Effect of soil and water conservation structures on smallholder farmers’ livelihood: Wenago district, Southern Ethiopia

Abstract Low standard of constructed structures, weak land management, unskilled human power, dependency on rain-fed agriculture and natural assets for household farmers’ survival make livelihood systems in the study area complex, dynamic and food insecure. The aim of this research paper was to assess and analyze the effect of physical structures on crop yield, annual income and smallholder farmers livelihood based on the data of 2020 and 2021 harvest years. The data was collected from a survey of 262 total household head farmers of which 132 were structureadopters while 130 were nonadopters (nonusers). Household heads were selected using simple random sampling techniques. The statistical data were analyzed using chi-square, t-test and SPSS version 20 software while the economic data were analyzed using econometric models such as propensity score matching, average treatment effect and logistic regression. Focus group discussions, key informant interviews and personal observations were also held to gather qualitative data. The results show that the adoption of physical structures has had a significant and positive impact on smallholder farmers’ incomes, crop productivity and livelihoods. Therefore, the adoption of physical structures in Wenago district, southern Ethiopia should be continued and encouraged as much as possible.


PUBLIC INTEREST STATEMENT
The study is aimed to provide the overall information regarding the role of implementing improved structures on smallholder farmers' livelihood improvement.The paper further attempts to show the importance of soil and water conservation structures (stone terracing and soil bunds) on increasing crop yield, smallholder farmers' income drawing lessons from Wenago district, southern Ethiopia.The paper has a significant contribution to soil erosion control, retaining soil moisture, agricultural productivity, rehabilitation of degraded farm lands and enhancing socio-economies.As conclusion, the paper argues that adoption of improved physical structures on farm lands have a great role in sustainable land management, soil erosion control, household farmers' income and livelihood improvement.For more advancement, the implementation of physical structures should be supported by continuous technical training, field demonstration, stakeholders' cooperation and community participation to ensure food security of rural smallholder farmers in local, regional and global levels.
data.The results show that the adoption of physical structures has had a significant and positive impact on smallholder farmers' incomes, crop productivity and livelihoods.Therefore, the adoption of physical structures in Wenago district, southern Ethiopia should be continued and encouraged as much as possible.

Introduction
Agriculture is the primary source of livelihood for an estimated 86% and 85% of rural people of Ethiopia and the world, respectively (Belay & Bewket, 2013;Bigsten & Tengsten, 2011;Hishe et al., 2019).The Ethiopian economy is primarily based on traditional agriculture, which is a subsistence economy (Abebe & Bekele, 2014;Adgo et al., 2013;Bewket, 2007).Agriculture provides jobs for nearly 85 million smallholder farmers and landless workers in Ethiopia (Bekele & Drake, 2003;Masha et al., 2021).Unfortunately, in the economy, the sector is characterized by a scarcity of factors of production (technology), farming methods and a low level of performance (Moges & Taye, 2017).Due to severe climate risks, weak land management, lack of technical skill in the implementation of improved soil and water conservation measures and low crop productivity in Ethiopia, in the past two decades research focus has been mostly on the on-site effects of soil erosion (Moges & Taye, 2017).As a result, the livelihood of its citizens has been threatened by these economic factors, often leading to serious food insecurity situations (Anley et al., 2007;Belayneh, 2023;Wolka et al., 2018).
Previous studies by (Hurni, 1991;Wolka et al., 2013) showed that various personal (attitude, perception, awareness), economic (financial income), institutional (fertilizer, credit, extension service) and biophysical attributes have determinant factors in smallholder farmers decision on adoption of improved SWC structures in Ethiopia.Other factors that affect Ethiopia and the study region are rainfed agriculture and investment which negatively influence smallholder farmers' livelihood status (Belay & Bewket, 2013;Chesterman et al., 2019).In southern Ethiopia and the study area, over 88% of the population depends on rain-fed agriculture for their livelihoods (Ellis, 2000b;Semu, 2018.According to Tadesse (2002), the livelihood of rural households in Ethiopia and the study district (Wenago Wereda) is primarily based on subsistence agriculture.Subsistence agriculture is also facing challenges such as land degradation, severe soil erosion and coffee pests, which result in low crop productivity, food deficit, low income and poor household livelihood status.
Recognizing all these serious threats of soil erosion problems and declining agricultural productivity, the government of Ethiopia began a program of natural resources conservation activity since 1973 supported by donors and nongovernmental organizations (Moges & Taye, 2017).Between 1976 and1988 through food-for-work (FFW) programmers, 800,000 km of soil and tone bunds on cultivated land were constructed; 600,000 km of hillside terraces were built; and 80,000 hectares were closed for regeneration and for a forestation of steep slopes (Hurni, 1991;FAO, 1993;Masha et al., 2021).However, reports indicated that these conservation structures have not been sustainably used by the farmers due to the fact that planners and implementing agencies ignored local level biophysical and socio-economic realities (Semu, 2018;Walie & Fisseha, 2016;World Bank, 2000).
In this research study, SWC structures refer to: stone terraces, soil bunds and contour ploughing and are scaled up through a concerted effort of extension workers, agricultural experts, household farmers and other relevant non-state actors.The physical designs of each improved structure have specified parameters or dimensions (length, width, height, depth and spacing) which are commonly implemented in Wenago district and guided by the Development agents (DAs), experts and other concerned bodies.However, as this determines the growth of crop productivity, there is a need for careful selection of appropriate soil and water conservation (SWC) technologies to ensure sustainability (Nyambose & Jumbe, 2013;Tadesse, 2002).Based on the information gathered from Gedeo Zone Department of Forest Environmental Protection and Climate Change Regulation Office (GZFEPCCRO), a continuous practice of SWC structures in different districts is significantly increasing from year to year.According to Wenago Wereda forest environmental protection and climate change regulation unit (WWFEPCCRU) annual report (2021), approximately 7, 15, 15 and 20 km of improved SWC structures were installed in 2018, 2019, 2020 and 2021, respectively, at the study district level.
On the other hand, traditional SWC practices are based on individual farmers' level of perception, farm size holding and availability of materials.So the dimensions of the measures vary from one farmer to another.As a result, traditional soil and water conservation practices in the study are suffered with weak land management systems, misuse of natural resources, technical skill scarcity compared to adoption of improved SWC structures.
Therefore, the goal of this study was to examine and investigate the impact of adoption of improved soil and water conservation structures on smallholder farmers' crop yield, income, livelihoods and sustainable land management by comparing adopters of structures and nonadopter households.In this study, livelihood is assessed in terms of farm-land holdings, main crop types grown, annual crop yield, income, livestock owing, off-farm activities and institutional support services in the study area.
Findings obtained from this research can aid or assist policy-makers, planners, stakeholders, professional researchers, experts, local communities to enhance adoption of improved SWC structures in Wenago district and other parts with similar climatic conditions to Ethiopia.

Location and description of the study area
The study was conducted in Wenago district, Gedeo Zone, Southern Ethiopia, 375 km south of the capital city of Addis Ababa.The study area is located between 6° 12' 30 '' and 6° 22' 30'' N latitude and 38°15'0'' and 38°20'0'' E longitude (Figure 1, and Ethio-GIS arc, 2017).Based on Gedeo Zone Finance And Economy Department Annual report (2020), the recent population projection of the study district has a total population of 152,000, with an area of 248 square kilometer giving a population density of 613 persons/km 2 .The majority of the populations in the study area (90.8%) live in rural areas and the rest (9.2%) live in small town centers.The most implemented structures in Wenago district are: stone terraces, soil bunds and contour ploughing.The study is bordered to the west by Abaya district; to the south by Yirga Cheffie district; to the north by Sidama Zone and to the east by Dilla Zuria Wereda.The study area has a favorable climate for agroforestry dominated agricultural activities.Rainfall ranges from 812 to 1634 mm, while mean annual temperature varies from 13°C to 27.5°C.Nitosols are dominant soil types covering highest proportion of the study area.The soils are in general derived from volcanic rocks which are important for coffee growing areas (Tadesse, 2002)."SNNPR" on the map of Figure 1, stands for South Nation Nationalities and People Region which means southern Ethiopia.The name "Kebele" in this study refers to the lowest administration units (sub-division of a district) in the administrative structure of Ethiopia.

Research approach
This study incorporated a hybrid methodological approach (both qualitative and quantitative) that combined conventional survey tools, such as household surveys, focus group discussions, interviews and personal observations.Farming household heads are the main bodies responsible for making dayto-day decisions on land use and land management (soil and water conservation).Thus, structure adopters and non-adopter households are the basic sample units of this study, and the survey was conducted at the household level in the selected sample sites.Adopter and non-adopter House Hold Heads (HHH) were identified from the list of Wenago Wereda Forest Environmental Protection and Climate Change Regulation Unit (WWFEPCCRU) report (2021).The collected data were analyzed separately for each sample group, allowing for the identification of inter-community differences and similarities.The targeted study district was selected based on its relatively uniform topography, agricultural systems highly affected by soil erosion problems, and active SWC intervention practices.A cross-sectional survey design was used (i.e.grouping the rural people in to control and treatment groups) because rural livelihood systems are not amenable to experimentation.The cross-sectional design involved the observation of a representative subset at a definite time.

Data sources
The primary data were generated from household head surveys, Development Agents (DA's), district agricultural experts, supervisors, key informants interview (KII), focus group discussions (FGD), observations and government and non-government organizations (NGOs).Secondary data were gathered from published materials (books, journals, magazines and newspapers), unpublished sources (government reports, and documents and internet).

Data collection tools
In order to collect the relevant primary data from the sample households, the authors used basic instruments such as structured Questionnaire, semi-structured interview, focus group discussion, and checklists for observation and photographs.The authors employed both quantitative and qualitative approaches to collect data from the primary and secondary sources.Quantitative data were collected from 262 randomly selected smallholder farmers.In the survey research, the researcher asked a random sample of individuals to respond to a set of questions about their livelihoods, assets, farming activities, main crop types, skill capacities, income, livestock holdings, and access to natural resources, off-farm activities, so forth.

Population and sample size determination
The target population of the study was 4463 household farmers, of which 2,237 were structure adopters and 2,226 were nonadopters.The survey sample size was determined using the formula proposed by Glenn (1992).n = N/1+N (e) 2 , where n is the required sample size, N is the total target population (4463), and e is the error limit (0.06).Therefore, 262 sample households were selected to represent the study district, assuming confidence level of 94%.

Sampling techniques
The study is a case-control study, because the researcher had used people in the control and treatment groups.In the first step, relevant information about which district is the most affected by soil erosion problems and active human interference was gathered from Gedeo Zone Department of Forest Environmental Protection and Climate Change Regulation Office (GZFEPCCRO) (WWFEPCCRU 2021).Based on the gathered information, of the eight rural districts of Gedeo Zone, Wenago district which has seventeen kebeles was purposely identified as one of the most affected districts by soil erosion problem.
In the second step, to get the right sample area, supportive information was again collected from Wenago Wereda Forest Environmental Protection and Climate Change Regulation Unit (WWFEPCCRU) (2021).Based on the information, three kebeles known as Tumata chericha, Kara Sodti and Dobota sample sites were selected purposely on the basis of their similar agricultural practice, high rate of soil erosion problem and active human SWC intervention that adopted improved SWC structures.According to (Wenago wereda forest environmental protection climate change and regulation unitWWFEPCCRU, 2021), the adoption of SWC structures in Wenago district are designed, budgeted and lead by the government regular program and technically supported by CARE Ethiopia (Domestic NGO).In the third step, after securing fresh list of household heads from the three kebeles, households were again stratified in to SWC adopter and non-adopter groups.In the last step, 262 sample household heads (132 SWC adopters and 130 non-adopters) were selected from each house using simple random sampling techniques to represent the study district.

Data analysis methods
Quantitative analysis in this study employed both simple descriptive statistics and econometric models.

Descriptive data analysis methods
Simple descriptive statistics were analyzed using mean, frequency, percentage, graphs, charts, tables and figures.Inferential statistics such as t-test, Chi-square (χ 2 ) test were used in SPSS version 20 software to compare adopters of improved SWC structures and non-adopters in terms of demographic, socio-economic characteristics, institutional, biophysical factors, major crop types, annual crop yields, livestock owing and off-farm activities of households.

Econometric data analysis
With regard to econometric methods of data analysis, parametric and semi-parametric models were used.Propensity Score Matching (PSM), the average treatment effect, and to estimate the logistic regression model, the dependent variable was soil and water conservation (SWC) practice participation status, which takes the value of 1 if a household adopted SWC structures and 0 otherwise, and was used to analyze the impact of adoption of SWC structures on smallholder farmers' livelihood regarding their natural assets, resources and income.In this study, PSM model was used to evaluate the impact of improved SWC practices on the annual income (in birr) of smallholder farmers in the study area.Propensity score matching constructs a statistical comparison group based on a model of the probability of participating in the treatment (adopters), using observed characteristics.Participants were then matched on the basis of this probability, or propensity score, to nonparticipants (nonadopters).
The assumption is that adopter farmers are those who involved in the implementation of SWC structures in their own farm land, whereas non-adopters are those who do not involve in the implementation of SWC structures.The hypothesis is that if SWC structure has a positive impact on livelihood, an increase on crop yield, income and food security should be observed.The average treatment effect of SWC practices was then calculated as the mean difference in outcomes between these two groups (Caliendo & Kopeinig, 2008;Herweg & Ludi, 1999).To understand the potential estimated intervention effect, the treatment impact across different observable characteristics, such as the position in the sample distribution of explanatory variables was examined.ATT or the Treated group was computed as the weighted average of the two groups.The ATT is given by: Where N T is the number of observations treated, Y T is the outcome with treatment, Y C is the outcome without treatment or the control group (nonadopter groups), and Φ ij is the weight factor used in the matching.PSM constructs a statistical comparison group based on a model of the probability of participating in treatment T conditional on observed characteristics X, or the propensity score: P(X) = Pr(T = 1|X) (Chamber and Conway, 1991;Rosenbaum & Rubin, 1983).Therefore, in this study, the PSM method was used to analyze the impact of SWC practices on the income of smallholder farmers.The dependent variables were selected based on how frequently they were used or the number of SWC technologies farmers received on their plots.The covariates include demographic, socio-economic, institutional and environmental characteristics.According to Rosenbaum and Rubin (1983), the bivariate probit model is generally specified as: In the data Y j1 andY j2 are only observable through the two discrete choice responses such that: As the second equation is based on the first question response, the error terms are correlated; hence, the two equations can be estimated jointly using the model developed by Chamber and Conway (1991), which assumes a Bivariate Normal distribution for the two valuations, As mentioned earlier the two questions had four possible pairs of responses: (Y ji , Y j2 ) = (1, 1), (1, 0), (0, 1), and (0, 0) (6) By combining the associated probabilities of all possible responses in the likelihood function, Equation 6 can be estimated using the unrelated bivariate probit model.

Farm land size holding
As shown in Table 1, the survey result showed that 32.82% of non-adopters and 23.28% of adopters owned less than 0.5 hectares of farmland.About 14.12% of the non-adopters and 22.52% of the adopters owned within the ranges of 0.5 and 1.5 hectares of farmland.The results presented in Table 3, show that structure adopters owned a mean of 0.85 hectares while nonadopters hold a mean of 0.70 hectares indicating adopters owned more farm land than those of non-adopters.The result of the t-test [t = −3.00]indicated that there is a statistically significant difference between adopters and non-adopters in terms of mean total land-holding at the 5% probability level (Table 1).The proximity of households to farmland was also assessed, and the results showed that the mean distance travelled by households from home to farm-land was 30 min on average.The result of the t-test [t = −0.52]indicated that statistically there is no significant difference between adopters and non-adopters in terms of mean distance to farm land at the 5% probability level.
The average distance between household respondents' residence and their farm land where structures constructed was measured in minutes and the data was collected from interview and group discussion.Finally, the average minutes was calculated and used in the analysis.The information gathered from KII revealed that the small size of plots or farmlands has also been shrinking year after year due to the rapidly increasing population, high demand for land, deforestation and severe soil erosion problems that are still ongoing.According to the information gathered from the key informant, the redistribution of farmlands had not occurred in the study area for centuries, so the young generations received a very small plot from their parents.

Institutional support services
As shown in Table 2, nearly 94.7% of adopters (users) and only 20.7% of non-adopters (non-users) received training on SWC structures.This means that nearly 80% of non-adopters do not receive training in SWC practices.The result of the chi-square test [X 2 = 22.1] showed that there is a significant difference between adopters and non-adopters in terms of training opportunities on SWC practices at the 5% level of significance, implying that training motivates farmers to invest in conserving natural resources in their locality.As shown in Table 2, 93.9% of SWC adopter and 86.2% of non-adopter households have received credit services, indicating that more adopters use credit services than non-adopters.The result of the chi-square test [X 2 = 4.4] depicted that there was a statistically significant difference between adopters and non-adopters in terms of rural credit services received at the 5% level of significance.This result was in line with the finding of  (Abebe & Bekele, 2014) who stated that there was a significant difference between adopters and non-adopters in terms of credit provision and services in Adama District, Ethiopia.
In this study, extension contact measures the frequency of yearly contact (visits) of development agents (DA's) with smallholder farmers during the survey of harvest year.Thus, extension contact is a way of building the human capital of farmers by exposing them to information on the use of sustainable land management that reduces uncertainty.The results in Table 4 show that the average frequency of extension contact during the cropping season was found to be 48 days per year for structure users and 5 days per year for non-structure user households, showing that almost all nonuser groups have not contacted agricultural agent experts (not received extension services) which might negatively affect agricultural productivity.The results of the chi-square test [X2 = 235] indicated that statistically there was a significant difference between the structure user and non-user groups in terms of extension contact with experts at the 5% significance level.This result was in line with the findings of Chesterman et al. (2019), who stated that there was a significant difference between structure users and non-users in terms of credit provision and extension services in the Ethiopian highlands.
In this study, agricultural inputs refer to fertilizer and selected seed inputs that the local farmers used in their farm fields to maximize their yield.As indicated in Table 2, the results indicated that the majority of users (75.3%) and non-users (68.2%) used fertilizer as their main input for their own farm fields, indicating that users used more fertilizers but not significantly compared to non-users.Thus, the result of the chi-square test [X2 = 2.9] indicated that there was no statistically significant difference between the two groups in terms of the agricultural input used.The respondents reported that they used fertilizers more than others because of the subsiding of organic and mineral fertilizers, as well as improved seeds, which may influence the adoption of SWC technologies directly or indirectly.
The results in Table 2, show that 26.5% and 30.7% of household adopters and non-adopters, respectively, have access to market information.The results of the chi-square test [X2 = 0.58] revealed that there was no statistically significant difference between the two groups in terms of market access at the 5% significance level.This result was contradicted with the finding of (Wolka et al., 2018) who stated that there was a significant difference between adopters and nonadopters in terms of market access in sub-Saharan Africa.

Livestock asset of sample household respondents
Livestock holding is a key indicator of the wealth of local and under-studied communities.
Livestock provides transactions and manure and is a source of cash that can be used to purchase consumption goods and production inputs.As shown in Table 3, for structure users, the mean livestock holding in TLU was 0.17 while it was 0.11 for non-users with a difference of 0.06.This number is relatively large in the structure user category.This may be because structured farmland enhances soil moisture and harvests grass to feed livestock.As indicated in Table 3, the results of the t-test showed that statistically there was a significant difference between structure users and non-users in terms of livestock in TLU, oxen and cow holding.

Main crop types, annual crop production and annual income of household respondents
Table 4, presents the primary crops that serve as a major source of food and income for adopter and non-adopter households in the study area.Based on the statistical results of this analysis, it is possible to generalize that coffee (49.2%), enset (23.3%) and cereal crops (9.5%) followed by other crops are the main sources of food and income of the households in the study area.As shown in Table 4, the study area is an agroforestry-dominated agricultural system, as a result, coffee is the main crop cultivated for commercial purposes, whereas enset is the staple food almost grown for food consumption in the study area.All of these crops are grown during the long rainy season (April to November).
The remaining root crops, such as potatoes, sugar beets, and other spices and sugar cane are also grown on a small scale during the short rainy season (July to September).In the study area, perennial tree crops are widely grown, particularly coffee and enset, followed by mango, avocado, Eucalyptus and other vegetables, such as boyna, potato, godere and cabbages to enhance a household's basic income.According to the survey results, farmers in the study area diversified their income sources by growing different food crops.This result is supported by (Tadesse, 2002) who stated that coffee has been under production for a long time in the study area.The analysis result of the chi-square [X2 = 6.99;P-value = 0.53] showed that there was no statistically significant difference between structure adopters and non-adopters in terms of major crop type production (Table 4).4, the amount of crop produced, consumed, sold and carry-over (kg) for risk time by both adopter and non-adopter sample household farmers in the harvest season of 2021 were quantified.Accordingly, 48.9% of the crops produced went to market sales.This study agrees with the findings of Nkhoma et al. (2017) who reported that minority of cereal crops produced (47.5%) went to sell evidence from Luapula province, Zambia.The share of home consumption in this study was 42.5%.This result contradicts the finding of Adgo et al. (2013) that the majority of crop produced (70%) went to home consumption in Anjenie watershed, Ethiopia.This figure is lower than that reported by the National level-80% (FAO, 1993).Carry-over (300 kg) of adopters was three times greater than that of non-adopters (100 kg) (Table 4).Farmers in the study communities sell a large portion of coffee produced compared to other products, and they might use the cash in turn to buy additional stable food, fertilizers, credit payments and fulfill other social obligation.

As indicated in Table
To analyze the impact and benefit of adopting of physical structures on annual crop yield (kg), a comparison of crop yield before adoption of structures in the harvest season of 2020 and after adoption of structures in the harvest season of 2021 were compared.The results in Table 4, showed that the household's average crop production (kg) before adoption of SWC structures in the harvest season of 2020 was about 1800 kg while the household's crop yield after the adoption of SWC structures in 2021 was 2400 kg with an enhancement of 600 kg which indicates that adoption of introduced structures increased crop yield by 14.28%.
According to the study district (WWFEPCCRU) annual report ( 2021), approximately 7, 15, 15 and 20 km structures were installed in 2018, 2019, 2020 and 2021, respectively, at the study district level.Thus, it is important to consider that even though the comparison is based on the harvest years of 2020 and 2021, the progress is the result of the last four consecutive years of soil and water conservation structures in the study area.
Key interview informants (Figure 2(a)) were asked to reflect on their opinions regarding the main types of agricultural activities and sources of food in their local area.Accordingly, they replied that their main agricultural activity was mixed farming (crop production and livestock keeping).Of the 33 key informants interviewed, 25 (75.75%)underlined that the primary yields (coffee and enset) declined considerably from year to year because of attack by pests and diseases, severe soil erosion, poor soil moisture and weak land management.The majority (72%) of the interviewed households reported that diversification activities were limited by poor access to credit services, fertilizer, late delivery, market access and shortage of capital.
As shown in Figure 2(b), 20 (83%) of the FGD participants explained that agroforestry is their main farming activity, indicating that agroforestry systems are the most dominant agricultural system, meaning that all types of crops are grown together within the same farmland in the study area.According to the information gathered from 33 KII and 24 FGD participants, 46 (80.7%) reported that improved structure adopter households harvested an average of 2000 kg per year in the harvest season of 2018/19, whereas non-adopter households gained an average of 1500 kg of crop yield in the same harvest year, revealing that structure adopter households produce more yield in increments of 500 kg.According to the household respondents, they abled to find wood, grass and medicinal plants around their homes after applying the introduced SWC structures.However, according to the improved structure adopter households point of view, there are some obstacle factors that hinder them during implementation of physical SWC structures such as lack of incentives, shortage of technical skill, small farm land size, unavailability of construction materials and the like.
Household survey was also made on non-adopter households to assess why they do not adopt SWC structures on their own farm land.Of the total 33 interviewed household respondents in this study, 16 of them were non-adopters.Therefore, of the 16 interviewed non-adopters, 16 (100%) of them reported that they did not adopt improved SWC structures on own farm lands.According to their assumption, physical structures occupy part of the farm plot which could be used for production purpose and harbor rodents as well as dangerous wild animals.This indicates that nonadopter households seem to have less awareness about the advantages of improved SWC structures compared to those of adopters.Thus, to increase non-adopter households' awareness, short trainings opportunities, workshops and field demonstration should be arranged by concerned bodies.
During the field observation, the researcher realized that the implementation of SWC structures has been practiced in individual farm plots and communal lands but adopted more in individual farm lands than communal lands.The type of soil and water conservation measures practiced were both traditional and improved structures including stones terraces, soil bunds, wind breaks, ditch-channels, micro basins, grass-strips, contour ploughing and exclosures, etc. with very limited fallowing methods, might be due to shortage of alternative farm lands.

Off-farm activities and income of sample households
Off-farm income is another type of capital for smallholder farmers as they help finance cash deficits by farm households.The household heads of the study area sought off-farm/non-farm /employment, such as engaging in daily labor, handicrafts, selling fuel wood, carpenter, local small-scale trade, etc. when on-farm income or food was inadequate to meet basic household needs.The income obtained from off-farm activities influences the farmers' decision to invest in soil conservation structures on their own farm-lands.As shown in Table 5, 43.9% of adopters and 81.5% of non-adopter households engage in various forms of off-farm activities.This statistical result shows that non-adopter households engage more in off-farm activities than structure adopters/implementers.
According to the information obtained from non-adopter households during the interview and focus group discussion, non-adopter households engage less on-farm work and soil conservation practices by reducing labor availability, whereas they engaged more on off-farm activities because of shortage of farm land and then they engage on off-farm activities as alternative to compensate deficit of on-farm income.The analysis result of the chi-square test [X2 = 39.5]showed that there was a statistically significant difference between adopters and non-adopters in terms of engagement in off-farm activities at a 5% level of significance (Table 5).However, this result was contradicted with the finding of (Belay & Bewket, 2013) who reported that statistically there was no a significant difference between the two groups in terms of mean annual income from off-farm activities.
The survey results also showed that the total annual mean income of non-adopters and adopters from off-farm activities was 2.63 and 1.45, respectively, with a difference of 1.18, indicating that non-adopter households earned more income from off-farm activities than adopters.As shown in Table 5, the result of the t-test [t = 16.8]showed that there was a statistically significant difference between adopters and non-adopters in terms of mean annual income from off-farm activities at a 5% level of significance.This result was in line with the findings of Belay and Bewket (2013) who reported that statistically there was a significant difference between adopters and non-adopters in terms of off-farm activities and mean annual income in northwestern highlands of Ethiopia.However, this result contradicts the findings of Semu (2018) who reported that there was no statistically significant difference between the two groups in terms of mean annual income from off-farm activities in Ethiopia.

Matching SWC adopter and non-adopter households
As shown in Table 6, the estimated propensity scores varied between 0.072 and 0.997 (mean = 0.535) for treatment households, while for the control households, the propensity scores varied between 0.010 and 0.977 (mean = 0.493).This requires the use of a common support region under the distribution of the propensity scores between the two groups.Therefore, in this study, our common support region was between 0.010 and 0.997.Thus, households outside this range were not included in the matching process because of their contribution to the bias in the estimation effect.

Choice of matching algorithm
The nearest neighbour, calliper and kernel matching estimator were used to obtain the best matching estimator for matching the treatment and control households in the common support region.Finally, the selection of the best matching estimator uses different criteria, such as the equal means test referred to as the balancing test, pseudo-R 2 and matched sample size.Table 7, presents the estimated results of the matching quality tests.Based on the selected best estimator (kernel matching with bandwidth of 0.1), the balancing test of covariates, before and after matching households from SWC practices and non-SWC practices areas, shows that there is no statistically significant mean difference between the two groups in terms of all explanatory variables under consideration.
As indicated in Table 8, the balancing tests of covariates were performed using the kernel matching estimator.The results also revealed that out of the 11 explanatory variables tested, six (55%) of them (sex, marriage, farm land size, off-farm activities, extension contact and owning livestock in TLU) were found to be determinant variables that have significant effects on both adopter and non-adopter groups.The remaining five (45%) variables were found to have no significant effect on either the treated or control groups in the study area.In the list of variables (Table 8), FS stands for family size, FLS refers to farm land size and FEX represents farming experience of farmers (in years).
The results presented in Table 9, show low pseudo-R 2 and insignificant likelihood ratio tests, supporting the hypothesis that both the treated and control groups have the same distribution in covariate X after matching.These results clearly show that the matching procedure can balance the characteristics of the treated and matched comparison groups.

Average treatment effect on the treated
As shown in Table 10, the effect of implementing physical structures on households' annual income was analysed.The estimated results provide supportive evidence for the statistically significant impact of SWC structures on a household's annual income.Pre-intervention differences (demographic and socioeconomic characteristics) among the sample households that adopted structures (called adopters or treated groups) on their own farm-land and those who did not adopt structures are known as non-adopters or control groups.In the first step of data collection, households' individual daily and monthly income was recorded and converted to their total annual income, and the total annual income of adopter and non-adopter groups was analysed and compared in Ethiopian birr.
As shown in Table 10, the total annual income of adopter and non-adopter households was about 18,481.48birr (342.24USA dollar) and 5,465.32birr (101.20 USA dollars), respectively, with a difference of 13,016.15birr (241.03USA dollars).The result showed that the total annual income of treated (adopters of SWC structures) is higher than that of non-adopters.This implies that boosting or continuous use of SWC structures is preferable to increase agricultural productivity and income and improve smallholder farmers' livelihoods in the study area.As shown in Table 10, the results of the t-test [t = 9.31] showed that statistically there is a significant difference between structure adopters and non-adopters in terms of mean annual income from on-farm activities at 1% level of significance.

Sensitivity analysis
The final step of the PSM analysis is to test the robustness of the estimated results to possible Conditional Independence Assumption (CIA) failures.Matching estimators work under the assumption that there is no convincing source of exogenous variation in treatment assignment does not exist.Likewise, a sensitivity analysis was undertaken to detect whether the identification of the conditional independence assumption to be satisfactory or affected by the dummy confounder or the estimated ATT (adopter groups) is robust to specific failure of the Conditional Independence Assumption (CIA).Table 11, shows the sensitivity results of the outcome ATT values to the dummy confounder.The comparison between the simulated and baseline estimates provides an idea of the robustness of ATT estimation results from possible failures of the CIA.Sensitivity analysis was performed on the outcome variables of annual income and consumption expenditure using kernel matching.This study revealed that the simulated ATT of the outcome variables was very close to the baseline ATT.This implies that it is only when U is simulated that it provides an implausibly large outcome effect, that is, estimates are almost free from unobserved covariates.Consequently, it can be concluded that overall, the results are remarkably robust, the two outcome variables are insensitive to unobservable variables and the analysis supports the robustness of the matching estimate.

Conclusion
This article assessed the impact of soil and water conservation structures on smallholder farmers' livelihood.Data were collected from 262 household head farmers in Wenago district, southern Ethiopia using a random sampled survey in 2020/21.The common types of soil and water conservation practices in the study area were both traditional and improved structures.The chi-square test also revealed that structured user households are more beneficiaries of institutional support services.The t-test results show that the mean land holding of adopters is slightly greater than that of non-adopter households.The statistical analysis results revealed that the crop yield of structure user farmers was increased by 14.28%.Statistically, there was a significant difference between adopters and nonadopters in terms of farming experience, farm-land holding, training, credit use, extension contact, livestock owing and engagement in off-farm activities.Of the 11 explanatory variables tested, six (55%) were found to be determinant factors that had a significant effect on the adoption of soil and water conservation structures.The estimated propensity score matching result indicated that the distribution of structure user farmers (0.535) varied from that of non-users (0.493), indicating that the two groups have no similarity.The simple descriptive statistics result showed that the total annual income of adopter and non-adopter households was about 18,481.48birr (342.24USA dollar) and 5,465.32birr (101.20 USA dollars), respectively, with a difference of 13,016.15birr (241.03USA dollars).The average treatment effect of the model shows that the annual income of structure user farmers is more than two-fold compared to that of non-users, which implies that the crop yield, income and livelihood status of structure users might be better than that of non-users in the study area.The results shown in this article suggest the significance and positive impact of adopting structures in farmers' farm land.The implication of these findings is that smallholder farmers' access to institutional support services such as credit use, fertilizer, technical skill training, market and education should be improved and promoted in the study area in particular and Ethiopia in general.

Figure 1 .
Figure 1.Location of the study area.
production and consumption of the major crop produced(Kg) in the sample study (during the harvest season of 2021) of adoption of SWC structures on crop yield by comparing the crop yield harvested before and after the use of

Figure
Figure 2. Interview with SWC adopter and non-adopter key informants held in farmer's individual house compound (a), focus group discussion with selected groups near agroforestry land (b).(photograph: March, 2022).

Table 1 . Farm land size holding of sample households by household type
Source: Field Survey Result (2021).Note: ** indicate significance; NS shows no significant difference at 5% significance.

Table 7 . Comparison of the three matching estimators by performance criteria Matching Estimator Performance Criteria Balancing test (n) Pseudo-R 2 Matched sample size Mean SB
Source: Field Survey Results (2021).Note: n, Number of explanatory variables with no statistically significant mean differences between the matched groups of SWC practice households and non-SWC practice households.NN = nearest neighbor, KM = kernel matching.

Table 11 . Results of simulation-based sensitivity analysis (kernel matching) Outcome variable
Source: Field Survey Results (2021).