• Tidak ada hasil yang ditemukan

The effects of continuous sustainable land management practices on agricultural land productivity in Central Ethiopia

N/A
N/A
Protected

Academic year: 2023

Membagikan "The effects of continuous sustainable land management practices on agricultural land productivity in Central Ethiopia"

Copied!
15
0
0

Teks penuh

(1)

Volume 10, Number 3 (April 2023):4389-4403, doi:10.15243/jdmlm.2023.103.4389 ISSN: 2339-076X (p); 2502-2458 (e), www.jdmlm.ub.ac.id

Open Access 4389 Research Article

The effects of continuous sustainable land management practices on agricultural land productivity in Central Ethiopia

Taye Alemu1*, Degefa Tolossa2, Feyera Senbeta3, Tesfaye Zeleke3

1 Department of Development Economics and Management, Ethiopian Civil Service University, Addis Ababa, Ethiopia

2 Center of Rural Development Studies, Addis Ababa University, Addis Ababa, Ethiopia

3 Center of Environment and Development Studies, Addis Ababa University, Addis Ababa, Ethiopia

*corresponding author: [email protected]

Abstract Article history:

Received 10 September 2022 Accepted 3 January 2023 Published 1 April 2023

The Ethiopian government has exerted efforts to rehabilitate degraded agricultural lands using a range of sustainable land management (SLM) initiatives to enhance agricultural productivity. One of the key components was improved structural soil and water conservation (SWC) technologies.

This study examines the effects of continuous SLM practices on agricultural land productivity, with particular emphasis on SWC technology adoption in Central Ethiopia. The analysis was based on the data collected from 525 sample household surveys in two districts, namely Kewet and Sebeta-hawas. A propensity score matching (PSM) model was used to investigate the effects on treated and non-treated plots. The study findings revealed a substantial and positive effect on treated agricultural plots compared to non-treated ones in the Kewet district. At the same time, the difference was found positive but not significant in the Sebeta-hawas counterpart. The insignificant impacts are justified because SWC efforts focused on constructing structures rather than tailoring them with soil- replenishment and productivity enhancement functions. The important conclusion is that continuous SWC adoption efforts positively impact agricultural productivity; however, its effect is more noticeable when SWC structures are integrated with productivity enhancement functions and applied in low moisture areas. Thus, policymakers and project planners should consider the role of integrating physical SWC structures with soil replenishment and agronomic activities.

Keywords:

continuous adoption land productivity plots

propensity score matching SWC measures

To cite this article: Alemu, T., Tolossa, D., Senbeta, F. and Zeleke, T. 2023. The effects of continuous sustainable land management practices on agricultural land productivity in Central Ethiopia. Journal of Degraded and Mining Lands Management 10(3):4389-4403, doi:10.15243/jdmlm.2023.103.4389.

Introduction

The agriculture sector has played and continues to play a leading role in economic growth, poverty reduction, and development. It provides essential ingredients for humanity, such as raw materials for industrial development, income and employment for the rural population, and a source of food for an inevitably growing population (Khanna and Solanki, 2014). In many third-world countries, particularly in Sub- Saharan Africa, the agricultural sector is still

considered an engine of economic growth (Danso- Abbeam et al., 2018). In Ethiopia, agriculture provides employment opportunities for more than 80% of the population, contributes about 50% to the national Gross Domestic Product (GDP), and accounts for 84%

of all exported goods (Tilahun and Belay, 2019;

Wordofa et al., 2020). It is not just a source of economic activity but also the foundation upon which society's welfare has been built (Holmatov et al., 2017). Ensuring sustainable development, among other things, requires increasing agricultural

(2)

Open Access 4390 productivity (Danso-Abbeam et al., 2018).

Nevertheless, land degradation has been continuously threatening the productivity of agricultural land (Belayneh et al., 2019; Moges and Bhat, 2020; Masha et al., 2021).

The severity of the problem is high in developing countries where natural resources are the primary source of livelihood for most people (Masha et al., 2021; Jiru and Wegari, 2022). Land degradation due to soil erosion is the leading factor for poor agricultural development in many developing countries (Gemechu, 2016). Worldwide, nearly 40% of arable land has been degraded due to soil erosion and continues to be lost at the rate of 5 to 10 million hectares annually (Asfaw and Neka, 2017; Limani, 2018). Sub-Saharan Africa is one of the most eroded regions in the world, with nearly 65% of its agricultural land affected in the last four decades (Mehretie and Woldeamlak, 2013).

Like elsewhere, cropland degradation in Ethiopia is continuously undermining crop productivity (Erkossa et al., 2018; Mekuriaw et al., 2018; Debie et al., 2022). If immediate action is not taken, it will continue to be a problem for future subsistence agriculture (Wordofa et al., 2020). Despite varying estimates of the extent of erosion, various studies have illustrated the severity of the problem. For instance, almost half of the country’s highland areas have faced considerable soil erosion (Miheretu and Yimer, 2017;

Abiy, 2022), and 66% of the cultivated land areas highly eroded (Tesfaye et al., 2014), with 25%

severely and 4% extremely eroded beyond recovery (Asfaw and Neka, 2017). Agricultural productivity has been declining by 2.2 % per year (Abiy, 2022). Studies conducted in central Ethiopia (e.g., Mengistu and Mekuriaw, 2015; Estifanos et al., 2022), where this study conducted revealed that soil erosion has been the primary cause of land degradation. Deforestation, overgrazing, and continuous cultivation coupled with the steep slope nature of the cultivated land area aggravated soil erosion problems in the area (Mengistu and Mekuriaw, 2015; Bekele et al., 2022; Biratu et al., 2022). As a result, agricultural production is steadily dropping while the need to increase food production to satisfy the growing human population and their needs has become more desperate than ever. This has put smallholder farmers in a vicious circle of poverty and Environmental degradation (Menberu, 2014).

Since the 1970s, to address the land degradation problem and to enhance crop productivity, different sustainable land management (SLM) practices have been adopted and implemented in many parts of Ethiopia (Limani, 2018; Tesfaye et al., 2018; Ewunetu et al., 2021; Masha et al., 2021). In recent decades, participatory watershed management has gained recognition as part of the national development strategy (Agidew and Singh, 2018). This framework has triggered the launching of SLM projects alongside other initiatives in selected areas of the country. The first part of the project lasted between 2009 and 2013,

while the second phase began in 2014 and ended in 2019. One of the program's key project components was rehabilitating degraded lands using structural soil and water conservation (SWC) technologies such as bench-terrace, stone-bound, soil-bound, cut-off drains, and area closures (Schmidt and Tadesse, 2017; Sileshi et al., 2019). The World Bank and other development partners contributed to the program's financing and implementation. Furthermore, the government instituted a national SWC construction campaign in 2011 to mobilize the community to build the necessary structures following watershed development principles (Mekuriaw et al., 2018). The expected benefits behind the introduction of projects have been to reduce environmental degradation, poverty, and renovate land productivity (Ararso et al., 2016; Atikilt et al., 2020; Masha et al., 2021).

Despite these efforts, however, several studies conducted in the country have shown mixed results on SWC technology adoption and its impacts on land productivity (Masha et al., 2021; Jiru and Wegari, 2022). For example, studies conducted by Erkossa et al. (2018) in Western Ethiopia, Atikilt et al. (2020) in Northern Ethiopia, and Masha et al. (2021) in Southern Ethiopia showed that plots with SWC measures such as soil-bound and stone-bound produced higher yields than those without. However, studies conducted by Adimassu et al. (2014) in central Ethiopia and Abera et al. (2020) in the highlands of Ethiopia; demonstrated no significant productivity differences between conserved and non-conserved plots. The review and synthesis conducted by Adimassu et al. (2017) and Guadie et al. (2020) in Ethiopia stated that soil bunds and stone bunds negatively impact crop yields because the built structures reduce the size of the cultivatable land. According to the studies of Haregeweyn et al.

(2017) and Masha et al. (2021), the effects of SWC on crop production vary based on agroecosystems and site-specific factors. Furthermore, despite efforts being made to introduce SWC technologies, the land productivity impacts of continuous adoption among adaptors were not well investigated (Teshome et al., 2016). Most past assessments were made by considering all adopters to be at the same level without differentiating between initial and continuous adoption. According to Nigatu et al. (2017), adopting conservation measures may not improve production unless technologies are utilized continuously for years.

The potential impacts of SWC practices on yield outcomes could be the result of long-time experience passed through the maintenance of structures (Schmidt and Tadesse, 2017; Tanto and Laekemariam, 2019;

Tilahun and Belay, 2019).

Based on the preceding discussions, it is apparent that there are critical but unexplored research and policy gaps that need to be understood. This study sought to contribute to two knowledge gaps identified in previous studies. First, the analysis focused on the effects of continuous adoption decisions of households

(3)

Open Access 4391 on agricultural land productivity. The prior adoption

studies overlooked the impacts of long-term adoption efforts on agricultural productivity, considering adoptions to be at the same stage. In this paper, continuous adoption refers to farmers’ self-motivated effort to continuously maintain already created structures for a long period and replicate them to other owned plots after project interventions. Second, the study was conducted on areas supposed to represent a variety of socio-economic and agroecological contexts. Most previous studies were conducted at specific sites, and findings have not produced consistent results. Therefore, impact assessments based on different agroecosystems and locations are essential for understanding the impacts and sustainability of the program initiatives. Thus, the principal objective of this study was to examine the

effects of continuous SLM technology adoptions on land productivity, with particular emphasis on the improved SWC technology adoptions in central Ethiopia.

Materials and Methods Description of the study area

This study was carried out in two designated districts in central Ethiopia, namely Kewet and Sebeta-hawas (Figure 1). Kewet district is in the Amhara Region's northern Shoa Zone, whereas Sebeta-hawas is in the Oromiya Region's southwestern Shoa Zone. These are some of the areas chosen to be addressed in the first phase of the sustainable land management program carried out between 2009 and 2013.

Figure1. Map of the study area.

The Ethiopian government, in collaboration with the World Bank and other development partners, implemented SWC technology as a pilot project in these areas. They also have different agroecological zones than the other central regions of the country participating in the program. Kewet, for example, is a lowland, semi-arid, and low-potential area (Mekuriaw et al., 2018), whereas Sebeta-hawas is a midland, highland, and high-potential area. Furthermore, because most studies tend to focus on the country's

northern highlands, these areas went unnoticed by most researchers for scientific analysis (Masha et al., 2021). Because of their differences, the two study areas can be considered representative of central Ethiopia.

Kewet district: one of the two study areas chosen for this study is the Kewet district, which is located 225 kilometers north of Addis Ababa. The main study area was the Robit watershed located within the districts that include five kebeles (the

(4)

Open Access 4392 smallest and the lowest administrative units of

government next to districts in Ethiopia), namely Ayaber, Alolo, Mengist, Debir, and Abomsana. The Kebeles were those included in the first phases of the sustainable land management program. This research area is located at an elevation range of between 1,001 to 2500 meters above sea level (masl) on average. The location extends from 9°50'0" N to 10°10'0" N latitude and 39°50'0" E to 40°0'0" E longitude (Figure 1). The average annual rainfall is 916 mm, with temperatures ranging from 16 to 31 °C. The vast majority of people in this region are rural smallholders who rely on mixed crop-livestock farming activities (Tesfay et al., 2022).

Based Ethiopian Central Statistical Service (CSS) estimation in 2022, the district has a total population of 160,500, of whom 82,700 were men and 77,800 were women (CSS, 2022). The area's dominant crops are sorghum, teff, wheat, and maize. Soil erosion resulted from rapid population growth, and agricultural expansion severely threatened agricultural production in the district (Asefa and Mindahun, 2019).

Various SWC measures have been implemented as part of a sustainable land management program in this area since 2009 (Word Bank, 2018).

Sebeta-hawas district: Sebeta-hawas is also one of the other districts identified for this study. It is 45 km from Addis Ababa, in the country's southwestern part. The main study area was the Atebela watershed located within the districts that include six peasant associations (Kebeles), namely Haro Jila, Bole, Mogle, Korke, Koche, and Jamo. The selected kebeles are those identified for intervention in the first phase of the sustainable land management program. This study site is a highland and humid area with an altitude range of 2,001-4,455 masl and is located between 8°40'0"N to 9°0'0"N latitude and 38°30'0"E to 38°40'0"E longitude (Figure 1). The land feature of Sebeta-hawas is characterized by mountains, hills, and marshy plains and is surrounded by the "Awash" watershed in the west (Belay and Assefa, 2021). The average annual rainfall is 1055 mm. The area lies in the temperate climatic zone, with a temperature range of 13.9 to 25.4

°C (Mekonen et al., 2015). Based on CSS's projection, the district's total population was 189,912, of whom 97,150 were men and 92,762 were women (CSS, 2022). Wheat, teff, barley, and beans are the dominant crops grown in the area. The agricultural systems in these watersheds are small-scale subsistence crop- livestock mixed farming systems. Soil erosion and soil nutrient depletion severely threatened agricultural production in the district. Various SWC measures have been implemented as part of this area's sustainable land management program since 2009 (Word Bank, 2018).

Data sources and methods

Household and plot-level data were gathered from primary and secondary sources. Farm households supplied primary information through a survey conducted between November 2019 and January 2020.

The secondary data was obtained from the agricultural office's annual plans and reports. These sources were used to identify the number and location of target respondents. A set of standardized closed and open- ended questionnaires and key informant interviews were employed to collect primary data from the respondents. The questionnaire used various questions to acquire information on demographic, social, institutional, and economic variables of households and farmland characteristics. Before the primary questionnaire survey was distributed, a pilot pre-test was conducted on a randomly selected group of 12 non-sampled respondents. Various studies recommend a sample size of 10-50 units. This study used the 12 minimum subjects suggested by Whitehead et al.

(2015) to save time and cost. Modifications have been made following the pilot survey. Then, enumerators were selected to carry out the survey. Key informant interviews have also been conducted to collect information from extension workers and district-level SLM project officials. The data was employed to supplement the interpretation of the how and the why part of the quantitative findings.

Sample size and sampling techniques

The target population of this study was farmers who partook in the first phase of the SLM program, initiated by the Ethiopian government in collaboration with the World Bank. Particularly those who adopted one or more of improved structural SWC practices on their own farmland between 2009 and 2013 were considered. The population size was estimated to be 1,557 improved SWC technology users in the two districts; Sebeta-hawas (665) and Kewet (892). The total sample size representing the study population was determined to be 525 units (276 from Kewet and 249 from Sebeta-hawas). The study applied a simplified formula developed by Yemane (1967) and recently used by various researchers, including Byamukama et al. (2019), Wordofa et al. (2020), Nkonki-Mandleni et al. (2022), and Tekle and Gemechu (2022) to determine the required sample. Accordingly, for a larger population whose size is known, sample size can be determined by:

N

1 + N(e ) (1) where N = population size, n = sample size, and e = the margin of error (determined to be 5%). Assume that n1

is the sample size from Sebeta-hawas and n2 is from the Kewet study population. Then, at a 5% significance level, n1 = (665/1+ 665 (0.05)2 is proximately equal to 249 units, and n2 = (892/1+ 892 (0.05)2 is equal to 276.

Accordingly, the total sample size has been increased to 525 (249 + 276). The sample sizes were separately treated to have representative samples from each study area to make statistical comparisons reliable. The required sample size was drawn from the population using a multi-stage sampling procedure. In the first

(5)

Open Access 4393 stage, study districts (Sebeta-hawas and Kewet) were

purposively selected because these were among those areas selected for pilot SLM program intervention in Central Ethiopia between 2009 and 2013. In the second stage, agricultural offices in the respective districts were contacted, and Kebeles that participated in the program were selected. Then, local extension workers supplied lists of those households included in the first phases of the SLM program. In the third stage, proportional households were taken from each Kebeles using a simple random sampling technique to select 525 total sample households. From the total sample, 177 were continuous adopters (treatment groups), while the remaining 334 were non-continuous adopters (control group). Continuous adopters are those households who continuously maintained the accepted SWC measures for at least five years and applied them to more than 50% of slopping farmlands on their motivation (adapted in part from Teshome et al., 2016).

Finally, treated plots (hereafter, adopters) operated by continuous adopters groups were taken as treatment groups. On the other hand, non-treated plots (hereafter non-adopters) from non-continuous adopters were taken as a control group. The two groups are similar regarding land degradation history, cropping system, soil type, and topography.

Variables selection and description

The outcome variable was agricultural land productivity, measured by net production values of major crop yields per hectare grown in the study areas.

Instead of using crop-specific production functions, the study used aggregate crop production using the 2019 average market price in Ethiopian Birr (ETB), the official currency of Ethiopia. To consider input use variations in the analysis, all input costs, including labor, fertilizer, seed, and agricultural chemicals, were reduced from the total value of crops. The description of explanatory variables is presented in Table 1.

Table 1. Description of explanatory variables.

Variables Variable description

Demographic factors

Sex The sex of the household head; 1 if male, 0 otherwise.

Age The age of the household head in years.

Grade The household head’s highest grade completed in years.

Family size The number of people living in the household.

Full-time labor size The number of people working full-time in agriculture.

Institutional factors

Tenure security Feeling of tenure security; 1 if feeling secure, 0 otherwise.

Assistance Availability of assistance programs for SWC measures in the area; 1 if available, 0 otherwise.

Access to DAs Contacts with development agents; 1 if yes, 0 otherwise.

Credit access Access to credit services; 1 if yes, 0 otherwise.

Social factors

Social group Membership to local institutions; 1 if a farmer is a member and 0 otherwise.

Erosion problem

recognition Farmer‘s recognition of soil erosion as a problem; 1 if perceived, 0 otherwise.

SWC profitability Farmer‘s attitude to the profitability of SWC; 1 if profitability is perceived, 0 otherwise.

Labor sharing Farmer‘s participation in labor sharing activities; 1 if participated, 0 otherwise.

Economic factors

Access to market Distance to the nearest market area in kilometers.

Off-farm work Farmer‘s engagement in off-farm activities; 1 if yes, 0 otherwise.

Livestock holding Number of livestock measured by Tropical Livestock Units (TLU).

Plot related factors

Farm size Size of total cultivated land in hectares.

Land

Fragmentation The number of plots divided by the total farmland size/hectare.

Gentle slope Gentle slope; 1 if the average slope plots are gentle, 0 otherwise.

High fertility High soil fertility; 1 if the average plots soil fertility is high, 0 otherwise.

Distance to plots Average distance in kilometers from a residential area to farmlands.

Analytical model specification

The study's principal objective was to examine the effects of continuous SLM technology adoption on agricultural land productivity, focusing on improved

SWC technology adoption. Thus, comparing the land productivity status of SWC-treated plots (adopters) and non-treated plots (non-adopters) is indispensable.

However, simple comparisons of mean differences in productivity on plots with and without using particular

(6)

Open Access 4394 SWC practices probably result in biased estimates.

That is because the independent variable could be affected due to various observable and unobservable characteristics of households and plots other than particular land management practices. Thus, estimating the effects of continuous SWC practices on productivity gains requires creating comparable observations of the treatment and control groups regarding their attributes. This study used a propensity score matching (PSM) model to overcome the econometric challenges and ensure robust results.

According to Heinrich et al. (2010), the PSM framework is a non-experimental approach that involves constructing a statistical comparison group based on observed characteristics. It is considered a popular approach to estimating causal treatment effects and the second-best alternative for minimizing selection bias next to true experimental design in program evaluation (Baker and Ichino, 2002). To use the model, this study first estimated the propensity scores of continuously treated and non-SWC-treated plots of households using a binary logistic regression model. This model estimates the association between a set of predictors and a categorical binary outcome variable (Tabachnick and Fidell, 2019). Since the dependent variable, SWC adoption status, for this study is dichotomous; it takes the value of 1 if the household is a continuous adopter (long-term treated plots) and 0 if the household is a non-adopter (non- treated plots) with SWC measures at all. The probability that a household continuously adopts SWC measures (Pi) is expressed in the binary logit form as:

Pi = e

1 + e (2) The probability that a household belongs to non-

continuous adopters group (1-Pi) is given by:

1 − Pi = 1

1 + 𝑒 (3) The odd ratio is given by:

Pi

1 − Pi = 1 + e

1 + e = e (4) Finally, the logit model is given by:

Y = ln Pi

1 − Pi = α + β1x1 + β2x2 + β3x3 + ⋯ βnxp + Ui (5) where "Y" is a dummy dependent variable indicating whether a plot received continuous treatment or not, pi is the probability of continuously adopting SWC measures or the probability of the plot receiving continuous treatment, α is intercept; β1, β2…, βn are coefficients or parameters to be estimated; x1, x2..., xp are observed explanatory variables; Ui is a disturbance term.

After matching, the average treatment effect on the treated (ATT) was estimated to assess the impact of conservation measures. Accordingly, the effects of continuous adoption of SWC measures on a given outcome variable (Y) are estimated by:

ATTi = Yi (Di = 1)– Yi (Di = 0) (6) where ATTi represents the average treatment effects of continuous SWC adoption, Yi represents the outcome variable (Agricultural Land Productivity), and Di represents whether the farm plots receive SWC technology treatment (D=1) or not (D=0). The fundamental problem with evaluating individual-level treatment effects is that it is impossible to simultaneously observe the two potential outcomes for the same individual. Either Yi (Di =1) or Yi (Di =0) are observed, but both can never be observed for the same households simultaneously. This leads to a missing data problem. As a result, estimating individual-level treatment effects is impossible (Smith and Todd, 2005). To estimate ATT based on equation (5), Yi (Di = 0) is the unobserved counterfactual outcome, i.e., an estimated outcome parameter indicating what land productivity would have been in the absence of the treatment for the treated. Thus, two basic assumptions are required to use the observed outcome of the control group in estimating the counterfactual outcome (Heinrich et al., 2010). The first is referred to as the "conditional independence assumption. According to this assumption, potential outcomes of adopters and non-adopters are independent of the treatment status after observable covariates (X) are controlled. If treatment (Y1) and control (Y0) potential outcomes are independent of treatment allocation conditional on X, then treatment conditional on the propensity score P(X) is also independent (Smith and Todd, 2005). It is given as follows:

P(D = 1|X) = P (X) (7) The second one is a common support condition, which assumes the existence of sufficient overlaps in observable characteristics of the treated and control so that households being compared have a common probability such that:

0 < p(X) < 1 (8) If the two assumptions are satisfied and p(X) is known, then the PSM treatment for average treatment effect on the treated (ATT) conditional on the propensity score can be written as:

ATT = E(Y1|D = 1) − (Y0|D = 1) (9) ATT = E(Y1|D = 1, P(X)) − Y0 D = 1, P(X) (10) where; ATT is the average treatment effect on treated, P(X) is the propensity score calculated on the set of covariates (X), Y1 and Y0 are the potential outcomes

(7)

Open Access 4395 in the treatment and non-treatment groups, and D (0,1)

is the indicator of treatment exposure. This means the PSM estimator is simply the mean difference in outcomes over the common support region, appropriately weighted by the propensity score distribution of treated participants.

In addition, various tests used to satisfy PSM estimation assumptions were performed. These include algorithm selection tests, tests for balancing conditions, common support conditions, and sensitivity analysis. Different matching algorithms were tried to match the treatment with control households having similar propensity scores. It was suggested that trying as many matching methods as possible and checking the overall convergence of results are essential to find out matching pairs (Khandker, 2010). The most frequently used matching methods in PSM are the nearest neighbor, kernel, and radius matching (Thoemmes and Kim, 2011) were employed in this study. Various criteria, such as caliper distance, bandwidth, and the number of nearest neighbors, were employed to determine the best matching method. The matching algorithm selection was guided by factors such as Pseudo-R2, equal mean test, mean bias test, and the sizes of the matched sample. Deheja and Wahba (2002) suggested that using a matching algorithm with a large matched sample size and a low pseudo R square value balances all explanatory variable groups. Regarding the test of balancing condition, this study implemented before and after matching mean comparisons, standardized bias, and overall measures of covariate imbalance to check if there is still a difference between the two groups after conditioning on the propensity score. The goal of testing a common support condition is to check if there is a shared set of characteristics between the treatment and control groups (Caliendo and Kopeinig, 2008). The estimated average adoption effects (ATT) sensitivity to hidden bias was also checked using the Rosenbaum bounds sensitivity approach.

Methods of data analysis

The collected data were analyzed using Stata software version 17. Descriptive statistics such as frequencies and mean; inferential statistics including non- parametric statistics (Chi-square test), parametric test (T-test), and likelihood-ratio (Chow tests) were computed. A likelihood-ratio Chow test was conducted to determine whether the regression models in the two study areas (Sebeta-hawas and Kewet) were statistically different. The test result was first obtained by fitting the logistic regression model for each study area sample. Then the result was nested and compared with the combined (pooled) sample model. A binary logit regression model was used to compute the PMS of the treated and controlled plots. All estimations in the model were performed using Stata psmatch2. The bootstrap method with 100 replications has been used to estimate standard errors and to test the statistical

significance of the treatment effect. Statistical analysis outputs were presented in table and figure format.

Results and Discussion

This study aimed to examine the effects of continuous SWC adoption measures on agricultural land productivity. To this end, those farmers in the selected study area who continuously implemented, maintained, and applied one or more improved SWC measures to their sloppy plots were considered adopters. The measures were soil-bound, stone-bound, bench terraces, and cut-off drains mainly applied in the study areas. For two important reasons, all of these clusters of technologies have been considered instead of just one. First, in most cases, a particular technology is selected based on material availability and the conditions of the resources on the ground. For example, stone-bound is used in areas where the stone is available. Where there are no stones, it is soil-bound.

Thus, a farmer can use more than one measure depending on the situation in a single plot. In other words, technology adoption is not mutually exclusive or independent of each other (Amare et al., 2014;

Mengistu and Asefa, 2019). Second, there are complementarities among SWC measures as far as factors affecting adoption statuses are concerned (Sileshi et al., 2019; Ewnetu et al., 2021). The following section provides a descriptive analysis of variables and propensity score matching (PSM) estimation results.

Results of descriptive statistics Agricultural land productivity

Agricultural land productivity in this paper was measured by the net value of output divided by land size. The quantity of production of crops from each plot was converted in terms of ETB per quintal, and the different crops were aggregated into an output index. Crop production is the major source of households' livelihoods in the study areas. The major crops grown in the study regions include wheat, teff, barley, beans, and sorghum in their order of importance. Crop revenue was used because the data was collected from different crops grown in the study areas. In order to consider input use variations in the empirical analysis, the costs of inputs were subtracted from the total value of crop production. Intermediate input costs such as chemical fertilizer, seed costs, agricultural chemicals (pesticide/ herbicide), and labor costs were measured in birr and aggregated into an input index. Finally, the net output value (the difference between the value of the crop and input values) was divided into the total plot size considered in the study to determine agricultural productivity.

Because the purpose of this section is to compute the propensity scores that will be used in the matching process, later on, it will not go into detail about why

(8)

Open Access 4396 and how each of the covariates affected households'

participation in the intervention.

A simple mean comparison test before matching, presented in Table 2, shows that the average value of crop production per hectare was 14,896 ETB ( 530 USD) on SWC conserved plots. In comparison, it was 15,680 ETB (558 USD) per hectare for non-conserved plots in Sebeta-hawas. Nevertheless, there was no

statistically significant difference at a 5% level between SWC-treated and non-treated plots regarding agricultural productivity. By contrast, the average value of agricultural land productivity was substantially higher on plots where structural SWC measures were applied, 22,323 ETB (794 USD) per hectare, than where measures were not applied, 19,347 ETB (689 USD) per hectare, in the Kewet district.

Table 2. Descriptive statistics of variables used in propensity score matching models.

Variables Sebeta-hawas

T/(χ2) Kewet

T/(χ2) Treated Untreated Treated Untreated

Outcome variable

Net land productivity 14,896 15,680 -0.68 22,323 19,347 3.18***

Explanatory variables

Sex (1 = male, 0 = female) 0.87 0.86 0.25 0.91 0.90 0.06

Age 45.83 46.13 -0.26 42.7 49.3 -5.79***

Grade 3.44 3.37 0.16 2.99 2.65 1.17

Family size 6.19 4.78 6.06*** 5.48 5.20 1.42

Full-time labor size 3.41 1.88 11.21*** 2.77 1.89 9.89***

Tenure security 0.90 0.75 2.59** 0.97 0.96 0.74

Assistance 0.52 0.78 -3.99*** 0.42 0.68 -4.4***

Access to DAs 0.98 0.92 1.72 0.93 0.94 -0.2

Credit access 0.57 0.39 2.58** 0.5 0.51 -0.2

Access to market/KM 7.94 4.59 8.03*** 12.63 7.58 9.09***

Off-farm work 0.08 0.54 -6.93*** 0.08 0.25 -3.67***

Livestock holding (in TLU) 4.41 2.72 9.60*** 3.4 2.79 5.48***

Shared labor 0.88 0.69 3.19** 0.92 0.75 3.78***

Soil erosion problem recognition 0.92 0.81 2.04** 0.91 0.88 0.94

SWC profitability 0.94 0.79 2.68** 0.90 0.77 2.99**

Social group 0.87 0.58 4.36*** 0.56 0.44 1.92*

Farm size/ha 1.39 0.84 8.13*** 1.23 0.63 12.88**

Land Fragmentation 2.06 2.93 -4.75 2.10 3.25 -7.77***

Gentle slope 0.06 0.28 -3.63*** 0.03 0.30 -6.03***

low fertility 0.78 0.31 7.16*** 0.67 0.39 4.71

Distance to plots/Km 1.24 1.21 0.40 0.81 0.87 -0.91

A t-test and chi-square test was performed to determine if the sample means between adopters and non-adopters groups are statistically significant; ∗p<0.05; ∗∗ p<0.01; ∗∗∗p<0.001.

Furthermore, it has been shown that there was a statistically significant difference between plots that received continuous treatment and plots that did not.

However, one must find out that such productivity differences could be due to the application of SWC practices, as the control groups are not restricted to treatment groups having similar characteristics.

Hence, further econometric analysis using the propensity score matching procedure is required to control other observable factors affecting the outcome variable.

Observable factors influencing treatment status One of the explanatory variables influencing the treatment status is the demographic characteristics of households. The most critical factors influencing SWC technology adoption include sex, age, education, family size, and full-time labor. Based on the finding presented in Table 2, in the Kewet district, statistically

significant differences in age and full-time labor were observed between adopters and non-adopters. More specifically, the average age of adopters was around 42 years, and full-time labor was 2.77 units, whereas non- adopters had a mean age of 49 and 1.89 units of the labor force.

These observations imply that the farm households that continuously used improved SWC technologies were relatively younger and had a higher number of full-time workers than those that did not.

This suggests that the likelihood of younger farmers is higher in continuously implementing SWC measures, and household labor size is an essential factor influencing adoption continuity. Similarly, in the Sebeta-hawas, continuous adopters had larger families and labor sizes than the control groups. The test statistics confirmed the differences at less than a 5%

significance level. Adopters had an average family size of 6.19 and a labor size of 3.41, while non-adopters

(9)

Open Access 4397 had 4.78 family sizes and 1.88 labor sizes. This implies

that households with more family sizes are likely to apply the SWC measures for the long term as it requires labor for structure maintenance and investment (Wolka and Negash, 2014). In both study regions, however, there are no significant dissimilarities between the two groups in terms of the gender and educational level of the household heads.

This implies that the impacts of SWC sustainable adoption on agricultural land productivity will not result from such variables.

The institutional variables selected in this study include access to tenure security, assistance programs, access to extension services, and access to credits.

Table 2 shows that the proportion of adopters who had access to assistance programs was lower than non- adopters, and the difference was statistically significant in both study areas. Adopters and non- adopters are statistically dissimilar in terms of tenure security, assistance programs, and access to credit in Sebeta-hawas. The findings further show that 90%, 52%, and 57% of SWC technology adopters and 75%, 78%, and 39% of non-adopters had a strong feeling of tenure security, had received government and Non- Government Organization (NGO) subsidies for their SWC efforts, and had access to credit services, respectively (Table 2). This implies that adopters in this area had a higher feeling of tenure security and lower access to subsidies and access to credit services compared to non-adopters. In the Kewet district, adopters and non-adopters are different regarding access to government incentives or subsidies.

However, there is no statistical difference between the two groups in terms of access to extension agents, tenure security, and credit access. This implies that the productivity impacts of SWC adoption efforts on land productivity cannot vary because of this variable.

The economic variables considered in this study were access to markets, off-farm income, and livestock holding. According to the results depicted in Table 2, statistically, significant dissimilarities were observed between the users and non-users in all economic variables. These results indicate that farm households with more livestock, low market access, and low off- farm opportunities had a better chance of continuously using the introduced SWC practices because such households are better at taking risks associated with new technologies. In both study regions, the test statistics show a statistically significant difference between users and non-users at less than a 5%

significance level. Social factors such as labor sharing for SWC practices, households' soil erosion risk recognition, attitudes toward the benefits of soil SWC structures, and membership in social groups were also incorporated into the descriptive statistics. Among the four social variables considered in this study, almost three were found to have a significantly different distribution between the users and non-users in both study areas (Table 2). Regarding the risk associated

with soil erosion, there was a significant difference between treatments and control groups in the Sebeta- hawas district. About 92% and 81% of the adopter and non-adopter farmers reported that their plots faced a risk of soil erosion. In the Kewet district, however, there was no significant difference between groups on the same factors.

The farm-pacific factors included in the empirical models of this study were cultivated land size, land fragmentation, slope status, soil fertility status of the plots, and distance from residence to farmland. Findings indicated in Table 2 reveal that adopters and non-adopters are different in such factors as farm size, slope of the plot, and soil fertility status in the Sebeta-hawas study region. The SWC technology continuous adopters had 1.39 hectares of cultivated land, with 94% of their plots located in sloppy areas and 78% of the cultivated plots having low fertility status. In contrast, non-adopters had 0.84 hectares, 72%, and 31%, respectively. In contrast, adopters in the Kewet district differ from non-adopters regarding land size, land fragmentation, and plot slope status. More specifically, adopters, on average, had 1.23 hectares of cultivated land, with 2.1 units of land fragmentation and 97% of their plots located in sloppy areas, whereas non-adopters had 0.84 hectares, 72%

and 31%, respectively. This implies that continuous adopters had relatively larger farm sizes, operated in steep plot areas, and on plots with lower fertility status.

Results of propensity score matching

According to the descriptive statistics presented in the preceding section (Table 2), continuous SWC adoption had a tentatively positive impact on agricultural land productivity in the Kewet district but a negative and insignificant impact on the Sebeta-hawas. However, attributing the effects to the treatment is difficult because productivity differences could be due to factors other than SWC practices. Thus, a PSM impact evaluation technique was used to control the observable household and plot level characteristics to determine the real productivity effects of adoption. A binary logit model was used to estimate the propensity scores of the intervention and control groups. To that end, model assumptions and multi-collinearity among variables were tested. As recommended by Tabachnick and Fidell (2019), independent variables with a high degree of correlation (r>0.70), tolerance value (0.1), variance inflation factor (>5.00), and low levels of coefficient values were excluded from the regression model. A likelihood-ratio Chow test has shown statistically significant models and coefficient variations (p-value 0.001) between regimes. This shows that comparing the two models based on explanatory variables would be more meaningful and statistically justifiable than using the pooled sample.

Furthermore, the log likelihood-ratio test conducted for each model was highly significant, indicating a good fit, and the pseudo-R-Square indicated

(10)

Open Access 4398 explanatory power of the model. In addition to PSM

estimation, the common support conditions, algorithm selection, tests for balancing conditions, sensitivity analysis, and average treatment effects on treated (ATT) between adopters and non-adopters were all performed.

Matching algorithm selection

After the propensity score was estimated using a binary logistic regression model, various matching algorithms were tried to match the treatment with control individuals having similar propensity scores.

The finding presented in Table 3 shows that for the Sebeta-hawas model, the Kernel method with a bandwidth of 0.4 was selected because it satisfied the balancing properties of explanatory variables. It produced a minimal pseudo-R square value of 0.128, the highest equal mean test value of 0.871, a lower mean bias value of 16.0, and a larger number observation size of 223 matched samples as compared to radius matching methods. Similarly, the best method in the Kewet sample was radius matching with a caliper distance of 0.2, as it resulted in the least pseudo-R square value of 0.098, the highest equal mean test value of 0.701, and the lowest mean bias of 13.4 and a matched sample size of 222, relatively. As a result, the kernel and radius matching methods have been selected as the best matching techniques under PSM in this study.

Covariate balance test

The mean comparisons between the two groups before and after matching indicate significant differences for many covariates. The likelihood ratio test for combined covariates shows that it was enormously significant before and insignificant after matching for

selected matching algorithms. However, the post- matching t-test results indicate no significant differences, suggesting that the different matching procedures enabled the balancing of the covariates (Table 3). The findings suggest that the matching procedure used effectively balanced the characteristics of the treated and control comparison groups. Thus, PSM can be used to assess the impact of continuous SWC adoption on land productivity among households with similar observed characteristics. This enables us to compare observed outcomes in treated plots with non-treated ones sharing common support. Thus, all the tests above suggest that the selected matching algorithms for the study regions were relatively the best for data analysis. Consequently, that has proved the possibility of estimating the average treatment effect on the treated (ATT) for sample households.

Common support regions

Figure 2 provides the histogram of the estimated propensity scores for SWC-treated and control units used to verify the common support condition. A visual examination of the histogram of propensity scores reveals that the common support condition is satisfied.

Furthermore, the result shows a good overlap between the distribution of the propensity scores of treatment and comparison groups, both for the Sebeta-hawas and Kewet models. Out of 525 respondents in the total sample, 37 and 186 in the Sebeta-hawas; 60 and 162 treatment and control groups in the Kewet district were found to have common support regions, respectively.

The bottom half of each figure shows the propensity score distribution for the non-treated, whereas the upper half refers to the treated individuals. Following the restriction of the common support region, choosing the matching algorithm was done.

Table 3. Performance criteria of matching algorism and matching quality analysis.

Matching Estimators

Sebeta-hawas Kewet

R2 LR X2

(p-value) Mean

(bias) Matched

sample R2 LR X2

(p-value) Mean

(bias) Matched sample Unmatched 0.688 0.000 62.4 249 0.651 0.000 44.4 276 Neighbor (1) 0.379 0.005 26.5 223 0.205 0.000 21.1 251 Neighbor (2) 0.207 0.339 20.5 223 0.175 0.002 19.0 251 Neighbor (3) 0.209 0.329 18.9 223 0.161 0.005 18.2 251 Neighbor (4) 0.205 0.349 20.0 223 0.183 0.001 20.7 251 Neighbor (5) 0.174 0.551 17.8 223 0.157 0.007 17.2 251 Kernel (.1) 0.183 0.535 20.0 223 0.158 0.007 18.3 251 Kernel (.2) 0.178 0.571 17.3 223 0.159 0.006 16.9 251 Kernel (.3) 0.149 0.757 16.8 223 0.159 0.006 15.2 251 Kernel (.4) 0.128 0.871 16.0 223 0.155 0.008 13.5 251 Radius (.01) 0.282 0.855 26.1 200 0.178 0.296 15.8 206 Radius (.02) 0.170 0.903 25.0 211 0.098 0.701 13.4 222 Radius (.03) 0.210 0.697 25.1 214 0.138 0.087 14.8 238 Radius (.04) 0.187 0.581 23.1 221 0.156 0.016 16.6 245

(11)

Open Access 4399 Sebeta-hawas Kewet

Figure 2. The common support regions of the propensity scores after matching.

Estimation of treatment effects

This study employed the procedure and assumptions required to apply the PSM model. The average treatment effect (ATT) of continuous SWC adoption was investigated using the Kernel matching model for the Sebeta-hawas sample and the Radius method for the Kewet region, as determined in the sections above (Table 3). The empirical result in Table 4 reveals that continuous adoption of SWC use impacts agricultural productivity in the Sebeta-hawas region positively. In other words, plots that received a continuous SWC technology treatment increased productivity by 1.5%

higher than those that did not adopt the measures.

In particular, the mean value of agricultural land productivity of adopter households was higher, 14,896 ETB (554 USD) per hectare) than the mean value of the non-adopters, 13,391 ETB (477 USD) per hectare, with a difference of 1,505 ETB (54 USD) per hectare on treated. However, it was not found to be significantly different at a 5% significance level between conserved and non-conserved plots. This study is in agreement with what was found by Abebe

and Bekele (2014), Adimassu et al. (2014; 2017), Abera et al. (2020), and Taye et al. (2021), who revealed that SWC intervention alone did not result in a difference between adopting and non-adopting households in terms of crop yield. One possible justification for the insignificant impacts is that SWC efforts focused on reducing soil erosion rather than enhancing land productivity, particularly in the Sebeta-hawas. Almost all respondents from this study area noted that soil replenishment functions, such as compost, manure, and other land management systems, needed to be integrated with the introduced measures. This is consistent with Adimassu et al.

(2017), who reported that integrating physical SWC measures with agronomic practices has yet to be emphasized throughout the country. The need to combine conservation efforts with soil replenishment biological and agronomic activities was recommended by Wolka et al. (2018), Belayneh et al. (2019), and Taye et al. (2021). The other reason was that Sebeta- hawas was considered a highland and high-potential area.

Table 4. Estimation of average treatment effects (ATT) using propensity score matching methods.

Variables Sebeta-hawas Kewet

KM (0.4) RM (0.02)

Matched (ATT) 0.015 0.078

Treated 14,896 22,323

Untreated 13,391 19,063

Difference 1,505 3,260

Bootstrap Standard Error 949.89 949.49

t-test 1.60 4.28

Bootstrap PV 0.112 0.000

Observations with common support 223 222

Number of treated on support 37 60

Number of treated off support 26 54

Number of controls 186 162

Total (N) 249 276

(12)

Open Access 4400 According to Etsay et al. (2019), moisture

conservation in high-potential areas contributes to the problems of water logging, weeds, and pests, and hence the contribution to land productivity is limited.

This is also confirmed by the qualitative data collected from farmers and experts in the study area.

Respondents reported that all the introduced measures hold too much water during the rainy season to the extent that it impedes free plant growth.

Study participants also stated that land lost by the structures was also a reason for the lack of yield improvements. This is consistent with Desta et al.

(2021) and Masha et al. (2021), who reported that farmers perceive structures occupying large areas of farmland that could be used for production purposes.

In addition, in some areas, the land was found to have been degraded heavily to allow easy recovery and return to its natural state. When land resources are severely degraded, restoring their productive capacity is very unlikely (Mekuriae and Hurni, 2015). By contrast, SWC adoption in the Kewet region is more productive and statistically significant at less than 5%.

About 7.7% per hectare of productivity improvement was observed on treated plots compared to control plots. More specifically, the mean value of agricultural land productivity of adopter households was higher, 22,323 ETB (830 USD) per hectare, than the mean value of non-adopter households, 19,063 ETB (678 USD) per hectare. Then, the ATT is found to be 3,260 ETB (116 USD) based on the 2019 market price. This implies that the average output per hectare could be increased by adopting SWC measures. The result is consistent with studies by Huang et al. (2020) and a review synthesis by Desta et al. (2021), who reported that SWC positively impacted productivity. The possible reason could be that some adopter households integrated physical SWC measures with composting and manure. Some households were found to move the

structures from one location to another. In other words, they relocate them to exploit fertile sediments accumulated behind the structures, which are thought to boost productivity.

The findings were consistent with those of Erkossa et al. (2018) and Masha et al. (2021), who found that SWC structures with agronomic activities increased grain yields over time. Another reason could be that Kewet was considered low land with limited agricultural potential. Adgo et al. (2013) and Erkossa et al. (2018) found a positive productivity effect in moisture-deficient soil and stone-bound plots. Physical soil and water conservation investments have lower productivity impacts in low agricultural potential areas than in high agricultural potential areas (Kassie et al.

2010).

Sensitivity analysis

There may be hidden biases against the result of matching estimators. Hence, testing the robustness of the result is recommended. A sensitivity test can address the problem of estimating the magnitude of the selection bias with non-experimental data. The primary issue in testing sensitivity is to check whether the estimated average treatment effect was purely based on the observed factor or not (Caliendo and Kopeinig, 2008). The Rosenbaum bounding sensitivity analysis was performed to check whether there is bias because of unobservable factors. The findings in Table 5 show that the productivity impacts of continuous SWC adoption do not differ between adopter and non- adopter households/plots when their odds of being treated differently by 0.5 up to Gamma = 3 in terms of unobserved covariates. This implies that the impact estimates (ATT) are insensitive to unobserved selection bias and are purely the effect of continuous SWC technology adoption.

Table 5. Rosenbaum bounds for treatments effects of SWC treated and non-treated plots.

Gama Sebeta-hawas Kewet

Q_mh+ Q_mh- p_mh+ p_mh- Q_mh+ Q_mh- p_mh+ p_mh-

1 2.21 2.21 0.01 0.00 2.96 2.96 0.000 0.00

1.5 1.06 2.96 0.05 0.00 2.34 3.35 0.000 0.00

2 0.75 3.05 0.23 0.00 1.08 4.08 0.02 0.00

2.5 0.51 4.35 0.31 0.00 0.23 4.67 0.43 0.00

3 0.09 5.08 0.54 0.00 0.08 5.19 0.53 0.00

Notes: Gamma is the log odds differential assignment due to unobserved factors. The upper (Q_mh+) and lower (Q_mh-) bounds are Mantel-Haenszel point estimates and; p_mh+ and p_mh- are the significance levels.

Conclusion

This study sought to investigate the impacts of continuous SLM adoptions on agricultural land productivity in central Ethiopia. To this end, the results from the PSM model have shown differential SWC productivity impacts across agro-ecosystems. There is evidence indicating substantial and positive effects on

treated plots compared to non-treated in the Kewet district, which is the low land and low moisture area.

Comparatively, it has positive but insignificant impacts on the Sebeta-hawas counterpart, which is the highland and high moisture area. The insignificant impacts are justified because SWC efforts focused on constructing structures rather than tailoring them with soil-replenishment and productivity enhancement

(13)

Open Access 4401 functions such as composting and the applications of

complementary biological and agronomic practices.

This implies that continuous SWC efforts positively impact agricultural productivity. However, its positive effect is more pronounced when SWC structures are integrated with productivity enhancement functions and applied in low moisture locations. In other words, physical SWC structures alone may not necessarily increase farm productivity, even if continuous maintenance and usage are ensured. Thus, policymakers and project planners should consider the role of integrating physical SWC structures with soil replenishment and productivity-improving agronomic activities. They should also pay attention to agro- ecosystem variations in scaling up measures that improve agricultural productivity.

Acknowledgements

The authors would like to acknowledge the Kewet and Sebeta-hawas districts sustainable land management project focal persons, development agents, and farmers who supplied valuable information for this study. In addition, our appreciation goes to Addis Ababa University for the financial support given to conduct the fieldwork.

References

Abebe, Y. and Bekele, A. 2014. The Impact of swc program on the income and productivity of farm households in Adama District, Ethiopia. Science, Technology and Arts Research Journal 3:198-203, doi:10.4314/star.v3i3.32.

Abera, W., Tamene, L., Tibebe, D., Adimassu, Z., Kassa, H., Hailu, H., Mekonnen, K., Desta, G., Sommer, R. and Verchot, L. 2020. Characterizing and evaluating the impacts of national land restoration initiatives on ecosystem services in Ethiopia. Land Degradation and Development 31:37-52, doi:10.1002/ ldr.3424.

Abiy, W. 2022. Soil and water conservation nexus agricultural productivity in Ethiopia. Advances in Agriculture 2022: Article ID 8611733, doi:10.1155/2022/8611733.

Adimassu, Z., Langan, S., Johnston, R., Mekuria, W. and Amede, T. 2017. Impacts of soil and water conservation practices on crop yield, run-off, soil loss and nutrient loss in Ethiopia: review and synthesis. Environmental Management 59(1):87-101, doi:10.1007/s00267-016- 0776-1.

Adimassu, Z., Mekonnen, K., Yirga, C. and Kessler, A.

2014. Effect of soil bunds on runoff, soil and nutrient losses, and crop yield in the central Highlands of Ethiopia. Land Degradation and Development 25:554- 564.

Agidew, A. and Singh, K.N. 2019. Factors affecting farmers’

participation in watershed management programs in the northeastern highlands of Ethiopia: the Teleyayen sub- watershed case study. Ecological Processes 7(1): Article number: 15, doi:10.1186/s13717-018-0128-6.

Amare, T., Zegeye, A., Yitaferu, B., Steenhuis, T., Hurni, H.

and Zeleke, G. 2014. Combined effect of soil bund with biological soil and water conservation measures in the Northwestern Ethiopian Highlands. Ecohydrology and

Hydrobiology 14(3):192-199, doi:10.1016/

j.ecohyd.2014.07.002.

Ararso, E., Geremu, T., Ayele, G., Mamo, D. and Diriba, A.

2016. Efects of level Fanya Juu and Fanya Chin structures on grain yield of maize in moisture stress areas of Daro Labu District, West Hararghe Zone, Ethiopia.

Journal of Biology, Agriculture and Healthcare 6:94-98.

Asefa, B. and Mindahun, W. 2019. Geospatial based flood risk assessment: the case of Kewet District, Amhara Region, Ethiopia. American Journal of Geographic

Information System 8(1):1-10,

doi:10.5923/j.ajgis.20190801.01.

Asfaw, D. and Neka, M. 2017. Factors affecting adoption of soil and water conservation practices: the case of Wereillu Woreda, South Wollo Zone, Amhara Region, Ethiopia. Journal of International Soil and Water Conservation Research 5:273-279.

Mekuriaw, A. and Hurni, H. 2015. Analyzing factors determining the adoption of environmental management measures on the highlands of Ethiopia. Civil and Environmental Research 7(12):61-72.

Atikilt, A., Legese, A. and Hailu, K. 2020. Impacts of stone bunds on selected soil properties and crop yield in Gumara Maksegnit watershed Northern Ethiopia.

Cogent Food and Agriculture 6(1):1785777, doi:10.1080/23311932.2020.1785777.

Becker, S and Ichino, A. 2002. Estimation of average treatment effects based on propensity scores. STATA Journal: Promoting Communications on Statistics and Stata 2(4): 358-377.

Bekele Goba, W., Muluneh, A. and Wolancho, K.W. 2022.

Evaluating farmers’ perception on soil erosion and management of physical soil and water conservation measures in southwest Ethiopia. Journal of Forestry and Natural Resources 1(2):39-53.

Belay, B. and Assefa, C. 2021. Pre-extension demonstration of improved chickpea varieties with their production packages in South West Shea Zone of Oromia Regional Estate, Ethiopia. Global Scientific Journals 9(8):304- 315.

Belayneh, M., Yirgu, T. and Tsegaye, D. 2019. Effects of soil and water conservation practices on soil physicochemical properties in Gumara watershed, Upper Blue Nile Basin, Ethiopia. Ecological Process 8:36, doi:10.1186/s13717-019-0188-2.

Biratu, A.A., Bedadi, B., Gebrehiwot, S.G., Melesse, A.M., Nebi, T.H., Abera, W., Tamene, L. and Egeru, A. 2022.

Ecosystem service valuation along landscape transformation in Central Ethiopia. Land 2022, 11:500, doi:10.3390/land11040500.

Byamukama, W., Ssemakula, E. and Kalibwani, R. 2019.

Factors influencing the uptake and sustainable use of soil and water conservation measures in Bubaare Micro- Catchment, Kabale District: South Western Uganda.

Journal of Environmental and Health Science 5:26-32.

Caliendo, M. and Kopeinig, S. 2008. Some practical guidance for the implementation of propensity score matching. Journal of Economic Survey 22(1):31-72.

CSS. 2022. Population Size by Sex, Zone and Woreda. Addis Ababa, Ethiopia.

Danso-Abbeam, G., Sedem, D, and Aidoo, R. 2018.

Agricultural extension and its effects on farm productivity and income: insight from Northern Ghana.

Agriculture and Food Security 7(74), doi:10.1186/s40066 018 0225-x.

Referensi

Dokumen terkait

Dalam program pemberdayaan masyarakat di Propinsi Bali, institusi tradisional yang dikenal dengan Desa Adat/Pakraman menjadi bagian penting dalam meningkatkan keberdayaan

yang secara umum signifikan dan tidak berhasil diatasi dengan baik oleh. manajemen serta menganggu kelangsungan

CaSO 4 :Dy was successfully prepared by co-precipitation method; and re-annealing treatment was investigated on properties of CaSO 4 :Dy and CaSO 4 :Dy with PTFE

Yani Semarang dengan Metode FAA dan LCN ” ini tidak terdapat karya yang pernah diajukan untuk laporan tugas akhir, dan sepanjang pengetahuan kami juga

Pusat Volkanologi dan Mitigasi Bencana Geologi, Departemen Energi dan Sumber Daya Mineral.. Penanggulangan Tanah

Percepatan gravitasi bumi dapat diukur dengan melemparkan sebuah benda vertikal ke atas dan mengukur interval waktu saat benda melewati titik A dan B ketika

Sebuah skripsi yang diajukan untuk memenuhi salah satu syarat memperoleh gelar Sarjana Pendidikan pada Fakultas Pendidikan Seni Dan Desain. ©Rizki Pratama 2016 Universitas

Yang jelas, kita tidak mendapati nas yang jelas dalam konteks Al-Qur'an yang menunjukkan kenabiannya dan kita juga tidak menemukan nas yang gamblang yang dapat kita jadikan