Analysis of Monetary and Multidimensional Poverty Drivers Among Agricultural Households in Togo Using a Weighted Logit Framework

Abstract

Assessments of poverty among agricultural households in Sub-Saharan Africa often rely on either monetary or multidimensional indicators considered separately, overlooking key structural constraints. This study investigates the determinants of both monetary and multidimensional poverty among agricultural households in Togo. Using nationally representative EHCVM 2021/22 data from 2893 households, monetary poverty is measured using the Foster–Greer–Thorbecke Index, while multidimensional poverty is assessed with the Alkire–Foster method. A survey-weighted logit model is employed to identify the drivers associated with each poverty dimension. Results show that multidimensional poverty (59.40%) is more widespread than monetary poverty (51.50%). Education substantially reduces poverty risk, whereas larger household size, limited market access, and residence in the Savannah region increase it. Economic and natural shocks are negatively associated with monetary and absolute poverty, while cooperative membership raises the likelihood of being poor. Investment in livestock (TLU) reduces monetary poverty but increases multidimensional deprivation. These findings highlight that poverty among agricultural households in Togo is shaped by interconnected socioeconomic and institutional constraints rather than income deprivation alone. Therefore, integrated strategies aligned with the Sustainable Development Goals, particularly those promoting education, rural credit access, market integration, and resilience-building, are essential for achieving effective and context-specific poverty reduction.

1. Introduction

Poverty remains one of the most persistent and complex challenges confronting global sustainable development. It is widely recognized as a multidimensional phenomenon that extends beyond income deprivation to include deficiencies in health, education, nutrition, housing, and access to essential services. According to the United Nations Development Programme [1], approximately 1.1 billion people worldwide experience multidimensional poverty, with 455 million residing in conflict-affected regions. These patterns highlight the need for evidence-based and context-specific poverty reduction strategies, especially in low-income and agriculture-dependent economies across Sub-Saharan Africa. Persistent poverty also undermines progress toward the Sustainable Development Goals (SDGs), particularly SDG 1 (No Poverty) and SDG 2 (Zero Hunger), reinforcing its central role as both a development and sustainability challenge.

Togo, a small West African nation, continues to face high levels of poverty, particularly in rural areas where agriculture remains the main source of livelihood and 56% of the population resides [2]. Rural households experience multiple and overlapping deprivations linked to their reliance on rain-fed agriculture, limited access to infrastructure and financial services, and repeated exposure to climatic and economic shocks. The agricultural sector is further constrained by unequal access to modern technologies, leaving many smallholders in low-productivity systems and reinforcing disparities with better-resourced farmers. As a result, inequalities persist, and poverty remains deeply entrenched among the most vulnerable households. Recent estimates show that 45.5% of the population lived below the national poverty line in 2018, while rural poverty exceeded 49% under the international $1.90-per-day threshold. Although the national Multidimensional Poverty Index declined from 37.1% in 2018–2019 to 28.9% in 2021–2022, nearly one-third of households remain deprived across several indicators of well-being [2].

Previous studies on Togo have further documented these structural disparities. [3] highlights persistent multidimensional deprivation across key welfare indicators, while [4] emphasizes deep regional inequalities, particularly in the Savannah region. Other empirical works show that financial inclusion, microfinance, and targeted fertilizer subsidies can reduce monetary poverty [5,6]. However, these studies typically analyze monetary and multidimensional poverty separately, offering limited insight into how different poverty drivers operate simultaneously across multiple deprivation dimensions.

Given that agriculture employs roughly 30% of the labour force and contributes over 20% of GDP [7], understanding the mechanisms that keep agricultural households in poverty is essential for designing effective and sustainable interventions. Existing development efforts have tended to prioritize production objectives while overlooking the broader household-level, institutional, and contextual constraints that shape poverty dynamics. A growing body of evidence highlights the importance of personal characteristics [8,9], household decisions [10,11,12,13], and structural factors affecting access to resources, exposure to risks, and opportunities for diversification [14], underscoring the need for a more integrated analysis of agricultural households’ poverty in Togo. Yet, evidence remains scarce on how these drivers jointly influence monetary and multidimensional poverty when assessed within a unified analytical framework.

To address this gap, the study pursues three interrelated objectives:

• measure monetary and multidimensional poverty among agricultural households;
• identify the household-level and institutional drivers associated with each poverty dimension;
• compare how these drivers differ across monetary, multidimensional, and absolute poverty indicators to provide deeper insights into rural deprivation in Togo.

These objectives motivate the following research questions:

• What are the levels of monetary, multidimensional, and absolute poverty among agricultural households in Togo?
• What factors drive each poverty outcome?
• Do these drivers differ across poverty measures?

By integrating these poverty assessments into a cohesive framework, the study enhances the analysis of poverty and provides critical insights for bolstering rural resilience, shaping effective poverty-reduction strategies, and refining social and institutional support systems.

2. Literature Review

2.1. Socio-Demographic Drivers

Socio-demographic drivers such as age, gender, education, household size, and marital status of the head of the household are among the most consistently analyzed variables in poverty research, although their effects vary across contexts. Gender differences remain a central concern: numerous studies [15,16,17] show that female-headed households tend to experience higher poverty rates due to systemic inequalities in land ownership, credit access, and participation in agricultural extension programmes. Structural barriers, including patriarchal inheritance systems, further restrict women’s ability to accumulate productive assets. Nonetheless, heterogeneity exists within this pattern, as some evidence suggests that widowed or divorced female heads may reinvest more heavily in human capital, especially education, as a resilience strategy [17].

Education is consistently cited as a critical driver of household welfare. Studies [18,19] demonstrate that higher educational attainment enhances income-generating capacity by facilitating access to off-farm employment and adoption of modern agricultural practices. Education also improves households’ ability to interpret climate information, engage with extension services, and participate effectively in markets. However, in environments with weak institutional support, other drivers such as access to credit, agricultural assets, or social protection can be more influential in shaping welfare outcomes [20]. Household size and composition also matter. Larger households may benefit from greater labour availability [15,16], yet higher dependency ratios can exacerbate consumption pressure and deepen poverty traps, depending on resource endowments and productivity levels.

2.2. Household Decision-Making Drivers

Beyond individual characteristics, household decision-making drivers, including land ownership, farm size, diversification, and technology adoption, play a critical role in determining welfare outcomes. Secure land tenure and larger holdings are positively associated with investment incentives, access to credit, and economies of scale [21,22]. Similarly, crop diversification can reduce vulnerability by stabilizing income and mitigating the effects of climatic and market shocks [23], although for resource-constrained farmers it may dilute labour and inputs, reducing overall efficiency [21].

Off-farm activities, including wage employment, petty trade, and remittances, are key income-smoothing mechanisms that help households cope with agricultural seasonality and shocks [24,25,26]. Nevertheless, participation in lucrative off-farm sectors often depends on education, capital availability, and rural infrastructure. For the poorest households, off-farm work may remain limited to low-return casual labour, offering minimal upward mobility. Adoption of modern agricultural technologies, such as improved seeds, irrigation, and mechanization, has also been linked to poverty reduction [27], but these benefits are unequally distributed, favouring better-endowed households with stronger institutional support. Cooperative membership enhances access to markets and financial resources through collective efforts [28,29], although governance weaknesses and elite capture undermine these advantages [13,30]. Moreover, ownership of livestock provides households with financial stability because it generates income and functions as an emergency fund during times of crisis [31,32,33].

2.3. Institutional and Contextual Drivers

Finally, institutional and contextual drivers, including access to credit and markets and exposure to shocks, significantly influence poverty dynamics. In Togo, recent initiatives promoting financial inclusion, value-chain integration, and climate resilience [4,34] illustrate efforts to address structural barriers to rural development. The implemented interventions have not been enough to overcome the challenges of unbalanced infrastructure and restricted access to formal credit and periodic climate-related disasters. The combination of droughts and pest outbreaks, with individual events such as illness or job loss, makes agricultural households more vulnerable to risks. The main policy focus now centres on building social protection networks and creating index-based insurance programmes and climate-resistant agricultural practices to protect against risks and achieve lasting poverty reduction.

Research on poverty outcomes among smallholder farmers continues to grow, yet there are still considerable gaps in understanding how household and contextual drivers interact to influence these outcomes. Existing studies in Togo and Sub-Saharan Africa focus on either monetary poverty [5,6,35] or multidimensional [3,36,37,38] measures of poverty in isolation, overlooking how the same set of drivers may influence both dimensions differently. This study, by applying identical explanatory variables across the two poverty frameworks, the FGT index and the Multidimensional Poverty Index (MPI), allows a direct comparison of their drivers and highlights how combining both measures into a composite “absolute poverty” indicator provides a more integrated understanding of rural deprivation.

Furthermore, while many empirical analyses rely on nationally representative secondary data, few explicitly account for the complex survey design and sampling weights in their econometric estimation. To address this limitation, the present study employs a survey-weighted logit model, ensuring that the estimates accurately reflect the national structure of agricultural households and increasing the robustness and policy relevance of the findings.

By integrating these methodological innovations, this paper contributes to the growing literature on poverty measurement and sustainable development in developing countries. It sheds light on how demographic drivers, household dynamics, and institutional constraints interact to shape welfare outcomes in Togo, providing practical guidance for inclusive and context-sensitive poverty reduction strategies among agricultural households.

3. Theoretical and Conceptual Framework

3.1. Theoretical Framework

To capture poverty’s complexity and identify sustainable solutions, this study draws on the complementary perspectives outlined by [39], which together provide a coherent foundation for analyzing socio-demographic, household decision-making, and institutional and contextual drivers of poverty among agricultural households in Togo.

The first perspective is the Individualistic theory, introduced by [40]. It demonstrates the role of individual characteristics, suggesting that abilities, behaviours, and life choices can increase or reduce the risk of poverty. Closely linked is the Cultural Theory of Poverty, advanced by [41] and rooted in [42] earlier work. This theory highlights how norms, values, and beliefs passed from one generation to the next may perpetuate poverty within communities [41,42,43]. The two perspectives demonstrate that poverty dynamics depend heavily on socio-demographic characteristics and household decision-making processes.

A second theoretical strand is provided by the Sustainable Livelihoods Framework (SLF) [44], which explains household well-being in terms of access to human, natural, physical, financial, and social assets within a specific vulnerability context. According to [45], the SLF conceptualizes a livelihood as sustainable when it can cope with and recover from shocks while maintaining or enhancing its asset base. This framework highlights how agricultural households mobilize and combine different forms of capital to make production, investment, and diversification decisions, and how these choices ultimately influence poverty outcomes.

Finally, poverty is shaped by broader structural and spatial conditions. The Structural Theory argues that inequality is reproduced through institutional barriers that limit access to markets, finance, and productive opportunities. Building on this view, the Capability Approach developed by Sen conceptualizes poverty as a deprivation of essential capabilities such as health, education, and security, which are necessary for achieving a dignified life [45]. Complementing these perspectives, the Geographical theory of poverty highlights how spatial inequalities influence welfare outcomes. Weber emphasizes that communities located in structurally disadvantaged regions often face limited institutional presence and reduced economic opportunities, which constrain their ability to improve living conditions [46]. Similarly, Foster’s work on spatial deprivation shows that regional disparities can create poverty traps, where limited access to services and markets reinforces persistent deprivation across territories [47]. These structural and spatial constraints restrict income diversification, reduce access to education and healthcare, and reinforce both monetary and multidimensional poverty.

3.2. Conceptual Framework

Building on the theoretical foundations presented above, the study conceptualizes poverty among agricultural households as the outcome of three interrelated domains: socio-demographic drivers, household decision-making drivers, and institutional or contextual drivers. These domains translate the theoretical insights from individualistic, cultural, livelihood, structural, and geographical perspectives into observable variables suitable for empirical analysis.

Socio-demographic drivers reflect the human capital and sociocultural mechanisms emphasized by individualistic and cultural theories. These include the age and education of the household head, household size, gender, and marital status. These variables represent personal endowments, capabilities, and social norms that shape households’ economic decisions and vulnerability. Accordingly:

H1: 

Higher education of the household head reduces the likelihood of monetary and multidimensional poverty.

H2: 

Larger household size increases poverty risk due to higher dependency pressures.

Household decision-making drivers are grounded in livelihood and behavioural theories, which explain how families allocate land, labour, and capital to secure and improve their livelihoods. These variables include land ownership, farm size, crop diversification, participation in off-farm activities, irrigation practices, ploughing methods, cooperative membership, and livestock ownership (TLU). These reflect strategic choices that influence income stability, production potential, and vulnerability. Hence:

H3: 

Participation in off-farm activities reduces the likelihood of monetary and multidimensional poverty by diversifying income sources and reducing reliance on agriculture.

H4: 

Larger farm size decreases poverty risk, as greater land endowment enhances production potential and livelihood security.

Institutional and contextual drivers capture the broader structural, environmental, and spatial constraints that shape welfare outcomes. Consistent with the Structural Theory and the Capability Approach, these include access to credit, access to markets, and exposure to covariate shocks (droughts, rainfall variability, pests) or idiosyncratic shocks (illness, job loss, livestock disease).

Additionally, geographical inequalities, as highlighted by Weber’s theory of spatial disadvantage and Foster’s work on spatial deprivation, remain crucial contextual drivers in Togo. Northern regions, particularly the Savannah region, face weaker infrastructure, limited market integration, lower service provision, and greater environmental vulnerability, all of which reinforce both monetary and multidimensional poverty. Therefore:

H5: 

Households located in northern regions, especially the Savannah region, are more likely to experience monetary and multidimensional poverty.

H6: 

Better market access reduces the likelihood of poverty among agricultural households.

This conceptual framework shows that agricultural household poverty does not depend on a single set of drivers, but rather emerges from the interaction between internal characteristics and external constraints.

Figure 1 presents the interaction between drivers and the poverty status of the agricultural household.

4. Materials and Methods

4.1. Data Source

The data used in this research originates from the 2021/22 Harmonized Survey on Household Living Conditions (EHCVM) in Togo, which was carried out by the National Institute of Statistics and Economic and Demographic Studies (INSEED) with assistance from the World Bank and the WAEMU Commission, as part of the regional Household Living Conditions Survey Harmonization Project (P153702). This nationally representative survey encompassed all administrative regions, including both urban and rural settings, and involved interviews with 6462 households (2459 urban and 4003 rural). Data collection occurred in two phases (October 2021–January 2022 and April–July 2022) to account for seasonal variations in consumption, employing household/individual and community questionnaires. For this analysis, 3772 agricultural households were identified, with 2893 households providing complete information for further examination.

Approximately 879 households (about 23% of the agricultural sample) were excluded from the analysis due to missing information on key variables, including household income, access to credit, and cooperative participation. Diagnostic checks indicated that these missing values resulted mainly from item non-responses during interviews rather than systematic exclusion of specific groups, suggesting that the data are missing at random (MAR). To mitigate potential bias, the analysis employed survey weights provided by INSEED, which adjust for unequal probabilities of selection and partially correct for non-response. Nevertheless, the exclusion of these households could introduce a minor reduction in overall representativeness. However, descriptive comparisons between included and excluded observations revealed no systematic differences across regions or household characteristics, suggesting that the potential selection bias is limited. This limitation is nonetheless acknowledged in the interpretation of results, particularly regarding the generalization of findings to the national population of agricultural households. Figure 2 describes the sample section process.

4.2. Study Area

This research focuses on the Republic of Togo, a small West African nation projected to have a population exceeding 9 million by 2024 and covering an area of approximately 56,600 km². Agriculture plays a pivotal role in the rural economy, engaging around 30 percent of the workforce and serving as the primary source of livelihood for many households. The country is administratively divided into five regions: Maritime, Plateaux, Centrale, Kara, and Savannah, each representing unique agroecological zones with distinct soil types, rainfall patterns, and vegetation. These ecological variations lead to diverse farming systems and income-generating opportunities, influencing household welfare. Although the study encompasses all five regions, agricultural activities are less significant in the Maritime region, which is more urbanized, resulting in a smaller sample size from this area, as shown in Figure 3.

The map shows the five administrative regions of Togo and its geographical location within West Africa. Source: [34]

The distribution of the final sample across Togo’s five administrative regions illustrates the diverse landscape of agricultural households within the country. As shown in Figure 4, the sample comprises 342 households from the Maritime region, 692 from Plateaux, 515 from Centrale, 682 from Kara, and 662 from Savannah.

4.3. Data Analysis

The socioeconomic characteristics of agricultural households were analyzed using descriptive statistics, including means, percentages, and tables. The poverty status of the different agricultural households was measured using the Foster–Greer–Thorbecke (FGT) poverty Index. Finally, the relationship between household poverty status and its drivers was assessed using a survey-weighted logistic regression model that accounts for the complex survey design, including sampling weights, primary sampling units (PSU), and stratification.

4.3.1. Estimation of the Foster–Greer–Thorbecke (FGTI) Poverty Index

To measure the monetary poverty status of agricultural households, the study employed the FGTI. [47] used this index to assess the poverty condition of smallholder agricultural households. This approach incorporates both the Headcount Ratio and the measures of the poverty gap, and it further allows poverty rates to be decomposed across different population subgroups. This makes the FGT index one of the most widely applied measures in poverty analysis [3,39,48]. In this study, household poverty was proxied by total household expenditure, from which the per capita expenditure has been derived by dividing the total household expenditure by the household size. The poverty line was defined according to the national poverty threshold for Togo in 2021/2022 in the database set at 295,182.06 FCFA per person per year.

The general formulation of the FGT index is as follows:

$$P\alpha ={\displaystyle \frac{1}{N}}\sum _{i=1}^n{\left({\displaystyle \frac{z-y_i}{z}}\right)}^{\alpha }$$

(1)

where: α = Foster, Greer and Thorbecke index (0 ≤ p ≥ 1); N = total number of households; p = number of households below the poverty line; z = poverty line (295,182.06 FCFA); $y_i$ = per capita expenditure of household i. The parameter α takes values of 0, 1, and 2 with different implications:

• P₀ (α = 0) measures poverty incidence (headcount ratio).
• P₁ (α = 1) measures the poverty gap, i.e., the average shortfall of the poor from the poverty line, representing the number of resources required for poor households to escape poverty.
• P₂ (α = 2) measures the severity of poverty, giving greater weight to the poorest households. Poverty levels increase with the FGT index’s proximity to 1.

In addition, the Headcount Ratio (HR) is often presented separately, as the proportion of the population living below the poverty line, and it is generally expressed as follows:

$$HR={\displaystyle \frac{q}{N}}$$

(2)

where q is the number of households below the poverty line, and N is the total number of households. A higher HR value, approaching 1, indicates a higher proportion of poor households in the population.

4.3.2. Estimation of the Multidimensional Poverty Index (MPI)

Multidimensional poverty is measured using the Alkire–Foster (AF) approach, suited to capturing non-monetary deprivations and enabling decomposition by subgroups and by indicators. Consistent with the national methodology of INSEED (Togo), the study adopts three dimensions: education, health, and living conditions, equally weighted (1/3 each), with equal weighting of indicators within each dimension (

Table 1. Dimensions, indicators, and weights used to construct the Multidimensional Poverty Index (MPI).

Dimension	Indicator (Deprivation Rule)	Dimension Weight	Indicator Weight
Health	Medical consultation: at least one household member fell ill and could not consult a qualified health professional	1/3	1/9
	Access to healthcare: the household is more than 5 km from the nearest health centre	1/3	1/9
	Chronic disease: at least one household member has a chronic disease	1/3	1/9
Education	Years of schooling: at least one household member has not completed five years of primary education	1/3	1/9
	School attendance of children: at least one school-age child in the household is not attending school	1/3	1/9
	Adult literacy: at least one member aged 15+ cannot read and write in French and in a local language (illiterate in both)	1/3	1/9
Living conditions	Access to electricity: the household has no access to electricity	1/3	1/24
	Drinking water: the household does not use potable/improved sources for drinking water	1/3	1/24
	Sanitation (toilets): household does not use improved toilets, e.g., internal/external WC with flush or manual flush, or latrines	1/3	1/24
	Wall material: main dwelling walls are not made of durable materials (mud, bamboo, or other temporary/low-resistance materials)	1/3	1/24
	Roof material: the main dwelling roof is not made of durable materials (palm leaves, plastic sheeting, or other temporary materials)	1/3	1/24
	Floor material: the main dwelling floor is not made of durable materials (earth, sand, dung, or other unimproved natural surfaces)	1/3	1/24
	Financial inclusion: no household member holds either a physical bank account or a digital account	1/3	1/24
	Ownership of durable assets: household owns fewer than two of the following: television, radio, car, motorcycle, bicycle, refrigerator	1/3	1/24

). The equal weighting follows the INSEED national methodology and the global MPI standard, which assumes that no single dimension of well-being is inherently more important than another. This approach ensures comparability across studies and avoids introducing subjective judgments into the aggregation process. Alternative weighting schemes were not considered, as the objective was to align with the national poverty measurement framework and maintain methodological coherence with [49]. Deprivation thresholds follow the national definition, and indicators are coded as binary (1 = deprived, 0 = not deprived). A household is identified as multidimensionally poor when its deprivation score is at least k = 1/3, meaning that the household is deprived in at least one-third of the weighted indicators. This cutoff reflects the conventional multidimensional poverty threshold used in both global and national applications of the AF method, indicating that a household experiences simultaneous deprivations across multiple dimensions of well-being rather than in a single aspect. Results report the incidence (H), intensity (A), and the MPI (M₀ = H × A), as well as the contribution of the dimensions (and, in Appendix A, of individual indicators). Estimates account for the survey design (weights, strata, clusters).

4.3.3. Measurement of Absolute Poverty

The absolute poverty measure used in this study follows the definition adopted by Togo’s national statistical institute [49]. It is defined as the intersection between monetary poverty and multidimensional poverty. A household is classified as absolutely poor if and only if it is simultaneously poor according to both measures (FGTI = 1 and MPI = 1). Formally, this can be expressed as follows:

$${ \mathrm{Absolute} \mathrm{Poverty}}_i=\left\{\begin{array}{c}1, \mathrm{i}\mathrm{f} FG{T}_{poor, i}=1 \mathrm{a}\mathrm{n}\mathrm{d} MP{I}_{\mathrm{poor}i}=1\\ 0, \mathrm{o}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{w}\mathrm{i}\mathrm{s}\mathrm{e}. \hfill \end{array}\right.$$

(3)

This composite definition captures households that experience deprivation across both income-based and non-income dimensions of poverty, thus representing the most vulnerable group.

4.3.4. Logistic Regression Model

To examine the drivers of agricultural household poverty, the study employed a binary logistic regression model that accounts for the complex survey design. The dependent variable equals 1 if the household is classified as poor and 0 otherwise [50]. Logistic regression is appropriate in this context because it models the probability of a binary outcome as a function of explanatory variables [51]

The general specification of the logit model is as follows:

$$Logit(P_i)=ln\left({\displaystyle \frac{P(Y_i=1\mid X_i)}{1-P(Y_i=1\mid X_i)}}\right)={\beta }_0+\sum _{i=1}^n{\beta }_i{X}_i$$

(4)

where Y_i denotes the poverty status of household_i, X_ji the explanatory variables, and β_j, the parameters to be estimated. The full specification includes the variables grouped into socio-demographic, household decision-making, geographical, and institutional categories, as detailed in

Table 2. Descriptive statistics of the variables used in the analysis.

Variables	Measurement	Mean	Std. Dev.	Percentage (%)
	Dependent variables
Poverty status FGT index (poor)	1 = Poor and 0 = non-poor			51.50
Multidimensional poverty status	1 = Poor and 0 = non-poor			59.4
Absolute poverty	1 = Poor and 0 = non-poor			35.36
	Independent variables
Age of the HH (year)	Continuous	47.28	13.99
Household size	Continuous	5.05	2.61
Education level of HH (0)	0 = Non educated; 1 = Primary school; 2 = Secondary School; 3 = High School; 4 = University;			44.80
Gender of HH	Dummy 1 if the household head is male, 0 if female			82.23
Marital status	Dummy 1 if the household’s head is married, 0 otherwise			76.32
Region
Maritime region	Dummy 1 if yes, 0 otherwise			11.82
Plateau region	Dummy 1 if yes, 0 otherwise			23.92
Central Region	Dummy 1 if yes, 0 otherwise			17.80
Kara region	Dummy 1 if yes, 0 otherwise			23.57
Savannah region	Dummy 1 if yes, 0 otherwise			22.88
Residence area	Dummy 1 if the household is in the rural area, 0 otherwise			90.91
Land ownership	Dummy 1 = Yes and 0 = No			63.33
Farm size (Hectare)	Continuous	3.89	9.87
Crop diversification	Dummy 1 = Yes and 0 = No			75.70
Participation in Off-farm activities	Dummy 1 = Yes and 0 = No			19.88
Irrigation use	Dummy 1 = Yes and 0 = No			1.73
Use of modern ploughing equipment	Dummy 1 = modern equipment and 0 otherwise			10.92
Membership in agricultural cooperatives	Dummy 1 = Yes and 0 = No			41.08
Tropical Livestock Units (TLU)	Continuous	1.48	5.79
Access to credit	Dummy 1= yes and 0 otherwise			9.37
Constraint to access the market	Dummy 1 = Yes and 0 = No			13.41
Economic idiosyncratic shock	Dummy 1 if the household experienced an economic shock during the past 12 months, 0 otherwise			9.30
Natural covariant shock	Dummy 1 if the household experienced a natural shock during the past 12 months, 0 otherwise			38.68
Farm income category	1 = Low 2 = Middle 3 = High			33.36

HH means head of household. Source: Authors’ conception. Farm income was categorized into low, medium, and high based on quintiles of the distribution. Specifically, Category 1 includes the bottom 20%, Category 2 the middle 60%, and Category 3 the top 20% of farm income among agricultural households.

.

Because the dataset was obtained from a complex survey design (multi-stage cluster sampling with stratification and unequal selection probabilities), the analysis was conducted using a survey-weighted logistic regression model. This approach adjusts for sampling weights, primary sampling units (PSU), and strata to ensure that the estimates are representative of the target population and that the standard errors are valid [52,53,54].

As described in Equation (3), the logistic model includes the term $P\left(Y_i=1∣X_i\right)=\pi \left(X_i\right)$, which represents the conditional probability that Y is equal to 1 given $X_i$. In the weighted logistic regression, survey weights $w_i$ are incorporated into the likelihood function to adjust for the unequal probabilities of selection. Then, the likelihood function for the weighted logit regression can be expressed as follows:

$$l\left(X,W,\beta \right)={\prod }_{i=1}^n {\left[\pi {\left(X_i\right)}^{y_i}{\left(1-\pi \left(X_i\right)\right)}^{1-y_i}\right]}^{w_i}$$

(5)

where W_i is the sampling weight of household i, the corresponding log-likelihood function, maximized in estimation, is as follows:

$$\mathcal{l}\left(\beta \right)=\sum _{i=1}^n w_i\left[y_i\ln \left(\pi \left(X_i\right)\right)+\left(1-y_i\right)\ln \left(1-\pi \left(X_i\right)\right)\right]$$

(6)

This pseudo-maximum likelihood estimation ensures consistency and unbiasedness under complex survey sampling.

Since the logit coefficients are not directly interpretable in terms of probabilities, marginal effects were also computed. For binary variables, the marginal effect is expressed as follows:

$$M{E}_j=\pi \left(X\mid D_j=1\right)-\pi \left(X\mid D_j=0\right)$$

(7)

This allows direct interpretation of results in terms of probability changes rather than log-odds or odds ratios.

The Variance Inflation Factor (VIF) was calculated to check for multicollinearity among the explanatory variables in the logit model, and the results are presented in Appendix A,

Table A2. Variance Inflation Factors (VIF) for multicollinearity diagnostics.

Variables	VIF	1/VIF
Gender of HH	1.8	0.554359
Marital status of HH	1.74	0.574167
Region	1.5	0.668619
Education level of HH	1.4	0.712046
Membership in agricultural cooperatives	1.29	0.772273
Household size	1.29	0.777116
Use of modern ploughing equipment	1.25	0.799699
Crop diversification	1.22	0.818073
Agricultural income category	1.2	0.83139
Age of the HH	1.19	0.838588
Participation in off-farm activities	1.18	0.845458
Land ownership	1.15	0.86705
Residence area	1.12	0.895809
Tropical Livestock Units (TLU)	1.09	0.919386
Constraint to access the market	1.07	0.938863
Farm size	1.05	0.952593
Natural covariant shock	1.03	0.967409
Access to credit	1.03	0.969631
Economic idiosyncratic shock	1.03	0.970649
Irrigation use	1.02	0.982171
Mean VIF	1.23

Source: Authors’ computation based on EHCVM 2021/22 data.

. In addition, a survey-weighted probit model was estimated as a robustness test to verify the consistency of the main logit results and presented in Appendix A,

Table A3. Probit regression: Drivers of poverty among agricultural households.

Variables	MPI			FGT Index			Absolute Poverty
	Margins	std. err.	p > t	Margins	std. err.	p > t	Margins	std. err.	p > t
Age of the HH	0.000	0.001	0.562	−0.002	0.001	0.006 **	−0.001	0.001	0.254
Household size	0.041	0.005	0.000 ***	0.075	0.005	0.000 ***	0.063	0.004	0.000 ***
Education level of HH	−0.106	0.011	0.000 ***	−0.058	0.012	0.000 ***	−0.064	0.010	0.000 ***
Gender of HH	0.002	0.031	0.936	−0.040	0.033	0.224	−0.036	0.029	0.215
Marital status of HH	−0.034	0.031	0.284	−0.013	0.031	0.677	−0.038	0.031	0.213
Land ownership	−0.032	0.023	0.161	0.003	0.025	0.906	−0.022	0.021	0.306
Farm size	−0.001	0.001	0.321	−0.004	0.001	0.003 **	−0.003	0.001	0.027 *
Crop diversification	−0.019	0.027	0.496	−0.025	0.027	0.365	−0.041	0.024	0.091 *
Participation in off-farm activities	−0.053	0.027	0.053 *	−0.100	0.026	0.000 ***	−0.103	0.025	0.000 ***
Irrigation use	−0.080	0.070	0.255	0.032	0.077	0.683	−0.011	0.075	0.881
Use of modern ploughing equipment	0.019	0.038	0.617	−0.094	0.052	0.070 *	−0.036	0.037	0.330
Tropical Livestock Units (TLU)	0.006	0.003	0.063 *	−0.010	0.002	0.000 ***	−0.006	0.002	0.002 **
Agricultural income category	0.000	0.013	0.990	−0.060	0.014	0.000 ***	−0.039	0.012	0.002 **
Membership in agricultural cooperatives	0.042	0.021	0.050 *	−0.033	0.025	0.183	0.009	0.023	0.693
Access to credit	0.044	0.036	0.217	−0.061	0.038	0.110	−0.029	0.035	0.403
Constraint to access the market	0.138	0.034	0.000 ***	−0.042	0.031	0.171	0.029	0.027	0.288
Economic idiosyncratic shock	−0.002	0.033	0.951	−0.183	0.034	0.000 ***	−0.105	0.034	0.002 **
Natural covariant shock	0.028	0.024	0.245	−0.076	0.021	0.000 ***	−0.039	0.020	0.055 *
Region	0.022	0.009	0.015 *	0.026	0.012	0.023 *	0.031	0.009	0.000 ***
Residence area	0.247	0.036	0.000 ***	0.069	0.042	0.101	0.196	0.041	0.000 ***
_cons	−0.805	0.225	0.000 ***	−0.1437234	0.2390836	0.548	−1.289514	0.250471	0.000 ***
Number of obs.			2893			2893			2893
F(20,384)			17.42			14.66			14.81
Prob>F			0.0000			0.0000			0.0000

Note: std. err. = standard error; p > t = p-value. ***, **, and * indicate the significance of the coefficient at 1%, 5%, and 10%, respectively. HH means head of household. Source: Authors’ conception.

.

5. Results

5.1. Descriptive Analysis

Table 2 reports the descriptive statistics of the variables used in the analysis. The results show that 51.5% of agricultural households are poor according to the monetary measure, while 59.4% are deprived under the multidimensional poverty index. Additionally, 35.36% of households are classified as absolutely poor, experiencing both monetary and multidimensional deprivation. The average age of household heads is 47 years, and the mean household size is approximately five members. A total of 44.8% of heads have no formal education, 82.2% are male, and 76.3% are married. Regarding productive resources, 63.3% of households own land, with an average farm size of 3.89 hectares, and 75.7% engage in crop diversification. Only 19.9% is engaged in off-farm activities. Only 1.7% of households use irrigation, while 10.9% use modern ploughing equipment. Cooperative membership is reported for 41.08% of households, and average livestock ownership is 1.48 TLU. Access to credit is limited (9.4%), and 13.4% of households report market access difficulties. 38.7% of households experienced natural shocks, and 9.3% economic shocks. Most households (90.9%) reside in rural areas, primarily in the Plateau (23.9%), Kara (23.6%), and Savannah (22.9%) regions.

These descriptive patterns highlight the structural characteristics of Togo’s smallholder agriculture, where production remains male-dominated, low-mechanized, and vulnerable to shocks. The high share of male-headed households (82.2%) and the extremely low adoption of irrigation (1.7%) reflect the broader conditions of traditional smallholder systems in West Africa, where limited capital, inadequate infrastructure, and dependence on rainfed farming constrain productivity. Likewise, the low rate of off-farm participation (19.9%) indicates restricted opportunities for income diversification and high dependence on agricultural income, which may amplify vulnerability to economic and climatic fluctuations.

5.2. FGT Index Results

Table 3. Foster–Greer–Thorbecke (FGT) poverty index for agricultural households.

Index	Index Values
Poverty incidence (P0)	0.5150
Poverty gap (P1)	0.3187
Poverty severity (P2)	0.1360

Source: Authors’ conception.

reports the FGT poverty indices for agricultural households. The headcount ratio (P₀) shows that 51.5% of households are below the poverty line. The poverty gap (P₁) is 0.32, and the severity index (P₂) is 0.14, indicating the depth and distribution of poverty among poor households. The binary poverty status derived from FGT0 is used as the dependent variable in the subsequent econometric analysis. The poverty incidence of 51.5% among farm households significantly exceeds the national rate of 43.8% reported by INSEED in 2021/2022, confirming that farm livelihoods are disproportionately deprived.

Table 3 reveals a high level of monetary poverty among agricultural households, with the majority falling below the poverty threshold. This outcome reflects several underlying drivers that shape household expenditure patterns, since the monetary measure is derived from daily consumption levels. However, income-based indicators alone do not capture the full extent of deprivation. To provide a more comprehensive understanding, the next section examines the multidimensional poverty status of these agricultural households.

5.3. Multidimensional Poverty Index Results

Using the Alkire–Foster approach (k = 1/3), the incidence of multidimensional poverty (H) is 0.594, and the average deprivation intensity (A) is 0.477, which means that households classified as multidimensionally poor experience almost half of the weighted deprivations. Roughly six to seven out of the fourteen considered indicators. Yielding an MPI of M₀ = 0.283. The dimensional decomposition indicates that education contributes most to M₀ (38.0%), followed by living conditions (34.6%) and health (27.4%), highlighting the central role of educational deprivations. All estimates and the decomposition are reported in

Table 4. Multidimensional Poverty Index (MPI): Incidence, intensity, and overall index.

Measure	Estimation
Incidence (H)	0.594
Intensity (A)	0.477
MPI = H × A	0.283

Source: Authors’ conception.

.

This illustrates the depth of poverty beyond income measures, highlighting the multiple and overlapping disadvantages faced by rural households.

Regarding the contribution of each dimension, Figure 5 illustrates the percentage share of the different indicators. Multidimensional poverty between agricultural households in Togo is predominantly influenced by educational factors, with adult literacy (20.1%), years of schooling (10.2%), and school attendance (7.7%) being the most significant contributors. Health issues follow closely, primarily due to inadequate access to medical consultations, which accounts for 16.3% of the poverty dimensions. While living conditions have a lesser impact, they remain important, particularly through financial exclusion (8.7%), sanitation (8.4%), and lack of electricity (7.2%). Deprivations related to housing materials and asset ownership play a relatively minor role in explaining the overall poverty level. The predominance of educational deprivation (combined contribution of 38%) suggests that human capital deficits are the main obstacle to multidimensional welfare, highlighting the need for interventions focused on education, in addition to income support.

5.4. Drivers of Poverty Among Agricultural Households in Togo

Table 4 presents the marginal effects from the weighted logistic regressions of poverty status among agricultural households in Togo. Results are reported across the three poverty measures: the MPI, FGTI, and absolute poverty. This threefold approach enables us to disentangle drivers that are robust across measures from those that are specific to a single dimension, thereby underscoring the multifaceted nature of poverty among agricultural households. The results reveal three categories of drivers: (i) those significant across both MPI and FGTI, suggesting structural drivers of poverty; (ii) those significant only under MPI, reflecting deprivations that extend beyond income; and (iii) those significant only under FGTI, capturing primarily monetary constraints. Together, these categories highlight how different sets of drivers shape poverty outcomes and reinforce the need to move beyond a one-dimensional measurement of welfare.

• Socio-demographic drivers

Household size is positively associated with all three poverty measures (1% level). Meaning each additional household member increases the probability of multidimensional poverty by 4.2%, monetary poverty by 7.9%, and absolute poverty by 6.4%. The education level of the household head is negatively associated with all poverty outcomes at 1% level. Each additional level of schooling of the head of household reduces multidimensional poverty by 10.6%, monetary poverty by 5.9%, and absolute poverty by 6.4%.

Moreover, the age of the household head is statistically significant only for monetary poverty. Each additional year of age reduces the probability of being income-poor by 0.2%. No significant marginal effect is observed for multidimensional or absolute poverty. Gender and marital status show no significant associations across the three poverty measures.

• Household decision-making drivers

Participation in off-farm activities correlates with a 5.4% reduction in the likelihood of multidimensional poverty (10% significance), a 9.9% decrease in monetary poverty (1% significance), and a substantial 10.4% lower chance of absolute poverty (1% significance). Conversely, livestock ownership (measured in Tropical Livestock Units) presents mixed results; it is linked to a 0.6% increase in multidimensional poverty (10% significance), while simultaneously reducing the probability of monetary poverty by 1.0% (1% significance) and absolute poverty by 0.5% (5% significance). Crop diversification does not significantly impact multidimensional or monetary poverty but is associated with a notable 4.1% decrease in absolute poverty at the 10% significance level. Additionally, farm size is related to a 1% reduction in monetary poverty and a 0.9% decrease in absolute poverty (5% significance), with no significant effect on the Multidimensional Poverty Index. Lastly, membership in agricultural cooperatives is associated with a 4.5% increase in multidimensional poverty (10% significance), with no significant effects on monetary or absolute poverty.

• Institutional and contextual drivers

The results presented in

Table 5. Drivers of monetary, multidimensional, and absolute poverty among agricultural households.

Variables	Multidimensional Poverty			Monetary Poverty			Absolute Poverty
	Margins	std. err.	p > t	Margins	std. err.	p > t	Margins	std. err.	p > t
Age of the HH	0.000	0.001	0.592	−0.002	0.001	0.006 **	−0.001	0.001	0.362
Household size	0.042	0.005	0.000 ***	0.079	0.005	0.000 ***	0.064	0.004	0.000 ***
Education level of HH	−0.106	0.011	0.000 ***	−0.059	0.012	0.000 ***	−0.064	0.010	0.000 ***
Gender of HH	0.000	0.031	0.989	−0.039	0.033	0.236	−0.039	0.030	0.194
Marital status of HH	−0.029	0.032	0.363	−0.016	0.030	0.594	−0.035	0.032	0.269
Land ownership	−0.037	0.023	0.112	0.008	0.025	0.759	−0.021	0.021	0.327
Farm size	−0.001	0.002	0.634	−0.010	0.003	0.001 **	−0.009	0.003	0.002 **
Crop diversification	−0.021	0.028	0.455	−0.021	0.027	0.424	−0.041	0.024	0.087 *
Participation in off-farm activities	−0.054	0.027	0.047 *	−0.099	0.026	0.000 ***	−0.104	0.025	0.000 ***
Irrigation use	−0.086	0.073	0.242	0.015	0.075	0.84	−0.026	0.077	0.736
Use of modern ploughing equipment	0.021	0.038	0.592	−0.084	0.053	0.111	−0.027	0.037	0.472
Membership in agricultural cooperatives	0.045	0.021	0.037 *	−0.030	0.024	0.215	0.013	0.023	0.578
Tropical Livestock Units (TLU)	0.006	0.003	0.098 *	−0.010	0.002	0.000 ***	−0.005	0.002	0.002 **
Agricultural income category	0.001	0.013	0.917	−0.057	0.014	0.000 ***	−0.035	0.012	0.005 **
Access to credit	0.039	0.037	0.291	−0.055	0.038	0.152	−0.022	0.035	0.529
Constraint to access the market	0.137	0.034	0.000 ***	−0.036	0.031	0.238	0.039	0.028	0.165
Economic idiosyncratic shock	−0.001	0.033	0.984	−0.187	0.035	0.000 ***	−0.106	0.035	0.003 **
Natural covariant shock	0.032	0.024	0.186	−0.074	0.021	0.001 **	−0.036	0.020	0.073 *
Region	0.022	0.009	0.016 *	0.021	0.012	0.069 *	0.028	0.009	0.002 **
Residence area	0.243	0.036	0.000 ***	0.073	0.043	0.091 *	0.211	0.045	0.000 ***
_cons	−1.371	0.383	0.000	−0.242	0.401	0.546	−2.257	0.444	0.000
Number of obs.			2893			2893			2893
F (20,384)			15.97			13.03			12.45
Prob > F			0.0000			0.0000			0.0000

Note: std. err. = standard error; p > t = p-value. ***, **, and * indicate the significance of the coefficient at 1%, 5%, and 10%, respectively. HH means head of household. Source: Authors’ conception.

indicate a strong correlation between poverty levels in Togolese agricultural households and drivers such as market constraints, exposure to shocks, regional disparities, and residence area. Constraints in market access are associated with a 13.7% increase in the probability of multidimensional poverty at a 1% significance level, while showing no significant effect on monetary or absolute poverty. Additionally, the findings suggest that exposure to economic shocks correlates with an 18.7% decrease in the probability of monetary poverty at the 1% significance level and a 10.6% decrease in absolute poverty at the 5% significance level, with no significant relationship identified for multidimensional poverty. Similarly, natural shocks are linked to a 7.4% reduction in monetary poverty and a 3.6% reduction in absolute poverty, with the latter being significant at the 10% level, while their impact on multidimensional poverty remains insignificant, indicating that households experiencing these shocks are statistically less likely to experience income or combined poverty in the short term.

The Variance Inflation Factor (VIF) was calculated to check for multicollinearity among the explanatory variables in the logit model, and the results are presented in Appendix A, Table A2.

Geographical Drivers

Turning to spatial disparities, the results in Table 5 reveal a strong and significant rural–urban divide. Rural households are 24.3% more likely to experience multidimensional poverty (1% significance). For monetary poverty, rural residence increases the probability of being income-poor by 7.3% (10% significance). The marginal effect of 21.1% in the absolute poverty model (1% significance) confirms their substantially higher vulnerability relative to urban households. Regional disparities also emerge as an important determinant of household welfare. Households located in disadvantaged regions are 2.2% more likely to experience multidimensional poverty (significant at the 5% level). For monetary poverty, the regional effect is 2.1% (significant at the 10%). Finally, regional residence increases the probability of absolute poverty by 2.8% (significant at the 1% level, p < 0.01), confirming that households living in structurally disadvantaged regions face systematically higher poverty risks across all dimensions.

To visualize these spatial disparities, Figure 6 presents region-specific poverty probabilities based on margins estimates for the monetary (FGT0), multidimensional poverty (MPI-H), and absolute poverty measures. The results reveal substantial contrasts across regions. The Central region shows the lowest predicted poverty levels across all three indicators, while the Savannah region consistently exhibits the highest incidence. Maritime and Plateau fall in intermediate positions, whereas Kara displays elevated poverty rates similar to Savannah, particularly under the MPI-H and absolute poverty measures.

6. Discussion

National statistics indicate [49] a slight reduction in monetary poverty in Togo, decreasing from 45.5% in 2018/19 to 43.8% in 2021/22. However, rural areas continue to experience high levels of deprivation, with poverty remaining nearly unchanged at around 58%, while urban poverty declined from 26.5% to 24.6% over the same period. The disparity is even more pronounced for multidimensional poverty: the national incidence is 28.9%, but it rises to 45.0% among rural households. Against this backdrop, agricultural households in the present study show substantially higher deprivation, with 51.5% living below the monetary poverty line and 59.4% classified as multidimensionally poor (MPI = 0.283). These figures highlight the structural vulnerability of agricultural households, whose poverty levels exceed both national and rural averages. While these patterns reflect correlations rather than causal effects due to the cross-sectional design of the data, they nevertheless reveal persistent structural inequalities affecting agricultural communities. Consequently, agricultural household poverty in Togo is shaped by a combination of structural conditions and household-specific drivers.

Drivers of Poverty Among Agricultural Households

• Socio-demographic drivers

These results are consistent with the evidence presented in the reviewed literature. Although the drivers were not grouped in the same way across previous studies, they consistently appear as variables influencing poverty levels, either positively or negatively, among agricultural or rural households in different contexts.

Household size plays a crucial role in increasing the risk of poverty, as larger families often encounter higher dependency ratios and greater consumption demands. This situation reduces per capita resources and limits the household’s capacity to invest in education, nutrition, and essential living conditions, leading to heightened monetary and multidimensional deprivation. The descriptive statistics reveal that the average household in the sample consists of around five members, suggesting that many agricultural families operate with relatively large units under constrained resources. In this study, the marginal effects indicate that each additional household member significantly raises the likelihood of experiencing multidimensional (4.2%), monetary (7.9%), and absolute poverty (6.4%). These moderate but persistent effects align with earlier research highlighting the economic burden associated with larger households [55,56,57,58,59]. While some studies propose that larger families might gain from enhanced labour availability and income diversification opportunities [60,61], these benefits seem inadequate in the Togolese context for this sample size. The reliance on subsistence agriculture, limited technological access, and widespread structural barriers restrict the potential contributions of additional household members to productive endeavours, thereby maintaining high dependency burdens as a significant factor driving both monetary and multidimensional poverty among agricultural households.

The educational level of the household head is linked to a reduced risk of experiencing monetary, multidimensional, and absolute poverty. This correlation likely stems from the fact that more educated household heads have better access to information, enhanced capacity to implement agricultural innovations, and increased opportunities for diversifying income sources. Previous research has highlighted these mechanisms as significant factors in poverty alleviation. For instance, [62,63,64] highlight its role in expanding livelihood opportunities and improving households’ ability to escape poverty. However, education alone may not be enough; its impact depends on complementary factors such as access to credit, market participation, and extension services that help translate knowledge into productive outcomes. The descriptive statistics show that 44.8% of household heads have no formal education. This significant educational gap highlights the urgent need for investment in rural education and adult literacy programmes. By strengthening educational foundations, households could improve their ability to adopt better practices, engage in more lucrative activities, and ultimately escape the cycle of poverty, especially if these initiatives are complemented by broader institutional and market support.

The age of the household head appears to have a nuanced association with poverty. Descriptive statistics reveal that the average age of household heads is 47 years, indicating that many agricultural households are managed by individuals with considerable life and farming experience. This experience likely enhances their decision-making abilities, bolsters social networks, and improves agricultural risk management, which helps to explain the negative correlation between age and monetary poverty observed in the econometric analysis. This interpretation aligns with previous studies suggesting that age is often associated with better adaptive and managerial capacities [65,66]. However, the literature also indicates that the benefits of age may wane after a certain point, as increased age can lead to diminished physical productivity and heightened vulnerability to economic shocks [67,68]. This complexity may clarify why age does not show a significant relationship with multidimensional or absolute poverty in this study; while experience may foster income stability, it does not necessarily lead to enhancements in education, health, or overall living conditions for the household.

Overall, the socio-demographic findings indicate that household size and education are critical drivers influencing poverty, primarily through their effects on dependency ratios, human capital development, and labour productivity. These insights emphasize the necessity of focusing on Sustainable Development Goal 1 (No Poverty) and Sustainable Development Goal 4 (Quality Education). Addressing educational inequalities and reducing household vulnerabilities are vital steps toward enhancing the welfare of agricultural households in Togo.

• Household decision-making drivers

The regression analysis indicates a consistent link between participation in off-farm activities and reductions in monetary, multidimensional, and absolute poverty, highlighting the crucial role of off-farm income in enhancing household welfare. This relationship can be attributed to several drivers. Off-farm employment offers a more stable income stream, which lessens reliance on rainfed agriculture and helps households manage consumption during times of climatic or market uncertainty. Additionally, it alleviates liquidity constraints, allowing families to invest in productive resources, education, health, and improved living conditions, thereby directly contributing to a decrease in multidimensional poverty. Furthermore, reallocating labour from low-yield subsistence farming to off-farm activities boost overall labour productivity and mitigates agricultural risks. However, despite these benefits, only 19.9 percent of households engage in off-farm activities, indicating that the majority of agricultural households are excluded from this important poverty-reducing pathway. Such limited participation may reflect structural barriers, including low educational attainment, limited labour mobility, and the scarcity of local employment opportunities. This finding is consistent with [69,70,71], who argue that off-farm participation not only diversifies income streams but also reduces households’ exposure to climatic and market shocks.

Membership in agricultural cooperatives is associated with a 4.5% increase in multidimensional poverty (p < 0.10) and shows no significant effect on monetary or absolute poverty. This counterintuitive result contrasts with most empirical studies, which generally find that cooperatives improve access to inputs, markets, and services [72,73,74,75]. Several mechanisms may explain this pattern. One possibility is adverse selection, whereby more vulnerable households join cooperatives seeking support, meaning that the association reflects pre-existing poverty rather than cooperative impact. Another explanation relates to internal governance challenges, including uneven participation or limited managerial capacity, which can reduce the effectiveness of service delivery for poorer members. Additionally, the financial and time costs of membership may disproportionately burden low-resource households, limiting their ability to benefit from cooperative activities. Similar findings in [76] highlight that only well-functioning cooperatives with strong service capacity are able to reduce multidimensional poverty.

The results indicate that farm size remains an important protective factor against poverty. Larger landholdings enable households to expand production, diversify crops, and generate marketable surpluses, thereby improving income stability and reducing vulnerability. They also strengthen financial security by serving as collateral for credit, facilitating investments in productivity-enhancing inputs. Although landownership is common among households (63.3%), the average cultivated area of 3.89 hectares masks considerable variation in both size and quality, which helps explain why the poverty-reducing effect, while present, remains modest. This finding is in line with previous studies showing that larger farms improve production capacity, market participation, and liquidity [77,78]. However, structural constraints such as unequal land distribution, variable soil fertility, and limited access to complementary inputs like labour, irrigation, and technology, may restrict the extent to which additional land translates into substantial welfare gains.

Investment in livestock (TLU) exhibits varying impacts on different poverty metrics. Specifically, an increase of one Tropical Livestock Unit correlates with a 0.6% rise in multidimensional poverty, while simultaneously contributing to a 1% decrease in monetary poverty and a 0.5% reduction in absolute poverty. This discrepancy indicates that livestock primarily serves as a financial buffer, enabling households to manage consumption during economic shocks or temporary income declines. Although this short-term financial relief alleviates income-based poverty, it does not necessarily lead to enhancements in health, education, or living standards. Often, the proceeds from livestock sales are directed towards immediate needs rather than long-term investments in human capital, which may clarify why multidimensional poverty remains unchanged or even increases despite improvements in monetary welfare. This aligns with evidence from [10,79], who show that livestock strengthens rural resilience, stabilizes income, and reduces poverty severity by providing a flexible asset that households can mobilize during shocks. At the same time, findings from [80] emphasize that these benefits are highly uneven, as many rural households own too few animals to reach viability thresholds, meaning that livestock ownership does not automatically improve multidimensional well-being and can even coincide with persistent deprivation.

Crop diversification is marginally significant in the absolute poverty model, where it decreases the likelihood of poverty by 4.1%. This indicates that while diversification may not influence poverty outcomes when assessed through a single dimension, its beneficial role becomes apparent when poverty is defined more rigorously, requiring households to be poor across multiple dimensions simultaneously. This pattern aligns with literature showing that diversification improves resilience through food security, dietary diversity, and risk reduction rather than through immediate income gains [21,81,82,83,84]. Because these benefits tend to accumulate gradually rather than translate into short-term income or welfare improvements, their impact becomes visible only under a stricter poverty definition that captures simultaneous deprivations. This helps explain why its effect emerges in the absolute poverty measure but remains muted in the monetary and multidimensional indices taken separately.

Agricultural income serves as a crucial mechanism for alleviating poverty, as increased earnings from farming enhance household liquidity, enabling families to better meet their daily consumption needs and reducing their dependence on low-yield coping strategies like asset liquidation or informal loans. This rise in agricultural revenue not only helps stabilize food security but also facilitates the purchase of necessary inputs and allows for smoother consumption during periods of economic fluctuation, thereby decreasing the risk of both monetary and absolute poverty. Nevertheless, the benefits of agricultural income often remain confined to financial aspects, as improvements in earnings do not inherently lead to enhanced access to education, healthcare, housing quality, or essential services, areas that necessitate structural investments beyond individual households’ capabilities. This limitation highlights why agricultural income has a restricted impact on multidimensional poverty, which encompasses more profound and enduring forms of deprivation. These observations align with research indicating that while agricultural income can effectively diminish monetary poverty, lasting enhancements in overall welfare necessitate complementary public investments in rural infrastructure, health services, and educational systems [85,86].

Overall, the results show that key household decisions strongly shape welfare outcomes among agricultural households in Togo. Off-farm participation remains the most consistent poverty-reducing strategy, while larger farm sizes and livestock ownership provide important buffers that stabilize income and reduce vulnerability. Crop diversification also contributes to resilience, though its effects appear only under the stricter absolute poverty measure. In contrast, cooperative membership is associated with higher multidimensional poverty, pointing to organizational or targeting challenges that limit its expected benefits. These patterns are consistent with individualistic and cultural theories of poverty, which emphasize the role of household choices, asset management, and adaptive behaviour in determining welfare outcomes [41,42,43]. They also reinforce the relevance of Sustainable Development Goals 1 (No Poverty) and 8 (Decent Work and Economic Growth), underscoring the need to strengthen household capabilities, improve access to productive assets, and expand viable livelihood pathways for rural populations.

• Institutional and contextual drivers

Market access emerges as an important institutional driver for multidimensional deprivation. Although only about 13% of households report facing market constraints, this minority group shows a significantly higher likelihood of simultaneous deprivations in health, education, and living standards. While these constraints do not influence monetary or absolute poverty, their effect on multidimensional outcomes highlights the essential role of market integration in enabling households to access both economic opportunities and welfare-enhancing services. The underlying mechanism is straightforward: when market access is limited, households struggle to sell agricultural produce at fair or timely prices, reducing income stability and restricting their ability to finance essential expenditures such as schooling, healthcare, and housing improvements. Limited market access also constrains the purchase of key agricultural inputs such as seeds, fertilizers, and tools, lowering productivity and keeping farmers trapped in low-return subsistence production. This result corroborates [87], who contend that weak market integration perpetuates subsistence-oriented production and constrains the diffusion of welfare improvements. Improving rural infrastructure, particularly roads and communication networks, and reducing transaction costs are therefore critical to connecting smallholder agriculture with broader socioeconomic development.

The effect of shocks on poverty appears to be complex and context-dependent. In the present analysis, exposure to economic and natural shocks shows negative marginal associations with monetary and absolute poverty, while having no significant influence on multidimensional deprivation. This counterintuitive pattern suggests that households reporting shocks are not necessarily the most deprived. This pattern indicates that those who report shocks often possess enough assets or coping capacity to buffer their short-term welfare, whereas the poorest, who lack resources, savings, or consumption margins, may experience shocks without recognizing or reporting them. Notably, only about 9 percent of households report experiencing an economic shock, compared with nearly 39 percent reporting a natural shock, reflecting substantial variation in exposure and in households’ ability to perceive or declare such events. Relatively better-off households may be more capable of withstanding short-term disturbances through financial buffers, social networks, diversified income sources, or stronger market participation, allowing them to maintain consumption in the aftermath of shocks. Evidence from the literature supports this interpretation. The CGE model analysis by [88] demonstrated that food price increases caused welfare deterioration primarily among low-income workers and self-employed agricultural employees. The research demonstrates that cash transfer programmes, together with food subsidies, specifically the former type, function as policy-based resilience mechanisms to reduce adverse welfare impacts from price increases. The research by [89] demonstrates that rural Ethiopian households who experience recurring shocks experience worsening structural and stochastic poverty, but their ability to maintain poverty avoidance depends on their access to irrigation, level of literacy, vegetation cover, and non-farm income. The research supports the concept that resilience acts as a mediator between shocks and welfare because it enables households with better resources to withstand shocks without becoming poor. The research supports previous studies [90,91,92], which demonstrate that multiple sequential shocks lead to the depletion of coping mechanisms, resulting in worsening poverty levels. The study reveals that shock-exposed households experience brief advantages which depend on their capacity to maintain adaptive and policy support systems. Given these dynamics, the observed associations should be interpreted cautiously. Several methodological factors may contribute to the unexpected direction of the estimates: (1) reverse causality (richer households have more assets at risk); (2) recall bias (poorer households may underestimate shocks); and (3) short-term measurement (cross-sectional data cannot capture long-term poverty impacts). A longitudinal analysis would be needed to disentangle these paths.

Taken together, these results provide strong empirical support for the structural theory of poverty and Sen’s entitlement approach. They highlight that poverty reduction cannot rely solely on household-level initiatives; instead, it requires systemic reforms in markets, financial systems, and public service provision [93]. Sustainable poverty alleviation thus depends on transforming the institutional environment in which rural households operate, ensuring that opportunities and entitlements are equitably distributed. This aligns with Sustainable Development Goals 10 (Reduced Inequalities) and 16 (Peace, Justice and Strong Institutions), which emphasize the need for inclusive institutions, fair resource distribution, and structural change to support long-term development.

7. Conclusions

This study used survey-weighted logit models and nationally representative EHCVM (2021/22) data to identify the determinants of monetary, multidimensional, and absolute poverty among agricultural households in Togo. The results show that multidimensional poverty is significantly more widespread than income poverty, underscoring the breadth of deprivations experienced by agricultural households. Key drivers include socio-demographic drivers such as household size, education, and age; household decision-making elements such as off-farm participation, land size, livestock assets, and crop diversification; and structural drivers, especially market access constraints and regional disparities in the Savannah region.

The findings confirm all hypotheses. Education reduces poverty, while larger households face higher deprivation. Off-farm participation and larger farm size both lower poverty risks by strengthening income diversification and expanding production capacity. Households located in the northern regions, particularly the Savannah region, show significantly higher probabilities of monetary and multidimensional poverty. In contrast, better market access lowers the likelihood of poverty among agricultural households. These patterns indicate that poverty among agricultural households arises from combined demographic, behavioural, and institutional constraints, consistent with Sen’s capability framework.

The policy implications are clear. Reducing poverty among agricultural households requires strengthening rural education, improving market connectivity, and supporting the acquisition and effective use of productive assets. Investments in rural roads, land governance, livelihood diversification, and access to credit, together with financial literacy and well-functioning agricultural markets, can enhance household resilience and welfare. Region-specific interventions in the Savannah region remain essential to address persistent structural disadvantages.

However, the study is not without limitations. The cross-sectional design of the data restricts the ability to draw causal conclusions or to analyze changes in poverty status over time. Potential endogeneity and unobserved heterogeneity may influence the estimated relationships, as it is not possible to fully control for all omitted variables. In addition, measurement limitations, reporting biases, and unobserved differences in soil quality, infrastructure, and market functioning may also affect the accuracy of the estimates. Future research should examine poverty dynamics using panel or longitudinal data to capture how households move into and out of poverty. Further work could also investigate how climate variability interacts with household resilience to shape welfare outcomes, in order to better distinguish chronic from transient poverty. Integrating geospatial information on infrastructure, soil characteristics, market connectivity, and climate exposure would also improve understanding of the spatial disparities observed among agricultural households.