Poverty and food security impacts of sustainable intensification: Evidence from Ethiopia

Sariyev, Orkhan; Asravor, Jacob; Zeller, Manfred Article — Published Version Poverty and food security impacts of sustainable intensification: Evidence from Ethiopia Food Security Provided in Cooperation with: Springer Nature Suggested Citation: Sariyev, Orkhan; Asravor, Jacob; Zeller, Manfred (2025) : Poverty and food security impacts of sustainable intensification: Evidence from Ethiopia, Food Security, ISSN 1876-4525, Springer Netherlands, Dordrecht, Vol. 17, Iss. 2, pp. 405-420, https://doi.org/10.1007/s12571-025-01517-9 This Version is available at: https://hdl.handle.net/10419/323637 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Food Security (2025) 17:405–420 https://doi.org/10.1007/s12571-025-01517-9 ORIGINAL PAPER Poverty andfood security impacts ofsustainable intensification: Evidence fromEthiopia OrkhanSariyev1 · JacobAsravor1 · ManfredZeller1 Received: 13 May 2024 / Accepted: 10 January 2025 / Published online: 12 February 2025 © The Author(s) 2025 Abstract As sustainable intensification is a major pathway for improving agricultural productivity and reducing the environmental impacts of land use, the Government of Ethiopia and international development organizations have been promoting several practices and technologies for sustainable intensification. Using panel data from 368 farming households in Ethiopia from 2014, 2016, and 2019, this study gauges the poverty and food security impacts of Integrated Soil Fertility Management technologies and their combined use with conservation agriculture practices, specifically minimum tillage and crop rotation.We find significant positive effects of ISFM adoption in terms of increasing dietary diversity and food expenditure and reducing food insecurity. In terms of poverty, ISFM adoption decreases the probability of being poor, the poverty gap, and the severity of poverty. When combined with CA practices, we find that the effects are consistently larger for farmers who integrate ISFM and CA for all food security and poverty measures. Our findings strongly suggest that the adoption of ISFM technologies has significant positive implications for poverty reduction and improved food security. These benefits are likely to gain a considerable boost if ISFM technologies are applied together with CA practices. Keywords Technology adoption· Integrated soil fertility management· Conservation agriculture· Multinomial endogenous switching regression 1 Introduction The number of people facing hunger has been growing since 2014, especially in Africa which has witnessed a recent sharp increase in all regions of the continent (FAO etal., 2023). In Africa, continued population growth has increased food demand and climate variability has further strained farming systems that have long been performing below their productive potential (The Montpellier Panel, 2013). Agriculture is the backbone of the economy in most African countries, particularly those in sub-Saharan Africa (SSA), and has a strong correlation with economic growth (Jayne & Sanchez, 2021). Thus, increasing agricultural productivity is an overarching goal for reducing hunger and sustaining economic growth in SSA countries. Since 2000, crop production in SSA has experienced significant growth. However, much of this growth originates from the expansion of the area under cultivation, which is not a sound strategy on environmental and ecological grounds (Jayne & Sanchez, 2021). In light of this, national governments, development organizations, and foundations have been promoting various sustainable intensification (SI) approaches as a “new paradigm for African agriculture” in many SSA countries (Petersen & Snapp, 2015; The Montpellier Panel, 2013). Although the term “sustainable intensification” lacks a precise definition (Petersen & Snapp, 2015), it is often linked to agricultural technologies and practices that achieve more produce from the existing area under cultivation, while reducing the environmental footprint of agricultural production (Godfray etal., 2010; Pretty etal., 2011). Besides SI, several other approaches seek to reduce environmental externalities and improve productivity, e.g., conservation agriculture (CA), agroecology, and ecological intensification. In terms of SI, Integrated Soil Fertility Management (ISFM) technologies, which align well with * Orkhan Sariyev o.sariye[email protected] Jacob Asravor jacob.asrav[email protected] Manfred Zeller [email protected] 1 Department ofRural Development Theory andPolicy, University ofHohenheim, Stuttgart, Germany 406 O.Sariyev et al. the principles of SI, have been recognized as means to expand agricultural productivity and prevent soil degradation (Pretty etal., 2011; Vanlauwe etal., 2015). Likewise, CA has received significant attention due to its potential to improve soil quality. Rural households are likely to adopt a combination of various approaches and practices. This paper seeks to answer the question of how SI impacts poverty and food security among rural smallholders in Ethiopia by utilizing three rounds of panel data from Ethiopia and concentrating on ISFM technologies and their combination with CA. Agricultural intensification via ISFM is considered the leading pathway to the African Green Revolution (Petersen & Snapp, 2015). ISFM aims to enhance the agronomic use efficiency of inputs. It is defined as a set of soil fertility management practices which include fertilizer, improved germplasm, and organic inputs, such as manure, compost, or crop residues adapted to the local context (Vanlauwe etal., 2010). CA is another SI approach, which is based on minimal soil disturbance, crop residue retention, and crop rotation or intercropping (Giller etal., 2009). The combination of CA practices is likely to improve soil quality and profitability, assuming the sufficient availability of necessary machinery (Johansen etal., 2012) and inputs, such as fertilizer and herbicides (Giller etal., 2009). ISFM requires upfront investment in purchased inputs, such as fertilizer and improved seeds. CA practices are labor-intensive; thus, their adoption may increase labor costs (Montt & Luu, 2020; Vanlauwe etal., 2010). While ISFM adopters are likely to experience higher yields in the short term, the benefits of CA usually occur in the long term (Giller etal., 2009). Considering the increasing efforts to promote SI in SSA, studies have investigated the implications of different SI practices on various livelihood indicators in different country contexts. However, despite existing studies (Hörner & Wollni, 2021, 2022; Kassie etal., 2018; Khonje etal., 2018; Maggio etal., 2022; Teklewold etal., 2013a, 2013b) that examine various CA or ISFM practices and technologies or their combinations, rigorous studies on the poverty and food security implications of ISFM technologies are scant. Among these studies, in a study from Uganda, Maggio etal. (2022) find that both organic fertilizer and maize-legume intercropping increases the total value of crop production. Using two rounds of data from Ethiopia, Kassie etal. (2018) show that different combinations of fertilizer, improved maize varieties, and legume diversification increase maize yield and income. In another study from Ethiopia, Teklewold etal. (2013b) report that maize farmers achieve the highest income when improved maize varieties, maize-legume rotations, and minimum tillage are adopted in combinations compared to in isolation. Tesfaye etal. (2021) further find that cereal-legume intercropping, reduced tillage, and their combination reduce poverty headcount, the poverty gap, and the severity of poverty in Ethiopia. Khonje etal. (2018) show that Zambian farmers achieve higher yield and income when improved maize variety and CA are jointly adopted. Motivated by the fact that most of the evidence on the impact of ISFM type practices is limited to crop yield and revenue, Hörner and Wollni (2021) study ISFM implications on household income, food security, and child education. They find positive effects of ISFM on these welfare indicators depending on the agro-ecological zone. Using the same dataset, Hörner and Wollni (2022) show that ISFM practices and their combinations improve land and labor productivity. Besides Hörner and Wollni (2021), we are not aware of any other study that investigates the effects of ISFM adoption on poverty or food security outcomes. Employing three waves of panel data from a technology adoption survey in Ethiopia, this study contributes to the literature by concentrating on ISFM practices and analyzing their adoption implications on poverty and food security outcomes and by examining the joint effects of ISFM and CA adoption on these outcomes. The paper proceeds as follows: Sect.2 presents the data, variables of interest, and the methodology employed for the analyses; results are presented in Sect.3 and discussed in Sect.4; Sect.5concludes the paper. We provide detailed description of the study context in the appendix. 2 Data andmethods This section first presents the sampling strategy and data utilized for this study before describing in detail the poverty and food security measures and the strategy for empirical analyses. 2.1 Sampling anddata We employ household-level data collected from a random sample of farm households in 15 woredas (districts) across 10 zones in southwestern Ethiopia. Figure1 shows the survey locations which have varying climatic and agro-ecological characteristics. The sample is derived from a sub-sample of a nationally representative baseline survey conducted in 2012 by the International Food Policy Research Institute (IFPRI) and the Agricultural Transformation Agency (ATA) of Ethiopia.1 Due to administrative and logistical constraints, the surveyed households were randomly selected from the list of surveyed farmers situated within a radius of approximately 200km around the town of Hawassa. The survey was different from the nationally representative baseline conducted in 2012 and was specifically designed for collecting data on smallholder farmers' risk management and 1 For more details on this sample, see (Minot & Sawyer, 2013). 407 Poverty and food security impacts of sustainable intensification innovation strategies and their impacts on poverty and resilience. The areas cover parts of Southern Nations, Nationalities, and Peoples’ (SNNP) and Oromia regions. Three follow-up surveys were conducted in 2014, 2016, and 2019. This study does not use the 2012 baseline by IFPRI and ATA because some important data collected in the follow-up surveys were not present in the baseline data. The sample consists of a balanced panel of 376 farm households; however, the data employed in this study covers 368 households due to missing consumption data for eight households in 2016. Using the geographical coordinates of the sampled households, we supplement the survey data with historical rainfall data extracted from the Climate Hazards Group InfraRed Precipitation with Station data. This is a quasi-global rainfall dataset that incorporates 0.05° resolution satellite imagery with insitu station data to create gridded precipitation data (Funk etal., 2015). We extract annual rainfall data between 1981 and 2018 and calculate standard deviations and historical averages for the mentioned period. 2.2 Poverty andfood security measures Adoption of SI technologies and practices is expected to promote productivity and resilience among rural farmers. Higher productivity and improved resilience are expected to contribute to reduced poverty and improved food security among the adopters of SI technologies and practices. In this study, we employ different measures of poverty and food security which reflect shortand long-term outcomes. We assume that poverty outcomes accrue to long-term sustained productivity gains from SI. Food security outcomes like food expenditure and dietary diversity are likely to derive from short-term productivity gains. Moreover, these outcomes can also reflect the long-term resilience of the households. Following Tesfaye etal. (2021), we employ the FosterGreer-Thorbeck method in calculating the various poverty indices (Foster etal., 1984): (1) P 𝛼=1 N N ∑ i=1 (z−yi z) 𝛼 I(yi<z) , Fig. 1 Location of sampled households 408 O.Sariyev et al. where yi denotes the per adult equivalent monthly consumption expenditure deflated by the consumer price index using 2016 prices as the base, N denotes the sample size, z denotes the national poverty line,2 and I(yi<z) takes the value of 1 when the ith household has a per adult equivalent monthly consumption expenditure below the national poverty line. Because we are not interested in the fraction of the population or sample below the poverty line, we remove 1 N from the formula. We calculate three poverty indices: when 𝛼=0 , P is the poverty headcount which indicates if the ith household is poor; when 𝛼=1 , P denotes the poverty gap index which shows how far the ith household is from the poverty line; and when 𝛼=2 , P indicates the squared poverty gap or poverty severity index which reflects inequality among the poor. We adopt three distinct indicators to ascertain the food security status of the sampled households. The household dietary diversity score (HDDS) is a commonly employed measure of household level dietary quality and food security. HDDS describes the household’s ability to access different food items. Ruel (2003) concludes that considering its association with energy availability and per capita income, capturing dietary diversity is a practical method of understanding the food security status of households. The survey employed a seven-day recall period to capture day-to-day variation in household diets. Using this data, we calculate HDDS based on 12 food groups: cereals; white tubers and roots; vegetables; fruits; meat; eggs; fish and other seafood; legumes, nuts, and seeds; milk and milk products; oils and fats; sweets; and spices, condiments, and beverages (Kennedy etal., 2011). Our second measure of food security is the probability of experiencing low food availability in at least one month over the past 12months. Finally, we calculate the adult equivalent real food expenditure based on the reported monthly food expenditure3 deflated by the consumer price index using 2016 prices as the base. 2.3 Empirical strategy Our objective is to unpack the adoption implications of ISFM practices and their combination with CA on poverty and food security. The CA practices considered in this study are minimum tillage and crop rotation. A household is considered an adopter of CA practices if either minimum tillage or crop rotation or both are adopted. To understand the impacts of technologies and practices adopted by households, for each household we must compare the respective outcomes from adoption to the outcomes from non-adoption. However, for each household, we can only observe one state at a time because each household either adopts a certain combination of ISFM practices (or their combination with CA) or does not adopt. Thus, we have a missing counterfactual which is the fundamental challenge for causal inference. In this case, we need to compare adopters with non-adopters, which leads to another challenge – selection bias or endogeneity (Imbens & Wooldridge, 2009). Agricultural technology adoption is a choice outcome which can be influenced by some unobservable factors, such as skills, motivation, care for family, bond between family members, and their dedication to a common goal, that can also influence food security and/or poverty outcomes. Availability of panel data solves the problem if the unobserved heterogeneity is timeinvariant. However, the bias still arises if the unobserved heterogeneity is time-variant. To overcome potential endogeneity, we apply the multinomial endogenous switching regression (MESR) model, which has been employed in several technology adoption studies (Biru etal., 2020; Hörner & Wollni, 2022; Kassie etal., 2015, 2018; Khonje etal., 2018; Montt & Luu, 2020; Teklewold etal., 2013b; Tesfaye etal., 2021). MESR estimates separate outcome regressions for adopters and nonadopters, which allows for an interaction between the adoption decision and observed and unobserved heterogeneity. This ensures that adoption has an impact on both the slope and intercept in the outcome equations. The estimated outcomes represent the returns to the characteristics of adopters and non-adopters (Biru etal., 2020; Di Falco etal., 2011; Kassie etal., 2018). Given that we examine the impacts of ISFM adoption and the combination of ISFM and CA, we run two separate MESR.4 Here, we take ISFM impact estimation as an example to describe the implementation of the method. MESR is implemented in two steps. In the first step, we model the determinants of ISFM adoption. We assume that at each period, famers select an ISFM technology set that maximizes their utility. Among the sampled households, chemical fertilizer is the dominant ISFM technology that is adopted. Hence, we consider chemical fertilizer as a core component and hypothesize two possible ISFM sets. One set entails the partial adoption of two out of three core ISFM technologies, specifically, either chemical fertilizer with an 2 The Central Statistical Agency (2017) reports that the national poverty line for 2015/2016 is 7,184 Ethiopian Birr per year, which is equal to 599 Ethiopian Birr per month per adult person. 3 This measure consists of two parts: food consumed at home which can be purchased or produced (based on the reported market price), and food consumed away from home. 4 CA practices only include conservation tillage and crop rotation. Crop residue (i.e. mulch or compost) is considered as an ISFM technology in this study. For mulch, crop rotation, and conservation tillage, households are considered as adopters if they apply the practice on at least 10% of their farmland. Please refer to the appendix for further detail on the technology combinations that constitute a combination of ISFM and CA practices. 409 Poverty and food security impacts of sustainable intensification improved variety or with an organic input.5 The second set comprises complete ISFM adoption, meaning that all three ISFM components are adopted: chemical fertilizer, improved seeds, and organic inputs. Since the ISFM principle requires the simultaneous use of these technologies, even if a household adopts one ISFM technology, we code that household as a non-adopter. Thus, assuming a technology set k=0,1, 2 , where k=0 indicates that no ISFM set was adopted, we can show farmers’ random utility at a certain time t as Ukt . A rational farmer will adopt a particular technology set given that the expected utility is higher than the alternatives, i.e., Ukt >Umt,m≠k . Following Kassie etal. (2018) under the assumption of independence of irrelevant alternatives (IIA),6 we estimate the first stage with the multinomial logit model. The probability that a household i with characteristics X will adopt ISFM set k is given as: where i denotes the household, k is the technology set, t indicates year, αk is a constant term of the technology set j,βk indicates represents parameters to be estimated, and Xit indicates various household socio-economic characteristics, survey year, and rainfall variables. To benefit from the panel structure of the dataset, we estimate the pooled selection and outcome (i.e., second step) models using the Mundlak approach and include the means of all time-varying variables into both equations. The Mundlak device helps to control time-invariant unobserved heterogeneity as with fixedeffects. Hi in Eq.2 indicates the means of the time-varying explanatory variables. In the second stage, outcome models are estimated for adopters of each technology sets and nonadopters separately: where Oit indicates the outcome (i.e., food security or poverty indicators) observed for household i at time t , Zit indicates various household socio-economic characteristics and (2) Prob� k � Xit,Hi � =𝑒𝑥𝑝 � 𝛼k+Xit𝛽k+Hi � ∑ k m=1 exp(𝛼 m +X it 𝛽 m +H i ) ,k= 0, 1, 2 (3) ⎧ ⎪ ⎨ ⎪ ⎩ Regime0∶Oit0=Zit0β0+  λit0σ0+Hi0+ε it0ifnon −adopter Regime1∶Oit1=Zit1β1+  λit1σ1+Hi1+ε it1ifpartialISFMadopter Regime2∶Oit2=Zit2β2+  λit2σ2+Hi2+ε it2ifcompleteISFMadopter , survey years, β denotes parameters to be estimated, Hi indicates means of time-varying explanatory variables, and σ represents the covariance between outcome and selection equation errors. An important term in Eq.3 is  λ , which represents the inverse Mills ratios derived from the selection equation to capture time-variant individual effects. Inverse Mills ratios generated from the first stage help improve identification; however, it is important that the first stage includes a selection instrument. We include two instrumental variables in the first stage: the standard deviation of rainfall between 1981 and 2018 and the historical average for the same period. We assume that historical rainfall patterns can play a role in technology choices and adoption decisions of farmers, but that they are not directly related to the outcomes of interest. However, there is a possibility that rainfall variation could affect outcome variables of interest through non-farm income. Drought (i.e., within the last growing season) leads to increased employment in off-farm self-employment but not in off-farm wage-employment in Ethiopia (Musungu etal., 2024). Although the evidence provided is for the last growing season, we can assume that historical variation can affect farmers’ expectations from agricultural production. Controlling for non-farm income, we block the effect of the instruments on outcome variables possible operating through nonfarm income and achieve conditional independence of the instruments. Teklewold etal. (2013a) show that the probability of manure application is high in areas where rainfall is reliable in terms of distribution, time, and amount. Dercon and Christiaensen (2011) find that fertilizer application is profitable for cereals in Ethiopia, given that rainfall does not largely deviate from its usual patterns. Farmers do not benefit from fertilizer if rainfall is not sufficient. Kassie etal. (2013) find that farmers in Tanzania choose to cultivate traditional varieties over improved varieties if the expected rainfall is insufficient. This implies that in the absence of a reliable rainfall amount, farmers do not invest in expensive inputs. Considering these results, we can assume that rainfall deviation and the historical average of rainfall are negatively and positively associated with the adoption of ISFM practices, respectively. In terms of CA practices, Teklewold etal. (2013a) report that similar to manure application, the probability of adopting crop rotation is high in areas where rainfall is reliable. Arslan etal. (2014) find a positive association between district-level historical rainfall variation and the adoption and intensity of minimum tillage. Given these findings, we again assume that rainfall variation and the historical average of rainfall are determinants of the adoption of the CA practices considered in this study. Equation3 is used to predict observed and counterfactual outcomes for adopters of the different technology sets. The 5 We consider manure, mulch, and compost as organic inputs under ISFM. Unfortunately, the data does not allow us to differentiate organic mulch materials from other mulch materials, such as plastics. 6 Bourguignon etal. (2007) show that selection bias correction based on the multinomial logit model provides sufficient correction for the second stage outcome equation even if IIA is violated. In this study, we employed Dubin and Mc Fadden’s Model variant 1 using the “selmlog” Stata command developed by Bourguignon et al. (2007) with 200 bootstrapped replications. 410 O.Sariyev et al. observed outcomes are computed given the information in the data as: where Zi denotes observed means of time-varying explanatory variables; β,σ, and θ are the coefficients estimated for adopters. To be able to estimate average treatment effects on the treated, it is important to also calculate the counterfactual outcomes for adopters of technology sets. The counterfactuals for adopters are calculated as follows: Parameters in Eq.5 are coefficients obtained from the outcome regressions for non-adopters. With this equation, we calculate the outcomes that adopters would have obtained if the returns to their characteristics had been the same as the returns observed by non-adopters. The standard errors are obtained by 200 bootstrapped replications to account for the first stage. The difference between Eqs.4 and 5 provides us with the average treatment effects on the treated: The term (βk−β 0)Zitk represents the change observed in the outcome due to differences in observed characteristics. The second and third terms, namely (σk −σ 0 )  λ itk and ( 𝜃 k −𝜃 0 )Z ik , denote the differences observed due to time-variant and time-invariant unobserved characteristics, respectively. Considering that we test multiple hypotheses, there is a probability of rejecting a true null hypothesis. Hence, we control for the false discovery rate and calculate and report the sharpened q-values as suggested by Anderson (2008). 3 Results This section provides an overview of the surveyed households before presenting the results of the impacts of SI on food security and poverty outcomes. 3.1 Sample characteristics This subsection reports data on important sample characteristics. Table1 lists relevant descriptive statistics, and Fig.2 depicts adoption trends for ISFM and CA practices and technologies. Among the outcome variables, HDDS was close to six food groups in 2014, while in the later rounds, it increased to seven food groups, indicating an improvement in food security. Although real per capita food expenditure has declined over the years, the probability of experiencing (4) E (O itK | k=K)=Z itk β k +  λ itk σ k +Z ik 𝜃 k ,k= 1,2 (5) E( O it0 |k = K )= Z itkβ0+ λitkσ0+ Z ik 𝜃 0 ,k =1,2 (6) ATT i =E(O itK |k=K)−E(O it0 |k=K)= (βk −β 0 )Z itk + (σ k −σ 0 )  λ itk +(𝜃 k −𝜃 0 )Z ik. food insecurity has also declined significantly from the first survey round to the later rounds. In terms of poverty indicators, all show that poverty and its severity are growing among the sampled households. Regarding household characteristics, the sampled households have, on average, six members and 17% are headed by women. It appears that households have extended their social networks by taking active roles in social organizations. Approximately, 16% of households received loans and around 34% had non-farm income in all survey rounds. Between 2014 and 2016, the average farm size increased, but did not change in 2019. The primary crops are barley and maize, with the share of farmers cultivating maize increasing to 40% in 2019, from approximately 33% in the previous survey rounds. Additionally, approximately 42% of the sampled households cultivate barley. Tropical livestock units (TLU) owned increased between 2014 and 2016, but declined in 2019. We observe growth in the number of farmers who perceive their soil quality to be either poor or good. Over the years, there have been improvements in infrastructure, as indicated by less time needed to reach the nearest periodic market and extension office. The share of farmers attending extension training has increased over the years. However, we observe a decline in the share of farmers receiving extension visits in 2019. On the other hand, the share of households referring to their fellow farmers for information regarding agricultural practices and technologies increased significantly, especially in 2019. In terms of rainfall, the year 2015 was, on average, drier than 2013 and 2018. We observe that the number of ISFM adopters7 increased from one survey round to the next. In terms of ISFM and CA combination, the number of farmers implementing both practices increased from 2014 to 2016, and in 2019 it remained close to the 2016 values. Overall, the number of farmers who do not practice any combination of ISFM or CA practices declined significantly from 2014 to 2016. The number of ISFM adopters who do not practice CA also increased from one survey round to the next. Moreover, the number of farmers who practice only CA increased from 2014 to 2016, but returned to 2014 levels in 2019. Among the ISFM technologies, fertilizer application was the dominant component, with an increased share of households using fertilizer over the years. In 2019, almost 80% of the sampled households indicated applying some fertilizer. The share was 72% in 2014 and 77% in 2016. Regarding manure, 23%, 44%, and 49% of the sampled households reported its use in 2014, 2016, and 2019, respectively. Households have also increased their use of improved varieties over the years. Mulch was utilized by 9%, 6%, and 8% of 7 Those who use chemical fertilizer with either manure or improved variety (partial ISFM) or both (complete ISFM). 411 Poverty and food security impacts of sustainable intensification Table 1 Sample characteristics Notes: ISFM is Integrated Soil Fertility Management; CA is conservation agriculture; HDDS is the household dietary diversity score; HH is household; and TLU is tropical livestock units. a distance, soil quality, and land certificate are for the biggest parcel; soil quality is based on farmers’ perceptions; b partial adoption of ISFM: either chemical fertilizer with improved variety or with organic input and complete adoption of ISFM: all technologies, chemical fertilizer, improved seeds, and organic input; c Partial or complete ISFM with crop rotation and/or minimum tillage as CA; 2014 2016 2019 Household characteristics HDDS 6.36 (1.68) 6.71 (1.51) 6.67 (1.63) Adult equivalent real food expenditure (ln) 6.25 (0.73) 6.18 (0.66) 6.09 (0.77) Food insecurity experience (= 1 if yes) 0.67 (0.47) 0.39 (0.49) 0.39 (0.49) Poverty headcount (= 1 if poor) 0.60 (0.49) 0.62 (0.49) 0.65 (0.48) Poverty gap 0.22 (0.25) 0.22 (0.22) 0.26 (0.26) Poverty severity 0.11 (0.16) 0.10 (0.14) 0.13 (0.18) Female head 0.17 (0.37) 0.16 (0.37) 0.17 (0.38) Head age 44.90 (13.85) 48.15 (13.41) 49.71 (14.53) Years of formal schooling of head 3.17 (3.61) 3.12 (3.58) 3.13 (3.35) Number of household members 6.37 (2.24) 6.45 (2.33) 6.33 (2.30) Dependency ratio 1.20 (0.88) 1.16 (0.84) 1.09 (0.86) Farm-specific characteristics Distance to periodic market (minutes) 52.63 (45.18) 47.84 (40.24) 40.90 (34.37) Distance to ext. office (minutes) 46.46 (46.04) 37.53 (39.04) 31.25 (31.90) Distance to parcel (minutes) a14.90 (22.31) 20.48 (34.65) 18.78 (32.62) Land certificate (= 1 if yes) a0.83 (0.37) 0.67 (0.47) 0.78 (0.42) Poor soil quality (= 1 if yes) a0.04 (0.20) 0.03 (0.17) 0.08 (0.27) Good soil quality (= 1 if yes) a0.52 (0.50) 0.52 (0.50) 0.64 (0.48) Farm size in hectares 1.47 (1.20) 1.67 (1.56) 1.67 (2.52) TLU owned 3.88 (4.32) 4.59 (4.64) 3.96 (4.10) Pesticides (= 1 if applied) 0.34 (0.47) 0.30 (0.46) 0.42 (0.49) Maize (= 1 if produced in prev. season) 0.34 (0.48) 0.32 (0.47) 0.40 (0.49) Barley (= 1 if produced in prev. season) 0.42 (0.49) 0.43 (0.50) 0.41 (0.49) Number of shocks 0.36 (0.58) 0.25 (0.51) 0.42 (0.61) No. of social organizations HH is active in 1.41 (1.06) 1.84 (1.45) 2.13 (1.52) Loan (= 1 if acquired in past 12months) 0.15 (0.35) 0.16 (0.37) 0.17 (0.37) Non-farm income (= 1 if yes) 0.34 (0.47) 0.33 (0.47) 0.36 (0.48) Extension visits (= 1 if yes in past 24months) 0.39 (0.49) 0.39 (0.49) 0.17 (0.37) Ext. training (= 1 if yes in past 24months) 0.44 (0.50) 0.43 (0.50) 0.54 (0.50) Fellow farmer as information source (= 1 if yes) 0.21 (0.41) 0.30 (0.46) 0.63 (0.48) Annual rainfall previous season (mm) 1286.47 (301.01) 1095.16 (269.27) 1328.32 (263.97) ISFM b (% in parentheses) No ISFM combination 207 (56.3%) 162 (44.0%) 142 (38.6%) Partial adoption of ISFM technologies 115 (31.3%) 122 (33.2%) 143 (38.9%) Complete adoption of ISFM technologies 46 (12.5%) 84 (22.8%) 83 (22.6%) ISFM and CA combination c No ISFM combination or CA 136 (37.0%) 77 (20.9%) 75 (20.4%) ISFM + CA 111 (30.2%) 135 (36.7%) 138 (37.5%) CA 71 (19.3%) 85 (23.1%) 67 (18.2%) ISFM 50 (13.6%) 71 (19.3%) 88 (23.9%) Observations 368 368 368 412 O.Sariyev et al. the sampled households in the years 2014, 2016, and 2019, respectively. Crop rotation was practiced by 48%, 59%, and 55% of the sampled households in 2014, 2016, and 2019, respectively. There was a significant growth in the share of farmers practicing composting over the years, i.e., 22% in 2014, 35% in 2016, and 42% in 2019. Very few households reported reduced tillage uptake. 3.2 Impacts ofSI The objective of this study is to evaluate the impacts of widely promoted SI practices on poverty and food security among rural smallholder farmers in Ethiopia. In this subsection, we concentrate on ISFM adoption and its combined use with CA before presenting their effects on outcome measures. Table2 presents the multinomial logistic regression results on ISFM adoption and the adoption of the combination of ISFM and CA. We perform the Hausman test of IIA and do not find any evidence of violating the IIA assumption. This implies that the multinomial logistic regression model can be applied to our data. We observe a significant positive effect of pesticide application on the adoption of ISFM technologies and CA practices, and their combined use. Maize and barley producers are likely to adopt ISFM technologies and their combination with CA practices. We find that farmers who are located in high altitude areas are less likely to adopt these practices. The results further show that participating in extension training programs and referring to a fellow farmer for agricultural-related information are positively associated with the uptake of the combination of ISFM and CA, as well as the adoption of CA alone. Regarding the instruments, as hypothesized, the results suggest that historical average rainfall has a positive link with ISFM adoption in both specifications. The standard deviation of historical rainfall is positively correlated with the adoption of CA practices and the combined use of ISFM and CA. To determine whether the instruments are admissible, we follow Di Falco etal. (2011) and perform a simple falsification test where the instruments are assumed to affect adoption decisions, but not the outcome for non-adopters. TablesA1 and A2 in the appendix report the results from this simple falsification test. The results do not provide any consistent evidence of a significant relationship between the instruments and the outcomes for non-adopters at 5% error probability. This suggests that the instruments are valid: they are jointly significant determinants of adoption decision but not the outcomes of households who do not adopt technology sets under consideration. Moreover, we run random effects, fixed effects, and Inverse Probability Weighted Regression Adjustment (separated for each round) impact estimation techniques to assess the robustness of the results. TablesA5 -A9 depict the results from these alternative techniques. Most of the results are in line with the MESR results. To visualize and compare the magnitude of the impacts, we depict the impacts of partial and complete ISFM adoption vs. no ISFM adoption in Fig.3 and Fig.4, respectively. These results are also available in table format in the supplementary appendix (in TablesA3 and A4). The figures illustrate the observed and counterfactual outcomes and the difference between them. There are positive average treatment effects for partial and complete adopters of ISFM technologies in terms of HDDS and adult equivalent per capita food expenditure. Also, both partial and complete adoption of ISFM technologies reduces the probability of experiencing food insecurity over the past 12months by 8% and 6%, respectively. Both partial and complete adopters of ISFM technologies would have around 13% Fig. 2 Adoption of ISFM and CA practices, by year 419 Poverty and food security impacts of sustainable intensification Scientific Data, 2(1), 150066. https:// doi. org/ 10. 1038/ sdata. 2015. 66 Giller, K. E., Witter, E., Corbeels, M., & Tittonell, P. (2009). Conservation agriculture and smallholder farming in Africa: The heretics’ view. Field Crops Research, 114(1), 23–34. https:// doi. org/ 10. 1016/j. fcr. 2009. 06. 017 Godfray, H. C. J., Beddington, J. R., Crute, I. R., Haddad, L., Lawrence, D., Muir, J. F., Pretty, J., Robinson, S., Thomas, S. M., & Toulmin, C. (2010). Food security: The challenge of feeding 9 billion people. Science, 327(5967), 812–818. https:// doi. org/ 10. 1126/ scien ce. 11853 83 Hörner, D., & Wollni, M. (2021). Integrated soil fertility management and household welfare in Ethiopia. Food Policy, 100, 102022. https:// doi. org/ 10. 1016/j. foodp ol. 2020. 102022 Hörner, D., & Wollni, M. (2022). Does integrated soil fertility management increase returns to land and labor? Agricultural Economics, 53(3), 337–355. https:// doi. org/ 10. 1111/ agec. 12699 Imbens, G. W., & Wooldridge, J. M. (2009). Recent developments in the econometrics of program evaluation. Journal of Economic Literature, 47(1), 5–86. https:// doi. org/ 10. 1257/ jel. 47.1.5 Jayne, T. S., & Sanchez, P. A. (2021). Agricultural productivity must improve in Sub-Saharan Africa. Science, 372(6546), 1045–1047. https:// doi. org/ 10. 1126/ scien ce. abf54 13 Johansen, C., Haque, M. E., Bell, R. W., Thierfelder, C., & Esdaile, R. J. (2012). Conservation agriculture for small holder rainfed farming: Opportunities and constraints of new mechanized seeding systems. Field Crops Research, 132, 18–32. https:// doi. org/ 10. 1016/j. fcr. 2011. 11. 026 Kassie, M., Jaleta, M., Shiferaw, B., Mmbando, F., & Mekuria, M. (2013). Adoption of interrelated sustainable agricultural practices in smallholder systems: Evidence from rural Tanzania. Technological Forecasting and Social Change, 80(3), 525–540. https:// doi. org/ 10. 1016/j. techf ore. 2012. 08. 007 Kassie, M., Teklewold, H., Jaleta, M., Marenya, P., & Erenstein, O. (2015). Understanding the adoption of a portfolio of sustainable intensification practices in eastern and southern Africa. Land Use Policy, 42, 400–411. https:// doi. org/ 10. 1016/j. landu sepol. 2014. 08. 016 Kassie, M., Marenya, P., Tessema, Y., Jaleta, M., Zeng, Di., Erenstein, O., & Rahut, D. (2018). Measuring farm and market level economic impacts of improved maize production technologies in Ethiopia: Evidence from panel data. Journal of Agricultural Economics, 69(1), 76–95. https:// doi. org/ 10. 1111/ 14779552. 12221 Kennedy, G., Ballard, T., & Dop, M. C. (2011). Guidelines for measuring household and individual dietary diversity. Food and Agriculture Organization of the United Nations. Khonje, M. G., Manda, J., Mkandawire, P., Tufa, A. H., & Alene, A. D. (2018). Adoption and welfare impacts of multiple agricultural technologies: Evidence from eastern Zambia. Agricultural Economics, 49(5), 599–609. https:// doi. org/ 10. 1111/ agec. 12445 Maggio, G., Mastrorillo, M., & Sitko, N. J. (2022). Adapting to high temperatures: Effect of farm practices and their adoption duration on total value of crop production in Uganda. American Journal of Agricultural Economics, 104(1), 385–403. https:// doi. org/ 10. 1111/ ajae. 12229 Minot, N., & Sawyer, B. (2013). Agricultural production in Ethiopia: Results of the 2012 ATA baseline survey. https:// reap. ifpr i. info/ files/ 2013/ 12/ ag_ produ ction_ in_ ethio pia. pdf Montt, G., & Luu, T. (2020). Does conservation agriculture change labour requirements? Evidence of sustainable intensification in Sub-Saharan Africa. Journal of Agricultural Economics, 71(2), 556–580. https:// doi. org/ 10. 1111/ 14779552. 12353 Musungu, A. L., Kubik, Z., & Qaim, M. (2024). Drought shocks and labour reallocation in rural Africa: Evidence from Ethiopia. European Review of Agricultural Economics, jbae020. https:// doi. org/ 10. 1093/ erae/ jbae0 20 Petersen, B., & Snapp, S. (2015). What is sustainable intensification? Views from experts. Land Use Policy, 46, 1–10. https:// doi. org/ 10. 1016/j. landu sepol. 2015. 02. 002 Pretty, J., Toulmin, C., & Williams, S. (2011). Sustainable intensification in African agriculture. International Journal of Agricultural Sustainability, 9(1), 5–24. https:// doi. org/ 10. 3763/ ijas. 2010. 0583 Ruel, M. T. (2003). Operationalizing dietary diversity: A review of measurement issues and research priorities. The Journal of Nutrition, 133(11 Suppl 2), 3911S-3926S. https:// doi. org/ 10. 1093/ jn/ 133. 11. 3911S Tambo, J. A., & Mockshell, J. (2018). Differential impacts of conservation agriculture technology options on household income in Sub-Saharan Africa. Ecological Economics, 151, 95–105. https:// doi. org/ 10. 1016/j. ecole con. 2018. 05. 005 Teklewold, H., & Mekonnen, A. (2017). The tilling of land in a changing climate: Empirical evidence from the Nile Basin of Ethiopia. Land Use Policy, 67, 449–459. https:// doi. org/ 10. 1016/j. landu sepol. 2017. 06. 010 Teklewold, H., Kassie, M., & Shiferaw, B. (2013a). Adoption of multiple sustainable agricultural practices in rural Ethiopia. Journal of Agricultural Economics, 64(3), 597–623. https:// doi. org/ 10. 1111/ 14779552. 12011 Teklewold, H., Kassie, M., Shiferaw, B., & Köhlin, G. (2013b). Cropping system diversification, conservation tillage and modern seed adoption in Ethiopia: Impacts on household income, agrochemical use and demand for labor. Ecological Economics, 93, 85–93. https:// doi. org/ 10. 1016/j. ecole con. 2013. 05. 002 Tesfaye, W., Blalock, G., & Tirivayi, N. (2021). Climate-Smart innovations and rural poverty in Ethiopia: Exploring impacts and pathways. American Journal of Agricultural Economics, 103(3), 878–899. https:// doi. org/ 10. 1111/ ajae. 12161 The Montpellier Panel. (2013). Sustainable intensification: A new paradigm for African agriculture. Vanlauwe, B., Bationo, A., Chianu, J., Giller, K. E., Merckx, R., Mokwunye, U., Ohiokpehai, O., Pypers, P., Tabo, R., Shepherd, K. D., Smaling, E. M. A., Woomer, P. L., & Sanginga, N. (2010). Integrated soil fertility management. Outlook on Agriculture, 39(1), 17–24. https:// doi. org/ 10. 5367/ 00000 00107 91169 998 Vanlauwe, B., Descheemaeker, K., Giller, K. E., Huising, J., Merckx, R., Nziguheba, G., Wendt, J., & Zingore, S. (2015). Integrated soil fertility management in Sub-Saharan Africa: Unravelling local adaptation. The Soil, 1(1), 491–508. https:// doi. org/ 10. 5194/ soil-14912015 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Orkhan Sariyev is a postdoctoral researcher at the Institute of Agricultural Sciences in the Tropics (Hans-Ruthenberg-Institute), University of Hohenheim, Germany. His primary interests revolve around agricultural and development economics, with a particular focus on food security, sustainable development, and women's empowerment. He is especially intrigued by rural development related policy and program evaluations, seeking to evaluate their effectiveness and identify areas for improvement. 420 O.Sariyev et al. Jacob Asravor is a doctoral candidate at the Institute of Agricultural Sciences in the Tropics (Hans-Ruthenberg-Institute), University of Hohenheim, Germany. His research interest encompasses issues of agricultural productivity and efficiency analysis, food security, technology and innovation adoption, and poverty reduction. He has keen interest in issues relating to how the adoption of agricultural technologies and innovations translate into greater food security and reduced poverty for rural farm families. Manfred Zeller is a Professor and Chair of Rural Development Theory and Policy at the HansRuthenberg-Institute, University of Hohenheim, Germany. He served as the Director of the Food Security Centre (2009– 2014), worked as a Senior Research Fellow and Program Leader for the International Food Policy Research Institute (IFPRI) from 1993 to 1999 and again from 2014 to 2016, and held a professorship at the University of Goettingen (1999–2005). His academic work emphasizes applied research on the impact of food, agriculture, and rural development policies on income, poverty status, and food security. His academic journey includes studies in Agricultural Sciences at the University of Bonn, where he earned a doctoral title in Agricultural Economics in 1990.