={\mathit{\beta }}_{\mathbf{1}}{\mathit{z}}_{\mathbf{1}\mathit{i}}+{\mathit{\mu }}_{\mathbf{1}\mathit{i}} \mathit{s}\mathit{i} \mathit{F}=\mathbf{1}\\ .\\ .\\ .\\ \mathit{R}\mathit{e}\mathit{g}\mathit{i}\mathit{m}\mathit{e} \mathit{j}: {\mathit{Y}}_{\mathit{j}\mathit{i}}={\mathit{\beta }}_{\mathit{j}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\mu }}_{\mathit{j}\mathit{i}} \mathit{s}\mathit{i} \mathit{F}=\mathit{j} \end{array}\right.\mathit{j}=\mathbf{2},\mathbf{3},\mathbf{4},\mathbf{5}$$

(5)

The error terms in the selection equation ${\mathit{\epsilon }}_{\mathit{i}\mathit{j}}$ and those of the outcome equation ${\mathit{\mu }}_{\mathit{i}\mathit{j}}$ are not independent. A consistent estimate of βj requires the inclusion of the selection correction terms of the alternative choices in Equation (5); we have the following:

$${\mathit{Y}}_{\mathit{j}\mathit{i}}={\mathit{\beta }}_{\mathit{j}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\rho }}_{\mathit{j}}{\mathit{\lambda }}_{\mathit{j}\mathit{i}}+{\mathit{\mu }}_{\mathit{j}\mathit{i}} \mathrm{W}\mathrm{i}\mathrm{t}\mathrm{h} \mathrm{j} = 1,2,3,4,5$$

(6)

ρ is the covariance between error terms ${\mathit{\epsilon }}_{\mathit{i}\mathit{j}}$ and ${\mathit{\mu }}_{\mathit{i}}$. The endogenous regime-switching treatment model is augmented by the inverse Mills ratio λji in Equation (6) and which is calculated from the estimated probabilities in Equation (4) and can be represented as follows:

$${\mathit{\lambda }}_{\mathit{j}\mathit{i}}={\mathbf{\sum }}_{\mathit{m}\ne \mathit{j}}^{\mathit{J}}{\mathit{p}}_{\mathit{j}}\left[{\displaystyle \frac{{\widehat{\mathit{P}}}_{\mathit{m}\mathit{i}}\mathbf{l}\mathbf{n} ({\widehat{\mathit{P}}}_{\mathit{m}\mathit{i}})}{\mathbf{1}-{\widehat{\mathit{P}}}_{\mathit{m}\mathit{i}}}}+\mathbf{l}\mathbf{n} ({\widehat{\mathit{P}}}_{\mathit{j}\mathit{i}}\right];\mathit{p}$$

(7)

within the framework of multinomial choice, in order to improve identification, it is necessary that Xji include at least one instrumental variable (IV), which affects access to land but is excluded from zji. It does not directly affect yield. The distance of the plot from the residence served as an instrumental variable.

3.6. Estimation of Counterfactual Effects

The MESR model can be used to calculate counterfactual effects. To estimate the effects of land access on productivity, it is essential to know the real effects of land access and the counterfactual effects. The counterfactual is defined as the return to landowning households that would have been obtained if they did not own land and vice versa [57]. From Equation (6), the following conditional expectations for each return outcome variable can be calculated using the following four equations:

Real results for landowners:

$$\mathit{E}\left[{\mathit{Y}}_{\mathbf{1}\mathit{i}}|\mathit{F}=\mathbf{1},{\mathit{z}}_{\mathbf{1}\mathit{i}},{\widehat{\mathit{\lambda }}}_{\mathbf{1}\mathit{i}}\right]={\mathit{\beta }}_{\mathit{j}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\rho }}_{\mathit{j}\mathit{\epsilon }}{\mathit{\lambda }}_{\mathbf{1}\mathit{i}}$$

(8)

Real results for non-landowners:

$$\mathit{E}\left[{\mathit{Y}}_{\mathit{j}\mathit{i}}|\mathit{F}=\mathit{j},{\mathit{z}}_{\mathit{i}\mathit{j}},{\widehat{\mathit{\lambda }}}_{\mathit{j}\mathit{i}}\right]={\mathit{\beta }}_{\mathit{j}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\rho }}_{\mathit{j}\mathit{\epsilon }}{\mathit{\lambda }}_{\mathit{j}\mathit{i}}$$

(9)

Counterfactual results for landowners:

$$\mathit{E}\left[{\mathit{Y}}_{\mathit{j}\mathit{i}}|\mathit{F}=\mathbf{1},{\mathit{z}}_{\mathbf{1}\mathit{i}},{\widehat{\mathit{\lambda }}}_{\mathbf{1}\mathit{i}}\right]={\mathit{\beta }}_{\mathit{j}}{\mathit{z}}_{\mathbf{1}\mathit{i}}+{\mathit{\rho }}_{\mathit{j}\mathit{\epsilon }}{\mathit{\lambda }}_{\mathbf{1}\mathit{i}}$$

(10)

Counterfactual results for non-landowners:

$$\mathit{E}\left[{\mathit{Y}}_{\mathbf{1}\mathit{i}}|\mathit{F}=\mathit{j},{\mathit{z}}_{\mathit{j}\mathit{i}},{\widehat{\mathit{\lambda }}}_{\mathit{j}\mathit{i}}\right]={\mathit{\beta }}_{\mathbf{1}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\rho }}_{\mathbf{1}\mathit{\epsilon }}{\mathit{\lambda }}_{\mathit{j}\mathit{i}}$$

(11)

Equations (8) and (9) represent actual productivity for owner and non-owner households, while Equations (10) and (11) represent counterfactual productivity. Therefore, the effect of the average treatment on the treated (ATT) can be derived by subtracting Equation (9) from Equation (11):

$$\begin{gathered} \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\mathit{A}\mathit{T}\mathit{T}=\mathit{E}\left[{\mathit{Y}}_{\mathit{j}\mathit{i}}|\mathit{F}=\mathit{j},{\mathit{z}}_{\mathit{i}\mathit{j}},{\widehat{\mathit{\lambda }}}_{\mathit{j}\mathit{i}}\right]-\mathit{E}\left[{\mathit{Y}}_{\mathbf{1}\mathit{i}}|\mathit{F}=\mathit{j},{\mathit{z}}_{\mathit{j}\mathit{i}},{\widehat{\mathit{\lambda }}}_{\mathit{j}\mathit{i}}\right] \\ \begin{array}{l}={\mathit{\beta }}_{\mathit{j}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\rho }}_{\mathit{j}\mathit{\epsilon }}{\mathit{\lambda }}_{\mathit{j}\mathit{i}}-{\mathit{\beta }}_{\mathbf{1}}{\mathit{z}}_{\mathit{j}\mathit{i}}+{\mathit{\rho }}_{\mathbf{1}\mathit{\epsilon }}{\mathit{\lambda }}_{\mathit{j}\mathit{i}}\\ \mathit{A}\mathit{T}\mathit{T}={\mathit{z}}_{\mathit{j}\mathit{i}}({\mathit{\beta }}_{\mathit{j}}-{\mathit{\beta }}_{\mathbf{1}})+{\mathit{\lambda }}_{\mathit{j}\mathit{i}}({\mathit{\rho }}_{\mathit{j}\mathit{\epsilon }}-{\mathit{\rho }}_{\mathbf{1}\mathit{\epsilon }})\end{array} \end{gathered}$$

(12)

3.7. Data and Study Variables

3.7.1. Data Source

Data from the Harmonized Survey of Household Living Conditions in WAEMU countries (EHCVM) in 2018 are used for this test. The EHCVM surveyed a total of 8012 households, including 42,343 individuals grouped into 23 strata in Benin. Out of the 8012 households, there were 3120 farmers who represent the analysis sample in this study. The survey covers the whole country, and the results are significant in rural areas, urban areas and administrative regions of the country. The survey covers the entire resident population, but for this study we considered the 3120 farming households that have access to land by at least one mode of transport and whose main activity is farming.

3.7.2. Study Variables

We have the outcome variable, which is the yield of the first four main crops. Then, there is the selection variable, which is access to land. There are three groups of explanatory variables: socio-economic variables such as gender, level of education, access to credit, income and food and non-food expenditure. In addition, technological variables include agricultural yield, the total quantity of inputs used, chemical fertilizers, organic fertilizers, insecticides and fungicides, the use of family labor and the amount of wage labor. Finally, the characteristics of the land include land ownership status, plot management mode, mode of acquisition, distance of the plot from the residence and place of residence. Yield is production per hectare. It is a measure of partial productivity, also known as land productivity. The technologies listed are the inputs to the production process that are positively linked to yield. The way in which the plot is managed is linked to custom, with some communities managing land collectively and others individually. The different modes of ownership in this trial are landlord, free loan, tenant farming and sharecropping. The instrument used in the study is the distance of the plot from the residence, which is directly related to the land but not to the yield (

Table 1. Variable descriptions and measurements. Source: authors own work.

Variable			Description	Measures	Signs
Socio-economic variables	Gender		binary	0 = male, 1 = female	+/−
	Level of education		binary	0 = no, 1 = yes	+
	Access to credit		binary	0 = no, 1 = yes	+
	Income		continuous	F cfa	+
	Food and non-food expenditure		continuous	F cfa	+/−
Technological variables	Agricultural yield		continuous	Kg/ha
	The total quantity of input used		continuous	kg	+
	Manure/organic fertilizer		binary	kg	+
	Chemical fertilizer		binary	kg	+
	Salaried workforce		continuous	F cfa	+
	Farm labor		binary	0 = no, 1 = yes	+
The characteristics of the land	Land ownership	Owners	Categorical	0
		Free loan		1	−
		Farming		2	−
		Sharecropping		3	−
	Management of the plot		binary	0 = individual, 1 = collective	−
	Distance from the plot		continues	time	+/−
	residence		binary	0 = Rural, 1 = Urban,	−/+

). The selection of the distance of the plot from the residence as the instrumental variable in our study was based on the following considerations: in rural areas, the proximity of farmland to the household can significantly affect how land is accessed and managed. For example, farmers can choose between different arrangements (leasing or sharecropping versus direct ownership or free loan).

4. Results and Discussion

This section presents the descriptive analysis, the results of the selection model, the outcome model and the treatment effect and discussion. The selection equation presents the determinants of access to land, and the outcome equation presents its impact on the returns. The base category is landowners.

4.1. Descriptive Analysis

A descriptive analysis of the different types of land ownership shows that landowners are in the majority with a percentage of 64.04%, followed by those who have taken out free loans, while farmers and sharecroppers are in low proportions with 4.55% and 3.97%, respectively (Figure 2).

The descriptive analysis also shows that the majority of the owners acquired their land by inheritance (83.18%), 10.71% by gift and only 6.11% by purchase (Figure 3).

Tenant farmers have an agricultural yield of 521,094 kg/ha, followed by farmers with a yield of 618,715 kg/ha, free tenants with a yield of 917.21 kg/ha and finally landlords with the highest level of agricultural yield, 2269.641, for a total average yield of 1753.957 (Figure 4).

On average, 87.5% of the farmers manage their land individually and 12.5% collectively, with a female proportion of 14.1% and a male proportion of 85.9%. The collective management was 11.11% and individual management 88.89 for the owners, with a female proportion of 11.36. Those who have taken out free loans manage collectively at a rate of 15.07% and individually at 84.93%, with 18.57% being women. The individual management through tenant farming is 83.1% and collective is 16.9%, with the same proportion for women. Finally, 87.9% of the tenant farmers manage their land individually and 12.1% collectively, with a higher proportion of women (24.19%). Women are more likely to be sharecroppers, while there are fewer women owners. These statistics support the idea that women are marginalized by customary systems, especially in Africa (Figure 5a,b).

An analysis of

Table 2. Determinants of access to land and agricultural yield. Source: authors own work (model estimations).

Land Tenure	Income (cfa)	Food Expenditure (cfa)	Non-Food Expenditure (cfa)	Total Quantity of Input (kg/ha)	Distance From Plot (mn)	Access to Credit	Level of Education	Family Workforce
Owner	168901.5	12215.94	1668.412	451.7143	34.9965	6.96	6.36	88.79
Free loan	81551.75	8912.749	1687.013	106.3504	33.54673	6.54	11.33	84.93
Farming	80110.56	7370.922	1428.035	422.5063	35.02113	6.34	11.97	88.73
Sharecropping	45148.59	8369.758	1908.771	32.81129	36.07258 *	4.23	12.1 *	95.97 *
Total	135976.8	10937.76	1672.197	338.9826	34.64263	6.83	8.21	88.01

* This refers to the highest percentage of these variables at the level of the sharecroppers.

shows that landowners have a higher average income of XOF 168901.5, followed by those who have taken out free land loans (XOF 81551.75), land-renting households (XOF 80110.56) and sharecroppers (XOF 45148.59), for a total average income of XOF 135976.8. The landowners spent more on food (XOF 12215.94), while the tenant farmers spent less (XOF 7370.922). In terms of non-food expenditure, the tenant farmers spent more (XOF 1908.771) and farmers spent less (XOF 1428.035). Landlords and farmers use more agricultural inputs (451.7143 kg/ha and 422.5063 kg/ha), while tenant farmers and lenders use fewer agricultural inputs (32.81129 kg/ha and 106.3504 kg/ha). These statistics show a very wide gap in terms of input use between landowners and tenant farmers. The distance of the plot from the residence is at least 36 min greater for the tenant farmers, and they are better educated (12.1%) than the majority of the landowners (6.36%). These tenant farmers make greater use of family labor (95.97%), whereas the lenders make less use of family labor (84.93%). Finally, the landlords have more access to credit (6.96%), whereas the tenant farmers have less (4.23%).

4.2. Determinants of Different Land-Use Patterns

The results of the selection model in

Table 3. Selection model. Source: authors own work (model estimations).

Variable	Free Loan	Farming	Sharecropping
Food expenses	−0.253 ***	−0.269 ***	−0.148
	(0.051)	(0.083)	(0.099)
Non-food expenditure	0.160 **	−0.153	0.059
	(0.073)	(0.130)	(0.150)
Access to credit	−0.035	−0.062	0.085
	(0.195)	(0.380)	(0.377)
Gender	−0.580 ***	−0.376	−0.849 ***
	(0.136)	(0.253)	(0.239)
Level of education	0.700 ***	0.693 **	0.663 **
	(0.170)	(0.301)	(0.311)
Family workforce	0.368 **	−0.056	−1.536 ***
	(0.144)	(0.299)	(0.479)
Total input expenditure	0.031 ***	0.003	0.010
	(0.011)	(0.020)	(0.021)
Total expenditure on agricultural equipment	−0.005	−0.036	−0.100 *
	(0.020)	(0.043)	(0.059)
Distance from plot	−0.065	0.236 **	−0.001
	(0.047)	(0.099)	(0.096)
Management of the plot	0.443 ***	0.481 *	0.207
	(0.144)	(0.258)	(0.296)
Constant	0.099	−0.131	−0.001
	(0.579)	(0.935)	(1.255)
Comments	3120		Wald chi2 (41) 695.78
Log pseudolikelihood	−9735.7312		Prob > chi2 0.000

Standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

show that non-food expenditure, level of education and family labor positively determine lenders’ access to land, whereas income, food expenditure, gender and total quantity of inputs negatively determine it. This implies that being a landowner is more attractive to women and improves income levels, food expenditure and the total quantity of inputs than being a land lender. These results could be explained by the fact that women have more access to land through loans than through ownership. These results are in line with those from the descriptive analysis, which already justifies that the majority of the landowners acquired their land through inheritance. However, because of the customary systems, women are often marginalized in terms of inheriting agricultural land. Also, because landowners have more secure access to land, they can invest in land and inputs, which will enable them to have a higher level of income, so they can increase their food expenditure.

In the case of farmers, the level of education is a positive determinant, whereas non-food expenditure and the total quantity of inputs are negative determinants; i.e., these determinants favor landlords more. For sharecropping, the level of education is a positive determinant of access, while income, gender and labor determine access negatively. We find that being educated is more advantageous for tenant farmers than for landlords. These results can be explained by the fact that tenant farmers’ sharing contracts comply more with the standards when they are educated. It should be noted that the instrument of collective management suits the lenders and the farmers more than the owners, whereas the distance of the plot from the dwelling only benefits the farmers. Collective land management can lead owners to invest less in the land because the interest is collective. Also, the further away the plot is, the more it benefits the farmers. This can be explained by the fact that the cost of renting the plot is lower the further you are from home. It should also be noted that these instruments are valid because the coefficients are significant.

4.3. The Impact of Land on Farm Yields

The results in

Table 4. Estimation of land access modes and their impact on farm yields using the multinomial regime-switching model. Source: authors own work (model estimations).

Variable		Agricultural Yield	Variable	Agricultural Yield
Land	Lender	−0.680 **	Total expenditure on agricultural equipment	0.044 ***
		(0.314)		(0.017)
	Farming	−0.900 ***	Food expenditure	(0.240)
		(0.335)		0.182 ***
	Sharecropping	−3.069 ***	Non-food expenditure	(0.043)
		(0.240)		−0.175 ***
Sigma		0.252 *	Access to credit	(0.062)
		(0.142)		−0.150
Rho1		−0.014	Type	(0.164)
		(0.353)		0.328 ***
Rho2		−0.273	Level of education	(0.122)
		(0.293)		−0.589 ***
Rho3		1.872 ***	Family workforce	(0.154)
		(0.133)		−2.115 ***
			Total input expenditure	(0.129)
				0.007

Standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

show the impact of the different land access modes on agricultural yield. These results show the negative impact of the different access modes—loan, tenancy and sharecropping—on the agricultural yield. In other words, access by these different methods has a greater negative impact on the yield than land ownership. These results are in line with the classical and neoclassical view of Besley, which stipulates that secure access to land benefits farmers [14]. The results also confirm the Marshallian inefficiency hypothesis, according to which an agent deploys less effort than optimal if he or she does not have a full residual claim on the production [39]. In other words, in the absence of secure land, the threat of eviction on the returns from investments in production implies that tenant farmers will make less effort in production (Newman et al., 2015) [23]. Also, the neoclassical view is that the landowner can pledge the plots to have access to credit and invest in technologies to improve the agricultural yield. These results corroborate the work of Bergaly, Rashid, Chen et al., Keudem and Savadogo and Rogito and Makhanu, who have shown that land certification facilitates leasing and improves labor efficiency and agricultural productivity [11,42,46,47,48]. According to these authors, limited access to land is a major obstacle to young people’s involvement in agriculture and their impact on the agricultural value chain.

The positive impact of expenditure on agricultural equipment on yield means that investment in equipment through access modes improves the agricultural yield. Non-food expenditure, level of education and family labor have negative impacts on agricultural yield, whereas income, food expenditure and total quantity of inputs have positive impacts on agricultural yield. The positive effect of income can be explained by the fact that a higher level of income allows farmers to invest in land and inputs for a better agricultural yield. These farmers can also reduce their investments by spending more on non-food items. This may explain the negative effect of non-food expenditure on yield. The positive effect of food expenditure can be explained by the fact that additional access to food improves the cognitive and productive capacity of farmers, which in turn increases agricultural yields.

4.4. Results of Counterfactual Effects (ATT)

Table 5. Counterfactual analysis and Average Treatment Effect (ATT). Source: authors own work.

Access Mode	Group	Gain (kg/ha)	ATT (kg/ha) = 0−1	t-Value
Owner (result)		509.9342
Lender	0	618.7491	396.615	20.2438 ***
	1	222.1341
Farmer	0	532.0126	485.103	11.0858 ***
	1	46.90965
Sharecropper	0	525.4262	389.7975	8.2789 ***
	1	135.6286

*** p < 0.01. Annotations: Group 0 refers to landowners, which is the baseline category. Group 1 is either lender, farmer or sharecropper.

presents the results of the estimates of the productivity gains resulting from access to different types of land tenure. According to the results in Table 4, being a landowner has a positive and significant impact on the agricultural yield. The average effect of the treatment on the treated (ATT) is positive, reflecting the existence of a productivity gain linked to landowner access. This gain is estimated at 509.9342 kg/ha. This is followed by a gain of 396.615 kg/ha for lenders if they were landowners, 485.103 kg/ha for farmers if they were landowners and 389.7975 kg/ha for sharecroppers if they were landowners. We note a high gain for tenant farmers, which implies that being a tenant farmer is less convenient for farmers than other forms of access to land. In the light of these results, being a landowner favors farmers compared with other forms of access, such as being a farmer, a moneylender or a sharecropper. This may be due to the fact that farmers, whether lenders, farmers or sharecroppers, do not want to take the risk of investing in heavy equipment involving high investment costs with the added risk of expropriation. These methods of access do not instill confidence in non-owner farmers.

Improving agricultural productivity requires appropriate investment conditions. Investments based on the type of access to land improve farmers’ productivity. This is because the different types of access to land determine the motivations of each household or individual in making investment decisions. Land security is captured by land ownership, while land insecurity is captured by credit, tenancy and sharecropping. Households with insecure tenure invest less than households with secure tenure (landowners). What is more, landowners can pledge their land to obtain credit and finance production.

5. Conclusions and Implications of the Study for Public Policy, the Economy and Society

The results of the study show that access to land has a significant impact on agricultural yields. Access through free loans, tenant farming and sharecropping have negative impacts on agricultural yields. This means that being a landowner favors agricultural yields compared to not being a landowner. Also, non-food expenditure, level of education and family labor have negative impacts on agricultural yields, whereas income, food expenditure and total expenditure on agricultural equipment have positive impacts on agricultural yields. Better still, the counterfactual analysis indicates a positive gain in access to land for landowners.

These results show that land policies need to review the customary rights in order to reduce inequalities in access to secure land. Improving agricultural yields depends on securing land, which requires an effective land ownership system to be put in place. This means, for example, putting into practice the 2013 and 2017 laws prohibiting land grabbing by the elite, and also identifying cultivable areas by zone and setting prices per square meter. This implies that access to land, the formalization of land rights and investment in agricultural equipment must be prioritized by the institutions concerned in order to improve agricultural productivity. The policies must strike a balance between the legal and customary rights and historical practices to ensure a long-term vision. In other words, land ownership security must respect the historical rights, customs and legitimate sources in the interests of equitable and sustainable development.

In Benin, land is considered to be an inherited asset that is passed down from father to son and from one generation to another. Households with customary rights are therefore more dominant than those with legal rights. The very low level of education among the farmers explains the lack of interest among households in holding legal property rights. The positive impact observed at the landowner level proves that farmers’ security of ownership is not exclusively guaranteed by the possession of legal rights. Indeed, as the cost of obtaining titles is high, rational households often trade off the expected gains against the costs of holding a legal land certificate. As a result, these households are reassured about their inherited property, and they invest appropriately in their farms because they are sure of making a profit. The weakness of Benin’s institutions and the land market are also areas of action that can slow agricultural productivity. For example, investment in agricultural equipment, which should help to increase productivity and improve the market opportunities for smallholders, can provide healthy, nutritious food at reasonable prices to the ever-growing rural, urban and semi-urban populations. It can also increase opportunities for trade in agricultural products, raise incomes and ultimately improve the food security in Benin.

Finally, like all research, this study has its limitations. As the study focuses on agricultural land, the proportion of farmers with at least one agreement of sale is very low. Alternatively, at the level of landowners, we could separate those who hold title deeds or an agreement of sale from those who do not. This would make it possible to capture the difference in the effect of more secure access and simple inheritance at the landowner level. But, the proportion of farmers who have at least one agreement of sale would make econometric estimates very difficult, if not impossible. Only 6.11% of the land is acquired by purchase, and the majority of the land purchased continues to be the subject of land disputes because they do not hold formal title deeds. A larger sample for future research could solve these problems.