Targeting Rural Poverty: A Generalized Ordered Logit Model Analysis of Multidimensional Deprivation in Ethiopia’s Bilate River Basin

Abstract

Understanding the complex and multidimensional nature of poverty is essential for designing effective and targeted policy interventions in rural Ethiopia. This study examined the determinants of multidimensional poverty in Bilate River Basin in South Ethiopia, employing cross-sectional household survey data collected in 2024. A total of 359 households were selected using a multistage sampling technique, ensuring representation across agro-ecological and socio-economic zones. The analysis applied the Generalized Ordered Logit (GOLOGIT) model to categorize households into four mutually exclusive poverty statuses: non-poor, vulnerable, poor, and extremely poor. The results reveal that age, dependency ratio, education level, livestock and ox ownership, access to information and credit, health status, and grazing land access significantly influence poverty status. Higher dependency ratios and poor health substantially increase the likelihood of extreme poverty, while livestock ownership and access to grazing land reduce it. Notably, credit use and access to information typically considered poverty reducing were associated with increased extreme poverty risks, likely due to poor financial literacy and exposure to misinformation. These findings underscored the multidimensional and dynamic nature of poverty, driven by both structural and behavioral factors. Policy implications point to the importance of integrated interventions that promote education, health, financial literacy, and access to productive assets to ensure sustainable poverty reduction and improved rural livelihoods in Ethiopia.

1. Introduction

The interplay between poverty and environmental degradation has been widely debated in academic research. A prevailing perspective suggests that poverty exacerbates environmental degradation, as low-income communities rely heavily on natural resources for survival, leading to excessive exploitation and resource depletion a phenomenon known as the “poverty-environment nexus” (Barbier, 2010). However, an alternative viewpoint posits that wealthier households, through industrial activities and commercial agriculture, contribute more significantly to environmental depletion than impoverished communities (Sir Partha Dasgupta, 2021). These contrasting perspectives underscore the need for a comprehensive analysis of how poverty and environmental degradation interact within specific socio-economic and governance contexts.

In many developing nations, particularly those with agriculture-based economies, natural resources serve as the foundation of rural livelihoods (Barrett & Bevis, 2015; IFAD, 2016). Farmers depend on land, forests, and water sources for food production, energy, and income generation. However, environmental degradation manifested in deforestation, soil erosion, declining water quality, and loss of biodiversity poses a significant threat to agricultural productivity and food security. While existing studies highlight these concerns, they often assume a unidirectional link between poverty and environmental degradation, overlooking potential feedback loops and the role of governance structures (Jenkins et al., 2006; Ssekibaala & Kasule, 2023).

Empirical studies in Ethiopia have explored poverty–environment interactions, focusing on soil degradation, deforestation, and agricultural productivity (Gebrehiwot & Van Der Veen, 2013; Kassa et al., 2020). However, many studies rely predominantly on income-based poverty measures, failing to capture the multidimensional nature of poverty, which extends beyond financial constraints to include access to education, health services, and essential utilities (Goshu, 2019). A multidimensional approach provides a more holistic understanding of poverty–environment interactions, recognizing that factors such as inadequate sanitation and education levels significantly influence both resource use and conservation efforts (Ferraro et al., 2015; Goshu, 2019).

Furthermore, governance structures, institutional policies, and land tenure security play a crucial role in shaping poverty–environment dynamics (Gebrehiwot & Van Der Veen, 2013; Kassa et al., 2020). In Ethiopia, weak governance and unclear property rights discourage long-term investments in sustainable land management, perpetuating poverty and environmental degradation (Kassa et al., 2020). Conversely, policies that promote financial inclusion, encourage innovative agricultural practices, and secure land tenure rights can mitigate environmental degradation while improving economic well-being (Ferraro et al., 2015).

Despite previous efforts, few studies have rigorously modeled the ordered nature of multidimensional poverty categories while accounting for partial proportional odds violations a methodological limitation this study addressed by applying the Generalized Ordered Logit (GOLOGIT) model. The GOLOGIT model offers a flexible framework for analyzing ordinal dependent variables, such as poverty levels, while relaxing the restrictive parallel lines assumption that standard ordered models impose. This represents a novel methodological contribution in the Ethiopian context.

Therefore, this study aimed to explore the determinants of multidimensional poverty using the Generalized Ordered Logit model in the Bilate sub-basin of Southern Ethiopia. Specifically, it assesses poverty using multidimensional indicators and explores governance and institutional factors mediating this relationship. Unlike previous studies that focus solely on income levels, this research integrates economic, social, and environmental dimensions to offer a more comprehensive perspective. In line with this aim, the study specifically seeks to identify the socio-economic, demographic, and institutional factors influencing multidimensional poverty; analyze the spatial and environmental dimensions contributing to poverty in the study area; and assess the relevance and policy implications derived from the application of the Generalized Ordered Logit (GOLOGIT) model.

Based on these objectives, the study tests the following research hypothesis: “Socio-economic, demographic, environmental, and institutional factors significantly influence the likelihood of a household falling into different categories of multidimensional poverty in the Bilate sub-basin.”

By addressing theoretical and empirical gaps, this study contributes to existing knowledge in three key ways. First, it explores the socio-economic and environmental drivers of poverty and resource degradation; second, it employs a multidimensional poverty framework that captures deprivations beyond financial constraints; and, finally, it generates policy-relevant insights that could inform interventions aimed at breaking the cycle of poverty.

The findings from this study were expected to inform context-appropriate poverty alleviation strategies, enhance land and resource governance, and guide the design of multisectoral policies that address structural vulnerabilities while promoting environmental stewardship.

The rest of this paper is structured as follows. Section 2 provides a literature review, detailing the theoretical and conceptual frameworks. Section 3 outlines the research methodology. Section 4 presents the empirical results and includes a discussion of the findings. Finally, Section 5 concludes the paper with recommendations for sustainable development strategies.

2. Literature Review

2.1. Theoretical Approach

The relationship between poverty and environmental degradation has been explored through various theoretical frameworks. One prominent theory is the Environmental Kuznets Curve (EKC), which posits an inverted U-shape relationship between environmental degradation and economic development. According to this theory, as income rises, environmental degradation initially worsens but eventually improves after reaching a certain level of income (Dinda, 2005). While the Environmental Kuznets Curve (EKC) hypothesis has been widely used to explain the link between economic growth and environmental quality, its applicability to developing countries, like Ethiopia, is highly debated. Empirical studies (Stern, 2017; Vasanth et al., 2015) suggest that the EKC oversimplifies the complex relationship between poverty and environmental degradation by assuming that economic growth alone will eventually lead to environmental improvements. In reality, in low-income economies with weak institutions, environmental degradation may persist, even as income levels rise due to factors like poor governance, weak enforcement of environmental policies, and unequal wealth distribution (Zhang & Yan, 2022).

Another key theoretical framework is sustainable development, which emphasizes balancing economic growth with environmental protection and social equity (Burton, 1987). Similarly, the sustainable development framework is useful in advocating for a balance between economic, social, and environmental goals, but it lacks a strong empirical foundation on how poverty directly interacts with environmental degradation. Critics argue that the sustainable development goals (SDGs) often fail to account for the socio-political realities of resource distribution, corruption, and governance inefficiencies that hinder policy implementation in developing countries (Burton, 1987).

On the other hand, the multidimensional poverty framework (Alkire & Foster, 2008) provides a more comprehensive approach to understanding poverty by considering not just income but also education, healthcare, and access to basic utilities. Recent empirical studies (Zulkifli & Abidin, 2023) confirm that multidimensional poverty measures are better at capturing the various deprivations that contribute to both poverty and environmental degradation. For instance, lack of access to education and healthcare can limit households’ ability to adopt sustainable resource management practices, perpetuating poverty and ecological damage.

Given the limitations of the EKC and sustainable development models, this study adopts the multidimensional poverty framework as the most suitable theoretical foundation. This approach provides a holistic perspective by integrating economic, social, and environmental factors, making it more applicable to the Ethiopian context, where poverty is not merely a function of income but also of inadequate access to education, healthcare, and governance (Yıldırım et al., 2022). Moreover, it aligns with recent empirical findings that emphasize the importance of institutional quality, land tenure security, and policy-driven interventions in breaking the cycle of poverty and environmental degradation (Dalrymple, 2006; Leroy & Arts, 2014).

Thus, this study builds on the multidimensional poverty framework to assess how various deprivations contribute to environmental degradation, offering a more policy-relevant analysis for addressing poverty–environment linkages in Ethiopia.

2.2. Conceptual Approach

The conceptual framework for this study on multidimensional poverty in Bilate River Basin was grounded in both theoretical and empirical foundations. The study adopted the (Alkire & Foster, 2008) methodology for measuring multidimensional poverty, recognizing that poverty extends beyond income deprivation to encompass various dimensions such as education, health, and living standards. Empirical evidence from (Alkire & Santos, 2014) highlights the importance of a multidimensional approach in understanding poverty dynamics across developing countries, emphasizing that income-based measures alone fail to capture the full extent of deprivation. Additionally, the (Sen, 1999) capability approach provides a theoretical underpinning, arguing that poverty should be assessed based on individuals’ freedoms and abilities to achieve well-being rather than solely on economic metrics.

Key determinants of multidimensional poverty in Bilate River Basin include household income, education, access to basic services, and climatic vulnerabilities. Studies by (European Commission, 2016; Cooper, 2008) confirm that rural poverty is often linked to structural limitations such as limited infrastructure, poor access to credit, and lack of market integration. Climate variability and environmental shocks also exacerbate poverty, as evidenced by (Dercon, 2005), who demonstrated that weather-induced income fluctuations significantly impact food security and household welfare. Furthermore, recent studies in Ethiopia (Abay et al., 2022) emphasized the role of social capital and government safety nets in mitigating multidimensional poverty, suggesting that policy interventions must integrate both economic and social dimensions to be effective.

In this study, a multidimensional poverty index (MPI) framework was used to identify the prevalence and depth of deprivation across households, incorporating dimensions such as education (school attendance and literacy), health (child mortality and nutrition), and living conditions (access to clean water, sanitation, and electricity). Empirical findings from (Alkire et al., 2023) reinforce the necessity of adopting this multidimensional lens, as traditional monetary measures often underestimate the extent of deprivation in rural communities. The framework also accounts for the role of gender and household headship, drawing from empirical insights by (Huyer, 2016), which highlight that female-headed households tend to face higher multidimensional poverty due to systemic barriers to access to resources and decision-making power.

By integrating these empirical insights, this conceptual framework offers a comprehensive approach to analyzing multidimensional poverty in Bilate River Basin, guiding both analytical methods and policy recommendations. The framework in Figure 1 underscores the need for a holistic poverty alleviation strategy that combines economic growth with targeted social interventions to address underlying structural inequalities.

3. Materials and Methods

3.1. The Description of the Study Area

Bilate River Basin, situated in the southwestern escarpment of the Main Ethiopian Rift, is integral to the region’s hydrology and socio-economic activities (Edamo et al., 2022). Approximately 130 km northwest of Hawassa and 340 km from Addis Ababa, it spans multiple administrative zones, including Hadiya, Kambata Tembaro, Gurage, Silte, Wolaita, and Alaba within the South Ethiopia Region, as well as parts of Sidama and Oromia (Orke & Li, 2022). The basin covers about 5625 km², with elevations ranging from 1174 m to 3324 m (Edamo et al., 2022). Its topography transitions from steep, rugged highlands in the northwest to gentler slopes and lowlands in the south. The Bilate River originates from the southwestern slopes of Mount Gurage and empties into Lake Abaya.

The basin’s climate varies due to its topography, resulting in distinct agro-ecological zones. Annual rainfall ranges from 769–956 mm in the lower basin to 1280–1339 mm in the upper regions. Most of the basin falls within the Weyna Dega (moderate highland) and Kola (lowland) zones, supporting diverse agricultural activities. However, challenges such as water scarcity and reliance on rain-fed agriculture affect productivity and food security. The region is densely populated, with an estimated 2.4 million residents, many dependent on subsistence farming (Orke & Li, 2022).

Culturally, Bilate River Basin is home to diverse ethnic groups with distinct traditions and social structures. The upper and middle basin communities are predominantly Muslim, with a history of labor migration, particularly among women traveling to Arab countries for economic support (Mathewos et al., 2019). Marriage customs have evolved, with increased autonomy for women, though polygamy remains practiced in some areas. In contrast, the lower basin communities, mainly composed of Wolayita, Oromo, Sidama, and Kembata ethnic groups, follow Christian traditions and inherit land based on equitable distribution. Despite these differences, the various communities in the Bilate sub-basin have maintained strong cultural identities and interethnic harmony, with festivals and communal ties playing significant roles in their social lives.

Agriculture is the primary livelihood and employment source for about 70% of the Bilate community. Irrigation has improved incomes, particularly in the upper and lower streams, while the middle stream sees minimal benefits. Due to climate variability and limited irrigation access, many households rely on rain-fed farming, leading to frequent crop failures and food insecurity. This necessitates alternative income sources such as livestock rearing, trade, and aid from relatives, the government, and NGOs. Income disparities exist, with irrigation users earning significantly more than non-users, especially in the upper and lower streams. Non-agricultural income sources, including wood and wood products, husbandry, and small-scale trading, are essential but often insufficient due to low productivity and resource constraints. Poor infrastructure, limited education, health challenges, and inadequate sanitation further hinder economic development, while deforestation remains a concern due to high fuelwood dependency. Water access varies across the basin; rivers serve as the primary source in lower streams, while wells and pumps are predominant in the middle and upper streams, though issues such as water quality, cost, and availability persist. Sustainable interventions focusing on income diversification, improved infrastructure, and environmental conservation are necessary to uplift the Bilate community from chronic poverty. The administrative and geographic map of the study area is presented in Figure 2 below.

3.2. Selection of Sample Size and Sample Size Design

For this study, a multistage stratified sampling method was employed to collect primary data based on the diverse characteristics of the study area. These differences include variations in physical, resource, cultural, and socio-economic factors. The study area, the Bilate sub-basin, encompasses different socio-economic groups, with varying characteristics based on geographical location (upstream, middle stream, and downstream) and farming practices (irrigation vs. rain-fed agriculture). Due to these variations, the sample was divided into strata corresponding to these key features.

The first stage of sampling involved the purposive selection of three zones along the Bilate River in the Southern Nations, Nationalities, and Peoples’ Region (SNNPR): Silte (upstream), Alaba (middle stream), and Wolaita (downstream). In the second stage, a list of Woredas in each selected zone was compiled, and one Woreda was randomly chosen from each zone. In the third stage, two Kebeles from each selected Woreda were randomly chosen. Finally, in the fourth stage, a list of households near the river was prepared, and a systematic random sampling technique was applied to select 359 households. The sample was proportionally distributed across the different strata to ensure comprehensive coverage.

To determine the appropriate sample size, the (Cochrane, 2018) formula for large populations was applied. Initially, with a total population of 5496 households in the selected Kebeles, a sample size of 384 was calculated, with a 95% confidence level and a 5% margin of error. Since the population is finite, a correction formula was applied to adjust the sample size to 359. The corrected formula is as follows:

$$n={\displaystyle \frac{n_0}{1+\frac{(n_0-1)}{N}}}={\displaystyle \frac{384}{1+\frac{(384-1)}{5496}}}=359$$

(1)

where $n_0$ denotes the sample size obtained from Cochran’s formula for infinite populations (384); $N$ represents the total population size (5496); and $n$ is the final adjusted sample size (359). The details of the sample distribution are presented in

Table 1. Sample size distribution across zones, Woredas, Kebeles, and households.

Basin Cluster	Selected Zone	Selected Woreda	Selected Kebele (Village)	Total Number of Households in the Kebele	Number of Households Near the River	Sample Size per Kebele
Upstream Cluster	Silte	Sankura	Bonosha	705	455	30
			Regdina Kore	846	354	23
	Hulbareg		Mereb Kalkelo	2010	670	44
			Ambaricho Gimba	1698	972	63
Middle Catchment Cluster	Halaba	Wera	Andegna Choroko	761	573	37
			Tach Bedene	494	356	23
Downstream Cluster	Wolaita	Duguna Fango	Fango Boloso	864	550	36
			Fango Damot	1391	625	41
	Abala Abaya	Abala Abaya	Abaya Bilate	742	336	22
			Abaya Guricho	809	605	40
Total	3 Zones	5 Woredas	10 Kebeles	10,320	5496	359

Source: offices of selected districts, 2024.

below.

3.3. Data Type and Source

This study employed both primary and secondary data to achieve its objectives. Primary data were gathered through a semi-structured survey questionnaire administered to 359 sampled respondents, focusing on socio-economic conditions, natural resource degradation, and rural poverty. Secondary data were collected from various sources, including land use records, administrative boundaries, demographic statistics, and datasets from institutions, such as the Central Statistical Agency (CSA). Additionally, relevant information was sourced from peer-reviewed journals, academic textbooks, government and non-government reports, and bulletins to deepen the study’s analysis.

The data collection occurred between July and August 2024. A pretest was conducted prior to the full survey to ensure the clarity and reliability of the questionnaire. Ten trained enumerators under the supervision of three field supervisors and the researcher carried out data collection. A one-day training session was held to ensure consistency in administering the questionnaire. Ethical clearance for the study was obtained from the Ethics Review Committee of the College of Agriculture, Wolaita Sodo University (WSUCAERC). Due to cultural norms, verbal consent was obtained from participants after they were fully informed about the study’s purpose and assured of confidentiality and anonymity. The collected data were carefully reviewed for completeness, consistency, and accuracy before being analyzed.

3.4. Methods of Data Analysis, Variable Definition, and Hypothesis

3.4.1. Methods of Data Analysis

In the analysis of the determinants of multidimensional poverty in Bilate River Basin, various econometric models could have been considered to capture the relationship between independent variables and poverty status. Among the potential alternatives, the multinomial logit (MNL) model, the ordered logit (OLOGIT) model, and the Generalized Ordered Logit (GOLOGIT) model stand out as viable options. Each of these models presents distinct strengths and weaknesses in dealing with categorical dependent variables.

The multinomial logit (MNL) model is a widely used approach for modeling categorical outcomes where the response variable has multiple, unordered categories. This model does not impose a natural ordering of poverty levels (non-poor, vulnerable, poor, and extremely poor) and allows for independent effects of explanatory variables across outcome categories (Greene, 2017). However, one major limitation of the MNL model is the independence of irrelevant alternatives (IIA) assumption, which restricts the substitution pattern across choices and may not hold in many real-world applications (Train, 2003). The ordered logit (OLOGIT) model, on the other hand, assumes an inherent ordering among the dependent variable categories and estimates a single set of coefficients for all categories while using threshold parameters to differentiate among them (McKelvey & Zavoina, 1975). This model is efficient when the dependent variable follows a natural ranking, as is the case in multidimensional poverty classification. However, its primary limitation is the parallel regression assumption, also known as the proportional odds assumption, which constrains the effect of independent variables to be constant across all outcome levels (Long & Freese, 2014). This assumption may not always hold in practical applications, leading to biased estimates.

To address the limitations of the OLOGIT model, the Generalized Ordered Logit (GOLOGIT) model was chosen for this study. The GOLOGIT model relaxes the proportional odds assumption by allowing some coefficients to vary across response categories (Williams, 2016). This flexibility makes it more suitable for modeling multidimensional poverty, where the impact of determinants may differ across levels of deprivation. Empirical studies have demonstrated that the GOLOGIT model provides better fit and more nuanced insights into poverty determinants compared to the traditional OLOGIT model (Minyiwab et al., 2024). One notable strength of the GOLOGIT model in this analysis is its ability to capture the heterogeneous impact of independent variables across poverty categories.

The GOLOGIT model, also referred to as the Generalized Ordered Logit or Partial Proportional Odds Model, builds upon the foundational work of McCullagh, who introduced the cumulative logit model framework for ordinal outcomes. In this study, we implemented the GOLOGIT model using Stata version 17, utilizing the gologit² command developed by Richard Williams. This command allows for the flexible specification of which variables meet or violate the parallel lines assumption, providing a practical and statistically sound way to estimate the model.

Despite its advantages, the GOLOGIT model is computationally more intensive than the OLOGIT model and requires a larger sample size to achieve robust estimates (Williams, 2016). Additionally, the interpretation of coefficients is more complex due to the relaxation of the proportional odds assumption. To address these challenges, robust standard errors were employed to account for heteroskedasticity, mitigating the computational complexity of the GOLOGIT model. Since coefficient interpretation varies across outcome levels, marginal effects were computed to provide clearer insights into how changes in explanatory variables influence each poverty category. Additionally, given the study’s 359 observations, tests for model adequacy, including the likelihood ratio test and pseudo R-squared, were conducted to ensure that the sample size was sufficient for reliable estimation.

The Generalized Ordered Logit (GOLOGIT) model has been employed in various studies to analyze determinants of multidimensional poverty, particularly in contexts where the proportional odds assumption of the traditional ordered logit model may not hold. For instance, applied the GOLOGIT model to assess household poverty in Turkey, demonstrating its effectiveness in capturing varying effects of predictors across different poverty levels. Similarly, Lelisho et al. (2022) utilized the model to identify factors associated with household socio-economic status in Tepi Town, Southwest Ethiopia, highlighting its applicability in diverse socio-economic settings. However, despite these applications, a clear research gap remains in the context of rural Ethiopia. Specifically, few studies have applied the GOLOGIT model to investigate multidimensional poverty in geographically and institutionally unique areas like Bilate River Basin. Moreover, the existing literature often fails to simultaneously account for environmental and institutional determinants, which this study integrates. By addressing these gaps, this research contributes novel methodological and contextual insights to the poverty literature.

Generalized Ordered Logit Model (GOLOGIT) Specification

The Generalized Ordered Logit (GOLOGIT) model extends the standard ordered logit (OLOGIT) model by allowing some coefficients to vary across different poverty levels. The general form of the GOLOGIT model is specified as follows:

$$\ln \left({\displaystyle \frac{P\left(Y\le j\right)}{P\left(Y>j\right)}}\right)={\alpha }_j+{\beta }_{j}X$$

(2)

where $Y$ represents the ordered categorical variable for multidimensional poverty status (low, medium, or high). $P(Y\ge j)$ denotes the probability that a household falls at or below a given poverty category $j$; ${\alpha }_j$ represents the probability that a household falls at or below a given poverty category; ${\beta }_j$ is a vector of coefficients that determine how each independent variable influences the probability of being in a certain poverty category; and $X$ is the set of explanatory variables hypothesized to affect multidimensional poverty.

3.4.2. Description and Hypotheses of Variables

The study examined various socio-economic and demographic factors influencing poverty status among households. These factors include household characteristics, such as age, education, and dependency ratio, as well as economic variables, like land size, off-farm income, and credit access. Each variable was expected to have a significant impact on poverty dynamics, either positively or negatively. The definitions, measurements, and hypothesized relationships of these variables are presented in

Table 2. Description and hypotheses of variables.

Variable	Definition and Measurement	Hypothesis	Empirical Source (Google Scholar)
Poverty status	Dependent variable: 0 = non-poor, 1 = vulnerable, 2 = poor, 3 = extremely poor	–	Abunge et al. (2013)
Age	Age of household head in years	+	Komikouma et al. (2021)
Sex	Sex of household head (1 = male, 0 = female)	–	Woldehanna and Gebremedhin (2015)
Marital status	1 = married, 0 = others (single, divorced, widowed)	–	Woldehanna and Gebremedhin (2015)
Family size	Total number of individuals in a household	+	Teku and Eshetu (2024)
Dependency ratio	Non-working to working-age population ratio	+	Ginting et al. (2020)
Education level	0 = cannot read and write 1 = grade 1–8 2 = grade 9–12 3 = certificate and above	–	Bigsten and Shimeles (2004); Ginting et al. (2020)
Years of residence	Stay in the community in years	–	Keene et al. (2013)
Employment	0 = farmer 1 = non-farm worker 2 = daily laborer 3 = others	–	Sundaram (2007)
Livestock ownership	TLUs except for oxen	–	Bijla (2018)
Ox ownership	Number of oxen owned	–	Etim and Edet (2014)
Saving	1 = saves regularly, 0 = no saving	–	Steinert et al. (2017)
Information access	Access to market or extension information (1 = yes, 0 = no)	–	Cole et al. (2018)
Credit use	1 = accessed credit in the last 12 months, 0 = no credit used.	–	Das (2019)
Grazing land	Total area of grazing land owned (in hectares)	–	Briske et al. (2015)
Farm land	Total cultivated land in hectares	–	Gatzweiler and Baumüller (2014)
Health status	0 = in extreme hardship 1 = not very well 2 = neither good nor poor 3 = quite well off	+	Komikouma et al. (2021)

below.

4. Results and Discussion

4.1. Demographic and Socio-Economic Characteristics of the Sample Households

The survey of 359 households across Bilate River Basin (BRB) covered three zones, five districts, and ten Kebeles, ensuring diverse representation. Of these, 236 were male headed and 123 female headed. The results shown below in

Table 3. Descriptive statistics of continuous socio-economic and demographic variables.

Socio-Economic Variables	N	Minimum	Maximum	Mean	Std. Deviation
Age of Household Head (Years)	359	27	72	44.82	10.03
Family Size	359	2	12	5.32	2.07
Dependency Ratio	359	0	0.80	0.35	0.18
Livestock Ownership (TLUs)	359	0	7.69	2.06	1.95
Oxen Ownership (TLUs)	359	0	6	0.86	1.01
Farm Land Size (Hectares)	359	0	5	1.82	0.82
Grazing Land Size (Hectares)	359	0	5	1.31	1.29
On-farm	359	0	10	2.82	2.08
Non-farm	359	0	13	1.06	1.65
Off-farm	359	0	6	0.37	0.80

Source: own computation from survey data (2024).

examined socio-economic and demographic factors like age, family size, dependency ratio, landholding, and livestock ownership.

The age of household heads in the study area ranged from 27 to 72 years, with an average of 44.82 years, indicating that most were middle-aged and actively engaged in economic and farming activities. Additionally, household sizes varied between two and twelve members, averaging 5.32, reflecting the common reliance on family labor in rural agricultural settings.

Furthermore, the dependency ratio, measuring the proportion of non-working-age individuals, ranged from 0 to 0.80, with a mean of 0.35, suggesting moderate dependency. Livestock ownership varied significantly, ranging from 0 to 7.69 Tropical Livestock Units (TLUs), with an average of 2.06, highlighting the essential role of livestock in household economies for income, food, and farm labor.

Farm land ownership ranged from 0 to 5 hectares, averaging 1.82 hectares per household, while grazing land ownership averaged 1.31 hectares. These variations indicate differences in land accessibility, which impact agricultural productivity and livestock-rearing potential. Finally, oxen ownership, averaging 0.86 per household, was relatively low, possibly limiting draft power for farming and overall productivity.

In addition, household employment distribution across on-farm, non-farm, and off-farm activities reveals that on-farm employment remains the primary source of livelihood, with an average of 2.82 individuals engaged per household. Non-farm employment, with a mean of 1.06, shows more variability, while off-farm employment, averaging 0.37 individuals, contributes less to household income diversification. This highlights the centrality of agriculture in the region’s economy, despite some participation in non-farm and off-farm sectors.

The descriptive statistics of dummy and categorical variables are presented below in

Table 4. The descriptive statistics of dummy and categorical variables.

Variable	Category	Count	Valid Percentage (%)
Gender	Male	236	65.7
	Female	123	34.3
	Total	359	100%
Education	Cannot read and write	104	29.0
	Grades 1–8	164	45.7
	Grades 9–12	71	19.8
	Certificate and above	20	5.6
	Total	359	100%
Marital Status	Never married	18	5.0
	Married	304	84.7
	Divorced	22	6.1
	Widow	15	4.2
	Total	359	100%
Religion	Orthodox Christian	93	25.9
	Protestant	75	20.9
	Muslim	187	52.1
	Others	4	1.1
	Total	359	100.0%
Ethnicity	Silte	136	37.9
	Halaba	67	18.7
	Wolaita	120	33.4
	Others	36	10.0
	Total	359	100%
Stay in the Community	Less than 20 years	36	10.0
	21–40 years	69	19.2
	More than 40 years (whole life)	246	68.5
	Total	359	100%
Primary Employment	Farmer	331	92.2
	Non-farm worker	17	4.7
	Daily laborer	5	1.4
	Others	5	1.4
	Total	359	100%
Information Access	Yes	224	62.4
	No	135	37.6
	Total	359	100%
Credit Use	Yes	194	54.0
	No	165	46.0
	Total	359	100%
Saving Status	Yes	214	59.6
	No	145	40.4
	Total	359	100%
Health Status	In extreme hardship	9	2.5
	Not very well	48	13.4
	Neither good nor poor	124	34.5
	Quite well off	178	49.6
	Total	359	100%
Income Change	Decrease a lot	30	8.4
	Slightly decreased	90	25.1
	No change	99	27.6
	Slightly increased	112	31.2
	Increased a lot	28	7.8
	Total	359	100%

Source: own computation from survey data (2024).

. These statistics provide an overview of the socio-demographic and institutional characteristics of the surveyed households. The gender distribution showed that 65.7% of respondents were male, while 34.3% were female, indicating a male-dominated household structure.

In terms of education, 29% of respondents were illiterate, 45.7% completed primary education, 19.8% attained secondary education, and 5.6% had higher education. This highlighted significant educational disparities. Regarding marital status, 84.7% of households were married, while smaller proportions were divorced, widowed, or never married, indicating that marriage was the predominant family structure.

The ethnic composition was diverse, with 37.9% of respondents identifying as Silte, 33.4% as Wolaita, and 18.7% as Halaba, while 10% were from other ethnic groups. Most respondents (68.5%) had lived in their community for over 40 years, indicating deep-rooted ties, while others had lived there for shorter periods, reflecting limited migration.

Concerning institutional variables, 62.4% of respondents had access to information, and 54% used credit. In addition, 59.6% reported saving money, while 40.4% did not. Health-wise, 49.6% were in good health, while 34.5% were neither in good nor poor health. Income changes indicated a relatively stable economic condition for most respondents.

4.2. Major Environmental and Natural Resource-Related Problems

The survey results in

Table 5. Major environmental and natural resource-related problems.

SN	Main Environmental and Natural Resource-Related Problems	Level of the Problems
		High	Moderate	Low
1	Land degradation	152 (42.3)	141 (39.3)	66 (18.4)
2	Gully formation	91 (25.3)	193 (53.8)	75 (20.9)
3	Soil erosion	133 (37.0)	194 (54.0)	32 (8.9)
4	Storm water runoff	69 (19.2)	191 (53.2)	99 (27.6)
5	Sedimentation	66 (18.4)	187 (52.1)	106 (29.5)
6	Water scarcity	108 (30.1)	168 (46.8)	83 (23.1)
7	Deforestation	114 (31.8)	114 (31.8)	131 (36.4)

Source: own computation from survey data (2024).

revealed that land degradation is a prominent environmental issue, with 42.3% of respondents classifying it as a high-severity problem, indicating the need for immediate action. A further 39.3% rated it as moderate severity, suggesting that land degradation was widespread. While 18.4% categorized it as low severity, mitigation efforts were still necessary.

Gully formation emerged as another significant environmental concern, with 54.0% of respondents considering it a moderate severity problem. Additionally, 37.0% rated gully formation as high severity, underscoring its urgency. A smaller proportion of respondents (8.9%) saw it as a low-severity issue, but it still required attention and corrective measures.

Stormwater runoff was identified as a major environmental problem, with 53.2% of respondents perceiving it as a medium-level concern. A notable 19.2% viewed it as a high-level issue, while 27.6% regarded it as a low-level issue. This suggested that stormwater runoff was a widespread problem but with varying levels of perceived severity.

Finally, water scarcity and deforestation were also identified as significant environmental problems. Water scarcity was seen as a high-level concern by 30.1% of respondents, while deforestation was perceived as a high-level problem by 31.8% of respondents. Both issues showed a relatively balanced perception of severity, with a substantial portion viewing them as moderate problems as well.

4.2.1. The Extent of Natural Resource Degradation

The survey results in

Table 6. The extent of natural resource degradation.

SN	Main Problems	Extent of Natural Resource Degradation
		Large	Moderate	Less
1	Gully and land damage	152 (42.3)	141 (39.3)	66 (18.4)
2	Soil erosion	66 (18.4)	237 (66.0)	56 (15.6)
3	Water scarcity	35 (9.7)	172 (47.9)	149 (41.5)

Source: own computation from survey data (2024).

revealed varying perceptions of natural resource degradation. A significant portion of respondents (42.3%) viewed gully and land damage as a large-scale problem, while 39.3% considered it moderate, and 18.4% saw it as a lesser issue. For soil erosion, 18.4% saw it as a large-scale problem, 66.0% regarded it as moderate, and 15.6% viewed it as a low-scale issue. Water scarcity was perceived as a large-scale problem by 9.7% of respondents, with 47.9% considering it moderate and 41.5% regarding it as a low-scale issue. These results suggest varying levels of concern for these environmental challenges, with gully and land damage being viewed as the most severe, while water scarcity was seen as a relatively lesser problem for many.

4.2.2. The Rate of Natural Resource Depletion

The survey results in

Table 7. The rate of natural resource depletion.

What Is the Rate of Natural Resource Depletion in the District for the Last 10 Years	Frequency	Percent
Very rapidly	44	12.3
Rapidly	114	31.8
Moderately	198	55.2
Slowly	3	8
Total	359	100

Source: own computation from survey data (2024).

revealed varying perceptions of the rate of natural resource degradation over the last ten years. A small percentage of households, 12.3%, perceived the rate as very rapid, indicating significant concern over the alarming decline in natural resources. A larger portion, 31.8%, considered the degradation rate to be rapid, reflecting a noticeable decline in quality and quantity. The majority, 55.2%, viewed the rate as moderate, suggesting a concern for the long-term sustainability of the district’s resources. Only 8% of households perceived the degradation rate as slow, indicating minimal decline or relatively stable conditions during the past decade.

4.2.3. The Major Causes of Erosion

The survey results presented in

Table 8. The major causes of erosion.

SN	What Are the Major Causes of Erosion on Your Plot?	Rank
		No	1st	2nd	3rd	4th	Less
1	Heavy rainfall	22 (6.1)	160 (44.6)	73 (20.3)	52 (14.5)	1 (0.3)	51 (14.2)
2	Cultivation of steeper slop	26 (7.2)	96 (26.7)	59 (16.7)	45 (12.5)	10 (2.8)	123 (34.3)
3	Intensive cultivation without fallow	17 (4.7)	67 (18.7)	95 (26.5)	113 (31.5)	20 (5.6)	47 (13.1)
4	Wind	24 (6.7)	75 (20.9)	61 (17.0)	71 (19.8)	12 (3.3)	114 (32.3)
5	Overgrazing	26 (7.2)	101 (28.1)	64 (17.8)	72 (20.1)	19 (5.3)	79 (22.0)
6	Lack of sense of ownership	17 (4.7)	105 (29.2)	80 (22.3)	55 (15.3)	23 (6.4)	79 (22.0)
7	Fragmentation	24 (6.7)	58 (16.2)	33 (9.2)	46 (12.8)	25 (7.0)	173 (48.2)

Source: own computation from survey data (2024).

highlight the major causes of erosion on agricultural plots. Heavy rainfall was identified as the second most significant cause, with 44.6% of respondents attributing it as the primary factor contributing to erosion. Cultivation on steeper slopes followed closely, with 26.7% recognizing it as a primary cause, as steep gradients exacerbate water runoff and soil loss. Intensive cultivation without fallow periods was also a major contributor, with 26.5% of cases identifying it as a primary cause due to its impact on soil structure and organic matter. Wind erosion, overgrazing, and lack of land ownership were also significant, with 20.9%, 28.1%, and 29.2% of respondents attributing them as primary causes, respectively. Lastly, fragmentation was considered the least significant cause, with 16.2% identifying it as a primary factor. These results suggest that a combination of natural and human-induced factors is driving erosion, emphasizing the need for comprehensive soil conservation measures to address the issue.

4.3. Household Poverty Analysis

The analysis of household poverty in

Table 9. Proportion of deprivation in each indicator.

Dimension	Individual Dimensions	Deprivation (%)
Education	Years of schooling	30.4
	School attendance	61.6
Health	Child mortality	22.8
	Healthy facility access	44.4
	Health service quality	49.5
Living Standard	Living standard	60.4
	Drinking water	58.2
	Housing floor	57.7
	Housing roof	46.2
	Electricity	53.5
	Cooking fuel	39.3
Asset	Household property	25.6
	Land	50.4
	Livestock	56.5

Source: own computation from survey data (2024).

revealed significant deprivation across various dimensions. In education, 30.4% of individuals faced deprivation in terms of years of schooling, while a substantial 61.6% of children were deprived of regular school attendance. In health, 22.8% of children experienced mortality before reaching five years old, and 44.4% lacked access to healthcare facilities, with nearly half (49.5%) facing poor health service quality. Regarding living standards, 60.4% of individuals suffered from poor living conditions, 58.2% lacked access to safe drinking water, and 57.7% of households had inadequate flooring. Housing and electricity deprivation were also notable, with 46.2% and 53.5%, respectively, lacking adequate amenities. In terms of assets, 50.4% of individuals lacked access to land, and 56.5% were deprived of livestock ownership. These statistics highlighted the extensive deprivation in education, health, living standards, and assets, underscoring the need for targeted interventions to alleviate poverty and improve living conditions.

4.3.1. Disaggregated Deprivation by Bilate River Basins (Streams)

The disaggregated deprivation analysis across the Bilate River basins in

Table 10. Disaggregated deprivation by Bilate River basins (streams).

SN	Individual Dimensions	Deprivation by Streams (%)			X2 (p-Value)
		Lower Stream	Middle Stream	Upper Stream
Education	Years of schooling	66.88	38.33	44.60	21.47 (0.00)
	School attendance	44.38	86.67	70.50	40.64 (0.00)
Health	Child mortality	51.88	50.00	48.20	0.40 (0.81)
	Healthy facility access	49.38	48.33	56.12	1.70 (0.42)
	Health service quality	56.88	38.33	56.12	6.67 (0.03)
Living Standard	Living standard	53.75	73.33	62.59	7.43 (0.02)
	Drinking water	57.50	31.67	71.94	31.14 (0.00)
	Housing floor	60.00	55.00	56.12	0.66 (0.71)
	Housing roof	53.13	50.00	53.96	0.26 (0.87)
	Electricity	51.25	60.00	53.24	1.34 (0.51)
	Cooking fuel	42.50	45.00	64.03	14.92 (0.00)
Asset	Household property	51.25	50.00	42.45	2.47 (0.29)
	Land	55.00	51.67	44.60	3.26 (0.19)
	Livestock	79.38	21.67	45.32	70.76 (0.00)

Source: own computation from survey data (2024).

revealed significant variations in deprivation levels across different streams. In education, the middle stream had the highest deprivation in school attendance (86.67%), while the lower stream had the highest deprivation in years of schooling (66.88%). Health-related dimensions, such as child mortality and access to healthcare, showed no significant associations with streams, but health service quality had a notable association, with the lower stream facing the highest deprivation (56.88%). In terms of living standards, the middle stream had the highest deprivation in living standards (73.33%) and access to drinking water (71.94%), while the upper stream also showed high deprivation in drinking water and cooking fuel. These disparities highlight the uneven access to essential services across the streams.

Asset ownership also displayed stark differences, particularly in livestock ownership, where the lower stream experienced a high deprivation rate of 79.38%, compared to much lower rates in the middle and upper streams. In contrast, land and household property did not show significant variations between streams, suggesting a more uniform distribution of these assets. Overall, the analysis revealed that deprivation varied significantly across several dimensions, such as education, health service quality, and living standards, while other factors, like housing conditions and asset ownership, showed less variation. These findings emphasize the need for targeted interventions to address the specific deprivations in each stream, particularly in the most disadvantaged segments of the population.

4.3.2. Multidimensional Poverty Analysis

The multidimensional poverty analysis in

Table 11. Status of multidimensional poverty.

Status of Indicator Variables	Value
Poverty cutoff (k)	33.3%
Total sample size	359
Multidimensional non-poor household number	73
Multidimensional poor households (q)	286
Incidence/head count ratio (H)	1511/1912 = 0.79
The intensity of poverty (A)	75,961.40/1511 = 50.27
Multidimensional poverty index (MPI)	0.79 × 0.503 = 0.39.77

Source: own computation from survey data (2024).

revealed that out of 359 households, 286 were identified as poor, while 73 were non-poor. The incidence of poverty was 79%, with poor individuals experiencing an average deprivation intensity of 50.27. The multidimensional poverty index (MPI) was calculated at 39.77%, indicating a significant level of poverty in the population, reflecting both its extent and depth.

4.3.3. Contribution of Dimensions on Multidimensional Poverty

The analysis of the contribution of individual indicators to multidimensional poverty in

Table 12. Contribution of individual indicators.

Individual Dimensions	Assigned Weight	Contribution	Remark
Years of schooling	12.5	7.21	Below
School attendance	12.5	14.63	Above
Child mortality	8.33	5.43	Below
Healthy facility access	8.33	10.52	Above
Health service quality	8.33	11.71	Above
Living standard	4.17	14.36	Above
Drinking water	4.17	14.49	Above
Housing floor	4.17	13.70	Above
Housing roof	4.17	10.99	Above
Electricity	4.17	12.71	Above
Cooking fuel	4.17	9.33	Above
Household property	8.33	6.09	Below
Land	8.33	11.98	Above
Livestock	8.33	13.43	Above

Source: own computation from survey data (2024).

revealed varying degrees of influence across dimensions. Years of schooling, with an assigned weight of 12.5%, contributed 7.21% to the overall poverty measurement, categorized as “Below.” Similarly, child mortality, with a weight of 8.33%, contributed 5.43%, also falling in the “Below” category. In contrast, school attendance had a higher contribution of 14.63%, categorized as “Above,” indicating a stronger impact on poverty assessment.

Several health-related dimensions, such as health facility access and health service quality, showed significant contributions. Health facility access contributed 10.52%, while health service quality had a notable contribution of 11.71%, both categorized as “Above.” In terms of living standards, drinking water and housing-related factors, such as housing floor and roof quality, also had notable contributions. Drinking water and housing floor each contributed over 14%, while housing roof and electricity contributed 10.99% and 12.71%, respectively, reflecting their substantial influence on the multidimensional poverty measure.

Finally, asset-based dimensions, including land and livestock, made important contributions to the poverty measurement. Land ownership contributed 11.98%, while livestock ownership had the highest contribution at 13.43%. Household property, with a contribution of only 6.09%, was categorized as “Below,” indicating a lesser influence on multidimensional poverty. These findings illustrate the varying impacts of different poverty dimensions, emphasizing the importance of health, living standards, and assets in shaping the overall poverty landscape.

4.3.4. Contribution of Dimensions

The assessment of multidimensional poverty within the Bilate River basins in

Table 13. Contribution of dimensions.

Dimension	Contribution
Education	26.00
Health	21.15
Living Standard	28.33
Asset	24.52

Source: own computation from survey data (2024).

, revealed the significant contributions of four key dimensions: education, health, living standards, and assets. Education accounted for 26.00% of the overall poverty measurement, underscoring its critical role in determining well-being through factors such as years of schooling and school attendance. This finding aligned with the global multidimensional poverty index (MPI), which emphasizes education as a fundamental component in assessing poverty levels (Alkire & Santos, 2010).

Health contributed 21.15%, reflecting the impact of child mortality, access to healthcare, and service quality on poverty. This result is consistent with studies conducted in Southern Ethiopia, which found that health deprivations significantly influence multidimensional poverty (Eshetu et al., 2022).

Living standards had the highest contribution at 28.33%, indicating the substantial effect of essential services like access to clean drinking water, adequate housing, and electricity. Similar findings were reported in a study focusing on urban households in Wolaita Sodo, Ethiopia, where living standards were identified as a major determinant of multidimensional poverty (Joshua et al., 2017).

Finally, assets contributed 24.52%, highlighting the importance of economic resources such as land, property, and livestock in sustaining livelihoods. This observation aligns with research conducted in the Upper Blue Nile basin, Ethiopia, which emphasized the role of asset ownership in mitigating poverty (Abeje et al., 2020).

4.3.5. Severity and Vulnerability

The analysis of multidimensional poverty status in the Bilate River basins in

Table 14. The association between multidimensional poverty status and Bilate River basins.

	Categories	Bilate River Basins			Total	X2 (p-Value)
		Lower Stream	Middle Stream	Upper Stream
Poverty status	Non-poor	3	1	6	10	27.93 (0.00)
	Vulnerable	27	6	30	63
	Poor	84	36	38	158
	Extremely poor	46	17	65	128
Total		160	60	139	359

Source: own computation from survey data (2024).

revealed significant variations across different streams. Among the 359 households, 10 in the lower stream, 1 in the middle stream, and 6 in the upper stream were classified as non-poor, totaling 17. The number of vulnerable households stood at twenty-seven in the lower stream, six in the middle stream, and thirty in the upper stream, summing up to sixty-three. Poor households were 84 in the lower stream, 36 in the middle stream, and 38 in the upper stream, totaling 158. Severely poor households comprised 46 in the lower stream, 17 in the middle stream, and 65 in the upper stream, reaching 128. The chi-square test (X² = 27.93, p-value = 0.00) indicated a statistically significant association between poverty status and the streams. Similar patterns have been observed in other river basins. For instance, a study in the Upper Blue Nile Basin in Ethiopia found significant differences in multidimensional poverty across various agro-ecological settings, highlighting the influence of environmental and locational factors on poverty levels (Abeje et al., 2020). Additionally, research in the Tupiza River Basin in Bolivia demonstrated that remote rural communities experienced higher multidimensional poverty, emphasizing the role of geographic isolation and limited access to resources (Espinoza et al., 2022). These studies underscore the importance of considering geographic and environmental contexts when addressing poverty in river basin areas.

4.4. Determinants of Multidimensional Poverty

The analysis of multidimensional poverty in Bilate River Basin in

Table 15. Determinants of multidimensional poverty in Bilate River Basin.

Variables	Delta Method
	Poverty Status (0 = Non-Poor, 1 = Vulnerable, 2 = Poor, 3 = Extremely Poor).	dy/dx	Std.Err	z
Age	0	−0.001 ***	0.000	−2.400
	1	−0.004 ***	0.001	−3.130
	2	−0.002 ***	0.001	−2.640
	3	0.007 ***	0.002	3.260
Sex	0	0.004	0.006	0.700
	1	0.017	0.024	0.710
	2	0.007	0.009	0.700
	3	−0.028	0.039	−0.710
Marital status	0	−0.009	0.008	−1.040
	1	−0.037	0.034	−1.090
	2	−0.014	0.013	−1.050
	3	0.060	0.055	1.090
Family size	0	0.002	0.001	1.210
	1	0.008	0.006	1.290
	2	0.003	0.002	1.240
	3	−0.012	0.009	−1.300
Dependency ratio	0	−0.070 ***	0.024	−2.880
	1	−0.298 ***	0.067	−4.450
	2	−0.115 ***	0.036	−3.190
	3	0.484 ***	0.101	4.790
Education level	0	−0.006	0.004	−1.650
	1	−0.027 *	0.015	−1.850
	2	−0.010 *	0.006	−1.730
	3	0.044 *	0.023	1.870
Years of residence	0	−0.005	0.006	−0.700
	1	−0.019	0.027	−0.710
	2	−0.007	0.011	−0.690
	3	0.031	0.044	0.710
Employment	0	0.001	0.010	0.090
	1	0.004	0.044	0.090
	2	0.002	0.017	0.090
	3	−0.006	0.072	−0.090
Livestock ownership	0	0.006 **	0.003	2.190
	1	0.027 ***	0.010	2.700
	2	0.010 **	0.004	2.310
	3	−0.043 ***	0.016	−2.760
Ox ownership	0	−0.011 *	0.006	−1.710
	1	−0.044 *	0.023	−1.910
	2	−0.017 *	0.010	−1.750
	3	0.072 *	0.037	1.930
Saving	0	−0.010	0.006	−1.540
	1	−0.041 *	0.024	−1.690
	2	−0.016	0.010	−1.570
	3	0.066 *	0.039	1.700
Information access	0	−0.026 ***	0.009	−2.850
	1	−0.110 ***	0.025	−4.390
	2	−0.043 ***	0.013	−3.260
	3	0.179 ***	0.037	4.770
Credit use	0	−0.023 ***	0.008	−2.700
	1	−0.097 ***	0.024	−4.030
	2	−0.037 ***	0.013	−2.930
	3	0.157 ***	0.038	4.180
Grazing land	0	0.008 ***	0.003	2.550
	1	0.035 ***	0.011	3.290
	2	0.014 ***	0.005	2.510
	3	−0.057 ***	0.017	−3.350
Farm land	0	0.002	0.003	0.740
	1	0.010	0.014	0.760
	2	0.004	0.005	0.760
	3	−0.017	0.022	−0.770
Health status	0	−0.034 ***	0.012	−2.770
	1	−0.143 ***	0.036	−4.010
	2	−0.055 ***	0.018	−3.110
	3	0.231 ***	0.054	4.320

***, **, and * indicate the level of significance at 1%, 5%, and 10%, respectively; SE: standard error. Source: authors’ computation (2024).

revealed several significant determinants. Age has a marginally significant effect on poverty outcomes. Each additional year of age decreases the likelihood of being non-poor by 0.1% (dy/dx = −0.001, p < 0.01), vulnerable by 0.4% (dy/dx = −0.004, p < 0.01), and poor by 0.2% (dy/dx = −0.002, p < 0.01) while increasing the likelihood of being extremely poor by 0.7% (dy/dx = 0.007, p < 0.01). This suggests that older individuals are more susceptible to extreme poverty, potentially due to declining physical capabilities and limited income-generating opportunities. This finding aligns with studies indicating that aging can increase vulnerability to poverty due to reduced labor capacity and higher health-related expenses (Wang et al., 2024).

However, this finding must be interpreted with caution due to potential model limitations, such as omitted variable bias and the assumption of linear marginal effects across age brackets, which may not hold in diverse household contexts.

The dependency ratio exhibited a significant impact on poverty outcomes. Each unit increase in the dependency ratio decreases the likelihood of being non-poor by 7.0% (dy/dx = −0.070, p < 0.01), vulnerable by 29.8% (dy/dx = −0.298, p < 0.01), and poor by 11.5% (dy/dx = −0.115, p < 0.01) while increasing the likelihood of being extremely poor by 48.4% (dy/dx = 0.484, p < 0.01). This indicates that households with higher numbers of dependents relative to working members are more likely to experience extreme poverty, possibly due to increased financial burdens and limited income. Similar findings have been reported in studies emphasizing the strain of high dependency ratios on household resources (Ginting et al., 2020). Nonetheless, the model results here should be interpreted within the context of household heterogeneity and labor market access, which the model may not fully account for.

Education level has a marginally significant effect on poverty outcomes. Each unit increase in education level decreases the likelihood of being vulnerable by 2.7% (dy/dx = −0.027, p < 0.1) and poor by 1.0% (dy/dx = −0.010, p < 0.1) while increasing the likelihood of being extremely poor by 4.4% (dy/dx = 0.044, p < 0.1). This suggests that higher education levels are associated with reduced vulnerability and poverty, potentially due to better employment opportunities and income prospects. However, the increase in extreme poverty likelihood may reflect other underlying factors not captured in this analysis. The importance of education in poverty reduction is well-documented in the literature (Wang et al., 2024). This counterintuitive pattern calls for caution, as it may arise from unobserved heterogeneity or reverse causality, which could not be addressed due to model scope limitations.

Livestock ownership significantly influences poverty outcomes. Each additional unit of livestock increases the likelihood of being non-poor by 0.6% (dy/dx = 0.006, p < 0.05), vulnerable by 2.7% (dy/dx = 0.027, p < 0.01), and poor by 1.0% (dy/dx = 0.010, p < 0.05) while decreasing the likelihood of being extremely poor by 4.3% (dy/dx = −0.043, p < 0.01). This suggests that livestock serves as a valuable asset, contributing to improved economic status and reducing the risk of extreme poverty. This finding is consistent with studies highlighting the role of livestock in enhancing household resilience and income (Ginting et al., 2020). Yet, one should note the potential for omitted variables, such as livestock productivity and access to veterinary services, which could affect the strength of these relationships.

Similarly, ox ownership has a marginally significant effect on poverty outcomes. Each additional ox decreases the likelihood of being non-poor by 1.1% (dy/dx = −0.011, p < 0.1), vulnerable by 4.4% (dy/dx = −0.044, p < 0.1), and poor by 1.7% (dy/dx = −0.017, p < 0.1) while increasing the likelihood of being extremely poor by 7.2% (dy/dx = 0.072, p < 0.1). This counterintuitive result may reflect the financial burden associated with ox ownership or the possibility that households invest in oxen as a coping mechanism during economic hardship. Livestock act as a method of savings as well as a revenue source, enhancing household food security and general well-being. The empirical findings by (Beyene et al., 2023) revealed that livestock ownership significantly enhances household food security, thus aiding in poverty alleviation. However, this finding may be context dependent and limited in its external validity beyond the study area.

The access to information variable, which equips households with insights into market opportunities, agricultural methods, and financial management, significantly affects poverty outcomes. Each unit increase in information access decreases the likelihood of being non-poor by 2.6% (dy/dx = −0.026, p < 0.01), vulnerable by 11.0% (dy/dx = −0.110, p < 0.01), and poor by 4.3% (dy/dx = −0.043, p < 0.01) while increasing the likelihood of being extremely poor by 17.9% (dy/dx = 0.179, p < 0.01). This suggests that while information access generally reduces poverty, it may also expose households to new risks or challenges that could lead to extreme poverty, which can include exposure to misinformation, cybercrime, privacy breaches, and the exploitation of vulnerable populations through online scams (Dzator et al., 2023). This result was also supported by (Abate & Schaapp, 2022), who revealed that households with access to agricultural extension services and market information were more productive and had higher income, thereby decreasing their poverty risks. Still, due to data limitations, the quality and source reliability of the information accessed were not controlled, which may compromise the interpretation of this result.

Credit use has a significant impact on poverty outcomes. Each unit increase in credit use decreases the likelihood of being non-poor by 2.3% (dy/dx = −0.023, p < 0.01), vulnerable by 9.7% (dy/dx = −0.097, p < 0.01), and poor by 3.7% (dy/dx = −0.037, p < 0.01) while increasing the likelihood of being extremely poor by 15.7% (dy/dx = 0.157, p < 0.01). This indicates that while credit access can alleviate poverty, it may also lead to increased financial vulnerability if not managed properly. Households that access credit without adequate financial literacy or sustainable income sources may fall deeper into debt, exacerbating their poverty status. This paradox is reflected in recent findings that credit access alone is not a panacea for poverty alleviation and must be coupled with appropriate support services, like financial training and market linkages (African Development Bank & African Development Fund, 2006). It is important to recognize that the model does not capture repayment behavior or interest burden, which are critical in understanding credit outcomes.

Access to grazing land shows a statistically significant association with all poverty categories. Specifically, an increase in grazing land raises the likelihood of being non-poor, vulnerable, and poor by 0.8%, 3.5%, and 1.4%, respectively, and all are significant at the 1% level. However, it reduces the probability of being extremely poor by 5.7%, and it is also significant at the 1% level. This suggests that access to grazing land plays an important role in preventing households from descending into deeper levels of poverty (Moreda, 2023). This conclusion, however, is constrained by the absence of data on land quality and productivity, which limits the explanatory power of the result.

Lastly, health status significantly influences poverty outcomes across all categories. Improved health status is associated with a 3.4% decrease in the likelihood of being non-poor, a 14.3% reduction in being vulnerable, and a 5.5% decrease in the chance of being poor, and all are significant at the 1% level. Conversely, poor health significantly increases the likelihood of being extremely poor by 23.1%. These findings highlight the critical role of health in poverty dynamics, where poor health considerably exacerbates extreme poverty risks (Devine et al., 2014). Yet, health outcomes may be endogenous to poverty status, introducing a potential bias that the current model does not address.

5. Conclusions and Recommendations

In conclusion, this study provided a comprehensive exploration of the complex relationship between multidimensional poverty and environmental degradation in the Bilate sub-basin of Southern Ethiopia. By adopting a multidimensional poverty framework and employing a robust sampling strategy across diverse agro-ecological zones, the research offered valuable insights into how socio-economic, environmental, and institutional factors interact to influence household poverty status. The use of the Generalized Ordered Logit (GOLOGIT) model allows for a nuanced understanding of the determinants affecting varying degrees of poverty from non-poor to extremely poor. These findings underscore the need for integrated, context-specific policy interventions that address environmental sustainability alongside poverty alleviation, especially in vulnerable rural settings dependent on natural resources.

The analysis of multidimensional poverty determinants in Bilate River Basin revealed that various socio-economic and demographic factors significantly influence household poverty outcomes. Age and education were marginally significant, with older individuals and less educated households more likely to fall into extreme poverty, indicating that aging populations and educational deprivation contribute to heightened vulnerability, particularly in subsistence-based rural communities. High dependency ratios and poor health status drastically increased the probability of extreme poverty, as these conditions reduce household labor capacity and increase care burdens, making it harder to escape poverty. Asset ownership, particularly livestock and grazing land, appeared to offer resilience, reducing the likelihood of being extremely poor, while paradoxical effects were observed with ox ownership and information access, potentially reflecting deeper structural issues, such as disparities in resource quality or access to reliable extension services. Moreover, credit use and information access, although traditionally seen as empowerment tools, were associated with increased extreme poverty risks possibly due to mismanagement or insufficient support systems. This highlights the need for complementary capacity building and institutional safeguards to ensure these tools yield positive outcomes. These findings underscore the complexity of multidimensional poverty, highlighting that addressing it requires a multifaceted approach encompassing health, education, household composition, productive assets, and support mechanisms for credit and information use.

The findings suggested that multidimensional poverty in Bilate River Basin was deeply influenced by age, dependency burden, health, education, asset ownership, and access to services, such as credit and information. Policymakers should prioritize integrated poverty reduction strategies that address both economic and non-economic dimensions. Enhancing access to quality education and healthcare, particularly for aging and dependent household members, is critical. Furthermore, promoting livestock ownership and secure access to grazing land can strengthen household resilience. Credit and information services must be coupled with financial literacy, risk mitigation, and extension support to avoid deepening vulnerabilities. Targeted interventions that consider household structure and empower communities with sustainable livelihood assets and support systems are essential for effectively combating extreme poverty in the region.

Policy Implications: Strengthening educational and health services is necessary to reduce multidimensional poverty. Enhancing asset-based interventions, especially support for livestock and land access, can significantly improve household resilience. Additionally, reforming credit and information systems, ensuring they are accessible and linked to poverty reduction outcomes, is vital for long-term impact.