Are agro‐clusters pro‐poor? Evidence from Ethiopia

J Tabe‐Ojong, Ma in Paul; Du e i, Guyo Godana A icle — Published Ve sion A e ag o‐clus e s p o‐poo ? E idence om E hiopia Jou nal o Ag icul u al Economics P o ided in Coope a ion wi h: John Wiley & Sons Sugges ed Ci a ion: J Tabe‐Ojong, Ma in Paul; Du e i, Guyo Godana (2022) : A e ag o‐clus e s p o‐ poo ? E idence om E hiopia, Jou nal o Ag icul u al Economics, ISSN 1477-9552, Wiley, Hoboken, NJ, Vol. 74, Iss. 1, pp. 100-115, h ps://doi.o g/10.1111/1477-9552.12497 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/287798 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h p://c ea i ecommons.o g/licenses/by-nc/4.0/ 100 | J Ag ic Econ. 2023;74:100–115. wileyonlinelib a y.com/jou nal/jage Recei ed: 16 Janua y 2022 | Re ised: 22 Ma ch 2022 | Accep ed: 3 May 2022 DOI: 10.1111/1477-9552.12497 ORIGINAL ARTICLE A e ag o- clus e s p o- poo ? E idence om E hiopia Ma in PaulJ Tabe- Ojong1,2 | Guyo GodanaDu e i2,3 This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion- NonComme cial License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed and is no used o comme cial pu poses. © 2022 The Au ho s. Jou nal o Ag icul u al Economics published by John Wiley & Sons L d on behal o Ag icul u al Economics Socie y. 1De elopmen S a egy and Go e nance Di ision, In e na ional Food Policy Resea ch Ins i u e (IFPRI), Cai o, Egyp 2Rheinische F ied ich-Wilhelms-Uni e si ä Bonn, Bonn, Ge many 3E hiopian Ag icul u al T ans o ma ion Agency, Addis Ababa, E hiopia Co espondence Ma in Paul J Tabe- Ojong, In e na ional Food Policy Resea ch Ins i u e (IFPRI), Cai o, Egyp . Email: m.p. abe- ojong@cgia .o g Abs ac Go e nmen s and de elopmen agencies inc easingly p omo e ag o- clus e s as a pa hway o imp o ing small- holde incomes and ensu ing inclusi e u al de elopmen h ough mi iga ing p oduc ion and ma ke isks. Howe e , he e is e y limi ed empi ical e idence o suppo his p omise. We use a la ge a m household su ey o abou 4000 smallholde a me s in E hiopia g owing ce eals like e , maize, whea , mal ba ley and sesame o examine he ela ionship be ween ag o- clus e s and smallholde wel- a e and po e y. Using ins umen al a iable es ima o s, we es ablish a posi i e associa ion be ween ag o- clus e s, household income and pe capi a income. Ag o- clus e s a e also shown o educe po e y and po e y gaps. Ou esul s a e obus o e di e en ag o- clus e p oxies and al e na i e es ima o s, such as he augmen ed in e se p obabili y weigh ing es ima o . We also show ha ou indings a e unlikely o be d i en by omi ed a iable bias. Mo ing beyond a e age e ec s and in he in e es o unde s anding he e ogeneous e ec s, we use quan ile eg essions a di e en income le els. We ind ha ag o- clus e s a e associa ed wi h wel a e gains o all house- holds. Howe e , he mos signi ican gains a e obse ed o he weal hie households. Despi e his eg essi e as- socia ion, ou indings sugges ha ag o- clus e s may be use ul in making a ming mo e p o i able wi h signi ican wel a e implica ions. | 101 AGRO- CLUSTERS AND POVERTY 1 | INTRODUCTION Po e y is a pe sis en p oblem in many pa s o he wo ld, especially in sub- Saha an A ica (SSA), whe e ch onic po e y emains high, despi e some p o- poo g ow h spu s in he egion1 (Dang & Dabalen,2019). Po e y is excep ionally high in u al a eas, whe e de elopmen is challenging (Mwabu & Tho becke,2004). The as majo i y o people (app oxima ely 82%) in SSA li e in hese u al a eas and a e mainly employed in ag icul u e, which emains hei p i- ma y sou ce o li elihoods (De con & Gollin,2014). To p omo e economic g ow h and de el- opmen in u al a eas and achie e sha ed p ospe i y, many go e nmen s and de elopmen agencies ha e emphasised ag icul u e as a necessa y pa hway ou o po e y (Ba e e al.,2019). Go e nmen s and de elopmen agencies inc easingly p omo e ag o- clus e s as a pa hway o inc easing smallholde incomes and ensu ing inclusi e u al de elopmen h ough mi iga ing p oduc ion and ma ke isks (Di en,1999; Wa dhana e al.,2017). Ag o- clus e s a e de ined as a concen a ion o ag icul u al ac i i ies c ea ing income and employmen oppo uni ies in and a ound a pa icula egion (Gal ez- Nogales & Webbe ,2017). They can be e ec i e in linking smallholde a me s o eme ging ood alue chains and ma ke s, wi h signi ican wel a e implica ions (Bu ge ,1999; Poul on e al.,2010). Ag o- clus e s could make smallholde a me s p oduc i e, compe i i e and mo e gene ally a ain economies o scale. Fu he mo e, clus e s could lead o sus ainable u al de elopmen h ough hei e ec s on communi ies' economic, socio- cul u al and en i onmen al ac i i ies (B asie e al.,2007). Beyond such communi y- le el e ec s, ag o- clus e s can s imula e echnology adop ion (Jo e e al.,2019,2020), and because hey inc ease he use o imp o ed p oduc ion p ac ices, hey may also a ec ag icul u al p oduc i i y (Wa dhana e al.,2017). The ex an li e a u e highligh s inc eased in e ac ion and coope a ion as building us and leading o he impac s o ag o- clus e s (Jo e e al.,2019, 2020; Wa dhana e al.,2020). These clus e s a e usually cha ac e ised by mu ual social and economic in e ac ions be ween a me s, which help achie e p oduc ion and comme cialisa ion goals, and imp o e linkages wi h de elopmen agencies, esea ch ins i u ions and go e nmen . These clus e s could also imp o e smallholde wel a e and educe po e y by inc easing income and gene a ing employmen . Howe e , empi ical e idence o hese e ec s is p esen ly limi ed. We examine he ela ionship be ween ag o- clus e s, wel a e imp o emen s, and u al po - e y in E hiopia. E hiopia is an in e es ing case s udy because he go e nmen is using ag o- clus e s as ehicles o educing po e y and s imula ing u al de elopmen . Ou analysis is based on a a m household su ey o abou 4000 households g owing ce eals (maize, whea , e , mal ba ley, and sesame) which a e p io i y c ops o he clus e s and he coun y's main s aple c ops. We use o dina y leas squa es (OLS) and ins umen al a iable (IV) es ima o s o es ima e he associa ion be ween he sha e o alloca ed land o a clus e wi h household income, pe capi a income, po e y and po e y gaps. To he bes o ou knowledge, his is he i s s udy o examine he ela ionship be ween ag o- clus e s and po e y om a smallholde 1This g ow h was accompanied by a 9% educ ion in po e y and a g owing middle class. Abou 60% o he poo in A ica a e ch onically poo , and 40% a e in ansien po e y. KEYWORDS ag o- clus e s, E hiopia, po e y, wel a e JEL CLASSIFICATION C21; I32; Q12; Q13 102 | TABE- OJONG ANd dURETI a m household pe spec i e. Wa dhana e al.(2017) look a ag o- clus e s bu only assess he e ec s on po e y a es a a dis ic le el, looking closely a spillo e and neighbou ing spa ial e ec s. This analysis builds mo e on hei agg ega ed esul s. The es o he a icle is s uc u ed as ollows. Sec ion 2 desc ibes he concep o ag o- clus e s in he E hiopian con ex . The a m household su ey da a and a iable de ini ions a e p esen ed in Sec ion3, and Sec ion4 p esen s he es ima ion s a egy. The esul s a e p e- sen ed and discussed in Sec ion5. Policy implica ions om he s udy a e discussed in Sec ion6 and Sec ion7 concludes. 2 | THE CONCEPT OF AGRO- CLUSTERS Al hough ex an li e a u e has used he clus e concep o abou wo decades, he e m has no clea de ini ion as di e en ac o s (policy- make s, ins i u ions and economic sec o s) use i in di e en con ex s and in en ions (Gal ez- Nogales & Webbe ,2017; S e ens,2016). One o he mos commonly used de ini ions is a geog aphic concen a ion o in e connec ed i ms and ins i u ions (Po e ,1998). The o he ela ed de ini ion is ‘an agglome a ion o p oduc ion ne wo k’ whe e geog aphical p oximi y a ou s economies o scale, in e ac ions be ween di - e en ac o s, and in o ma ion lows and access o ma ke s (Gal ez- Nogales & Webbe ,2017; Wa dhana e al.,2017). In he con ex o he ag icul u al sec o , ag o- clus e is he p e e ed e m, de ined as he concen a ion o ag icul u al ac i i ies c ea ing income and employmen oppo uni ies in and a ound a pa icula egion (Gal ez- Nogales & Webbe ,2017). Acco ding o Gal ez- Nogales(2010), ag o- clus e s a e cha ac e ised by h ee pilla s o in- e ac ion be ween he ac o s in ol ed: ho izon al, e ical and suppo i e. Ho izon al in e - ac ion is a ecip ocal ela ionship be ween a me s o o he ac o s a a pa icula alue chain le el; hese ela ionships allow alue chain agen s o educe compe i ion among hemsel es and imp o e hei ba gaining posi ion in he ma ke . The e ical pilla is an in e ac ion along he alue chain, such as ag icul u al p oduc ion, ood p ocessing and ood ma ke ing, and is abou imp o ing p oduc lows and adding alue. The suppo i e linkage in ol es agen s such as esea ch ins i u ions, uni e si ies and go e nmen s. Based on he combina ion o hese in e ac ions, ag o- clus e s could c ea e syne gies, enabling small- scale p oduce s o become mo e compe i i e, o e come he cons ain s o poo access o se ices (including inpu s and in o ma ion), sha e he cos s o adop ing echnologies and p ac ices, and mi iga e a ious sou ces o isks (Gal ez- Nogales,2010). Many de eloping coun ies, especially in Asia, ha e shown signi ican success wi h ag o- clus e s (Min en e al.,2020). Zhang and Hu(2014) show he success o a po a o ag o- clus e in boos ing po a o p oduc ion and os e ing u al de elopmen in China. In he Philippines, ege able ag o- clus e s ha e been shown o imp o e access o a m inpu s and inc ease ma - ke su plus (Mon i lo e al.,2015). Ag o- clus e s ha e also been shown o educe po e y in Indonesia, highligh ing he ole o localisa ion ex e nali ies in encou aging ag icul u al ans- o ma ion (Wa dhana e al.,2017, 2020). Examining ag icul u al and ood indus y clus e s in he Uni ed S a es, Goe z e al.(2004) highligh he p o i abili y and p oduc i i y implica ions o clus e s. They u he a gue ha clus e s p o ide smallholde a me s wi h coun e ailing ma ke powe and o e egions a sou ce o compe i i e ad an age. Beyond s aple and ege- able c ops, ag o- clus e s ha e also been used in he aquacul u e sec o . In Fiji, ag o- clus e s ha e been shown o be ins umen al in inc easing ish p oduc ion, enabling a me s o enjoy he bene i s acc uing o economies o scale (Va awa e al.,2014). Using he case o sh imp a ming in he Mekong Del a egion o Vie nam, Jo e e al.(2020) examine he pa icipa ion o small- scale aquacul u e a me s in p oduc ion clus e s. They show ha clus e ing is a p om- ising a enue o os e ing in e ac ions among a me s, and inducing he adop ion o be e aquacul u e p ac ices. | 103 AGRO- CLUSTERS AND POVERTY In E hiopia, ag icul u e s ill plays a c ucial ole con ibu ing o 33% o GDP, 83.9% o expo s and 84% o employmen (Na ional Bank o E hiopia,2019). Howe e , he sec o is s ill la gely subsis ence- based, plagued wi h low p oduc i i y, ou da ed echnologies, inadequa e in as uc u e and ins i u ions, and depending on e a ic ain all, in e alia (CSA,2020). In he las decades, he coun y has emba ked on se e al ini ia i es o add ess hese challenges and imp o e smallholde wel a e and educe he po e y quagmi e (Tabe- Ojong e al.,2021, 2022). In i s mos ecen de elopmen e o s, he coun y in oduced Ag icul u al Comme cialisa ion Clus e s (he ea e ACC) as he main app oach o enhance smallholde li elihoods and d i e u al indus ialisa ion (ATA,2019; Louhichi e al.,2019). ACC is a policy in e en ion ha a ge s speci ic geog aphic loca ions and some p io i y c ops ac oss he coun y (ATA,2019; Louhichi e al.,2019). The app oach add esses he key challenges o scale and poo in eg a ion o smallholde a me s by imp o ing p oduc ion and p oduc i i y while p omo ing and in eg a ing comme cialisa ion ac i i ies (ATA,2019; Louhichi e al.,2019). Wi hin he ACC app oach, a mo e speci ic in e en ion was in oduced called Fa me P oduc ion Clus e s (FPC), whe e smallholde a me s wi h adjacen a m plo s olun a ily pool a po ion o hei land o bene i om a ge ed go e nmen suppo o selec ed c op alue chains and clus e economic agglome a ion (ATA,2019). Smallholde s pa icipa ing in he same FPC a e expec ed o coope a e by p oducing simila c ops and bene i om asso- cia ed incen i e packages including he p o ision o basic inpu s (e.g., e ilise s, imp o ed seeds, c edi , mechanisa ion, e c.), s o age and anspo acili ies, and ma ke linkages (e.g., con ac a ming) (ATA,2019). Mo eo e , smallholde a me s bene i om clus e economies o scale such as g ea e a o dabili y o mode n echnology (e.g., sha ing he o e head cos s o ac o s), s onge ba gaining powe (e.g., nego ia ing a ou able p ices o hei p oduc s), s onge ma ke linkages o se e bulk buye s o la ge- scale buye s (e.g., con ac a ming wi h la ge p ocesso s), and as e dissemina ion o bes p ac ices and ex ension se ices among a me s (ATA,2019). 3 | DATA COLLECTION AND VARIABLE MEASUREMENT 3.1 | Fa m household su ey We use a a m household su ey o smallholde households ac oss ou main egions in E hiopia: Amha a, O omia, Sou he n Na ions, Na ionali ies and Peoples' Region (SNNPR), and Tig ay. A o al o abou 4000 households we e in e iewed o e wo consecu i e pe i- ods in 2019 and 2020, as pa o he assessmen o he pe o mance o ag o- clus e s. A mul- is age sampling echnique was used o selec households in he wo su ey pe iods. In he i s s age, 75 ea men and con ol wo edas we e andomly selec ed p opo ional o size. Wo edas whe e clus e a ming has been p omo ed a e ou ea men wo edas wi h he con- ol wo edas being wo edas whe e clus e a ming does no exis a he ime o he su ey. These con ol wo edas a e simila o he ea men wo edas excep o he ac ha he clus e ing app oach has no been p omo ed and does no exis in hem. The ea men and con ol wo edas ha e simila a ming sys ems, cul i a e simila c ops and belong o simila ag o- ecological zones. In ac , some o he con ol wo edas a e a eas whe e he go e nmen in ends o scale up he clus e app oach bu , a he ime o he su ey, no clus e s ha e been es ablished he e.2 F om hese 75 wo edas, kebeles we e andomly selec ed and households we e u he andomly selec ed o in e iews. Gi en ha households we e andomly se- lec ed e en in he ea men wo edas, we in e iewed bo h households ha belong o 2Ou da a con i ms his as we do no eco d any pa icipa ion in ag o- clus e s om he con ol wo edas. 104 | TABE- OJONG ANd dURETI ag o- clus e s and hose ha do no . Abou 25% o households did no pa icipa e in ag o- clus e s in he ea men wo edas. This sampling s a egy has he ad an age o signi ican ly educing selec ion bias wi hin each wo eda (Ruml & Qaim,2021). Al hough ou ea men and con ol wo edas a e simila , we con ol o wo eda ixed e ec s in he eg essions as pa o ou empi ical s a egy. Beyond wo eda di e ences, we u he add ess he issue o sel - selec ion bias below. The in e iews we e ca ied ou by a g oup o well- ained enume a o s. The su ey was designed and adminis e ed on su ey- based able s, which enabled eal- ime quali y checks and con ols. All ac i i ies we e conduc ed, adminis e ed and supe ised by he E hiopian Ag icul u al T ans o ma ion Agency. The su eys cap u ed in o ma ion on he household socioeconomic cha ac e is ics and alue chain ac i i ies. Speci ically, he su ey included household socio- demog aphic cha ac e is ics (gende , age, educa ion and amily size), household a m asse s (land size, o - a m ac i i ies, o al p oduc ion and ma ke su plus ou pu ), and social ne wo k (neighbou pa icipa ion in ag o- clus e s, awa eness and membe ship in sel - help g oups). In o ma ion was also cap u ed on access o ex ension se ices and c edi . Al hough we ha e a wo- pe iod da a, hese da a canno be ea ed as a panel as di e en households we e in e iewed in each yea , so we ea ou da a as c oss- sec ional. Howe e , we include yea dummies in all eg essions o con ol o yea e ec s. We eached 3978 households o e he 2 yea s, bu due o some missing en ies, we only used 3969 households in he analysis. 3.2 | Measu ing ou come a iables Fou ou come a iables a e p oxies o smallholde wel a e and po e y; household income, pe capi a income, income po e y and income po e y gap. Income was measu ed in ETB3 annually, adjus ed o in la ion using consume p ice indices; i was con e ed o USD pu - chasing powe pa i y using he 2017 In e na ional Compa ison P og am con e sion a es. Pe capi a income was used o unde s and and e alua e he s anda d o li ing and quali y o li e o sampled households. Fo income po e y, he in e na ional po e y line o US$1.90 a day was used as a baseline o compa e wi h household pe capi a income. Households wi h a pe capi a income abo e his line we e gi en a ze o alue, while households below we e gi en a alue o one. Simila ly, he income po e y gap was calcula ed by sub ac ing he pe capi a income o households om he in e na ional po e y line, di ided by he po e y line. Some households had pe capi a incomes abo e he po e y line, which led o some nega i e gaps; hus, ze os we e assigned o hese households. Calcula ing he po e y gap his way cons ains he alues be ween ze o and one, enabling compa ison. 3.3 | Measu ing ag o- clus e s Pa icipa ion in ag o- clus e s was measu ed using h ee di e en p oxies. Fi s , pa icipa ion was measu ed as a dummy, which akes he alue o one o households ha belong o ag o- clus e s and ze o o he wise. Pa icipa ion in ag o- clus e s equi e households o o e a mini- mum o 0.25 hec a es o land and g ow some key c ops. Thus, e e y a me 's land con ibu ion is used as a second measu e o ag o- clus e pa icipa ion. 3ETB s ands o E hiopian bi , which is he E hiopian cu ency. | 105 AGRO- CLUSTERS AND POVERTY Using he amoun o land alloca ed o a clus e makes compa ison be ween households no possible. As he amoun depends on he o al landholding, we use he sha e o o al land con ibu ed. 4 | ESTIMATION STRATEGY Gi en ha we ha e c oss- sec ional da a, we es ima e he ollowing eg ession model: whe e Yi is he wel a e and po e y indica o o household i , Ci is he sha e o land alloca ed o he clus e by he household, Xi is a ec o o con ol a iables, and 𝜀i is he s ochas- ic e o e m. Two di e en models a e es ima ed o each ou come a iable; we con ol o a ious a m household cha ac e is ics associa ed wi h u al po e y h ough channels o he han pa icipa ing in ag o- clus e s. We use he OLS es ima o o all he con inuous dependen a iables. Fo income po e y, which is a dummy, we also use a linea p oba- bili y model, which usually a oids iden i ica ion by unc ional o m common in logi and p obi models. The main pa ame e o in e es om Equa ion(1) is 𝛿 , which indica es he ela ionship be ween ag o- clus e s and u al po e y. We hypo hesize a posi i e ela ionship be ween ag o- clus e s and income measu es, and a nega i e ela ionship wi h po e y and po e y gap; he pa ame e es ima e should be posi i e o income indica o s and nega i e o po e y ou - comes. Assuming s ic exogenei y, he OLS es ima ions should p o ide unbiased es ima es o he ela ionship be ween ag o- clus e s and u al po e y. Howe e , s ic exogenei y does no e lec eali y, and he sha e o land alloca ed o a clus e may be po en ially endogenous. Endogenei y o ag o- clus e s may a ise om measu emen e o , e e se causali y and unob- se ed he e ogenei y. In he case o measu emen e o , i is always challenging o claim he accu acy o he da a gene a ing p ocess. S ill, we a e ce ain ha clus e pa icipa ion was well cap u ed wi h he ac ual amoun o land alloca ed by a me s, gi en ha hese p ocesses we e well supe ised and moni o ed. Be o e joining he clus e s, he ac ual amoun o land con ibu ed by a me s o he clus e s was measu ed using GPS echniques. Fo unobse ed he e ogenei y, i is pos- sible ha unobse ed ac o s, such as isks, p e e ences and manage ial abili ies, may d i e he amoun o land ha households alloca e o clus e s and be co ela ed wi h po e y. As we only use c oss- sec ional da a, i is di icul o sa is ac o ily con ol o his e en uali y beyond including di e en con ols and obse ing he s abili y o he coe icien s. Addi ionally, we ollow Os e (2019) o e alua e he obus ness o he es ima ed coe icien s o omi ed a iable bias. Finally, he e could be e e se causali y issues be ween ag o- clus e s and he wel a e and po e y indica o s. Alloca ing mo e land o clus e s may inc ease income and educe po - e y h ough he associa ed bene i s o pooling esou ces om o he households and enjoying economies o scale when i comes o inpu pu chases and comme cialisa ion. Households ha a e gene ally mo e ad an aged in income may alloca e mo e land han hei pee s. Bu can inc ease in income lead o land expansion? This ela ionship may only hold o households who alloca e only small sha es o hei landholdings. The e may be li le o no income– land alloca ion e ec o households ha alloca e mos o all o hei lands o he clus e limi ing h ea s o e e se causali y. Ne e heless, we employ IV es ima o s o con ol o any esidual endogenei y especially gi en non- andom sel - selec ion in o ag o- clus e s in he ea men wo edas. We speci y wo ins umen s: awa eness o he exis ence o ag o- clus e s and neighbou pa icipa ion in (1) Yi =𝛽 0 +𝛿C i +𝛾 � X i +𝜀 i, 106 | TABE- OJONG ANd dURETI ag o- clus e s. These ins umen s a e mo i a ed by he p emise ha ne wo ks ease in o ma ion low and educe he ba ie s acing a me s in u al a eas (Di Falco e al.,2020). Gi en ha ag o- clus e s a e a new concep in he s udy a ea, awa eness o hei exis ence and unc ioning a e necessa y p e- condi ions o pa icipa ion. Awa eness o ag o- clus e s is signi ican ly co ela ed and posi i ely de e mines pa icipa ion in ag o- clus e s (R2=26%, p= 0.00). Neighbou hood pa icipa ion is also posi i ely ela ed o pa icipa ion in ag o- clus e s (R2=26%, p=0.06). These wo indings al eady alida e he ins umen s based on he ele ance condi ion ( ull esul s a e p o ided in he Appendix S1). Examining he second condi ion o ins umen exogenei y equi es e i ying i he in- s umen s di ec ly a ec po e y; in ui i ely, no e ec s a e an icipa ed, excep h ough ag o- clus e s. Being awa e o an ag o- clus e o knowing i a pee is pa icipa ing in hese clus e s, seemingly does no a ec ou ou comes. Apa om concep ually mo i a ing his condi ion, he e is ypically no s a is ical es o his. Howe e , because his s udy has wo ins umen s, Woold idge's sco e es o o e - iden i ying es ic ions, which is he e oscedas ici y- obus is conduc ed4 (Woold idge,1995). As shown in he Appendix S1, s a is ically insigni ican es ima es a e ob ained. Thus, we ail o ejec he null hypo hesis ha he ins umen s a e alid. 4.1 | E ec he e ogenei y Gi en ha he e ec s o ag o- clus e s may ha e a he e ogeneous associa ion wi h household wel a e and u al po e y, we u he examine his he e ogenei y and es ablish which g oup o households bene i mos using quan ile eg essions. Using he same conno a ions as in Equa ion(1) abo e, we es ima e he ollowing eg ession model: Xi is a ec o o explana o y a iables, including ag o- clus e s. ( Y i| X i) is he condi ional quan ile o Yi a quan ile 𝜑 . We es ima e he associa ion be ween ag o- clus e s and u al po e y using nine di e en quan iles ( 𝜕𝜑 = 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.70, 0.80 and 0.90). 𝜕𝜑 = 0.10 ep esen s he poo es g oup o households. 5 | RESULTS AND DISCUSSION 5.1 | Cha ac e ising he sample Table1 p esen s he summa y s a is ics o he ou come a iables and he explana o y a iables used in he eg ession amewo k. Households in he s udy a ea epo an annual income o US$1340, which alls o US$235 when iewed pe capi a. App oxima ely 60% o he house- holds can be e med poo based on he US$1.90 in e na ional po e y line, wi h an associa ed income po e y gap o 35%. Rega ding pa icipa ion in ag o- clus e s, abou 57% o households in he s udy a ea be- long o ag o- clus e s, whe e hey alloca e an a e age o 0.60 hec a es o land o he clus e s. Rega ding hei o al landholdings, which a e abou 2.25 hec a es, households alloca e a sha e o 0.27 o he clus e . The e a e abou 17 membe s pe clus e , and he o al land pe clus e is app oxima ely 12 hec a es. Mos household heads a e middle- aged, a e aging abou 43 yea s 4Though no an ins umen exogenei y es , an o e - iden i ica ion es p o ides s a is ical alida ion ha addi ional ins umen s a e exogenous. (2) Yi =X� i 𝜕 𝜑 +𝜀 i , ( y i| X i) =X� i 𝜕 i | 107 AGRO- CLUSTERS AND POVERTY o age. Households a e mos ly male headed, wi h a household size o abou se en membe s. Ex ension access in he s udy a ea is widesp ead (90%), and 73% o household heads ha e achie ed p ima y educa ion. Signi ican di e ences a e obse ed be ween households based on income and po e y ou comes (Table2). Speci ically, households in ag o- clus e s ha e highe household and pe capi a incomes han hei non- pa icipa ing coun e pa s. Simila ly, such households appea less impo e ished unde bo h income po e y and income po e y gap measu es. Households in clus e s gene ally ha e mo e land, and hei heads a e younge han hei pee s who do no belong o ag o- clus e s. Al hough hese esul s o e some in e es ing insigh s, hey do no con ol o po en ial con oundings. TABLE 1 Summa y s a is ics Mean S d. de . Ou come a iables Income (US$) 1340.94 2724.22 Pe capi a income (US$) 235.40 576.91 Income po e y (dummy) 0.61 0.48 Income po e y gap (0– 1) 0.35 0.35 Va iables o in e es Ag o- clus e s (dummy) 0.57 0.49 Plo alloca ed o clus e (hec a es) 0.60 0.87 Ag o- clus e s (0– 1) 0.27 0.31 O he con ol a iables To al clus e size (hec a es) 11.46 16.88 Clus e size (numbe ) 16.63 24.56 Age o household head (yea s) 42.67 11.02 P ima y educa ion (dummy) 0.73 0.44 Household head is emale (dummy) 0.09 0.29 Household size (numbe ) 6.51 2.44 Landholding (hec a es) 2.25 1.96 G oup membe ship (dummy) 0.37 0.48 C edi access (dummy) 0.29 0.45 Ex ension access (dummy) 0.90 0.29 S o age acili ies (dummy) 0.59 0.49 O - a m income (dummy) 0.40 0.49 Whea (dummy) 0.29 0.45 Te (dummy) 0.14 0.34 Sesame (dummy) 0.07 0.26 Mal ba ley (dummy) 0.14 0.34 Maize (dummy) 0.34 0.47 Neighbou hood pa icipa ion (dummy) 0.22 0.42 Awa eness o ag o- clus e s (dummy) 0.78 0.41 Obse a ions 3969 3969 114 | TABE- OJONG ANd dURETI Bu ge , K. (1999) Clus e ing o small ag o- p ocessing i ms in Indonesia. The In e na ional Food and Ag ibusiness Managemen Re iew, 2(3– 4), 289– 299. h ps://doi.o g/10.1016/S1096 - 7508(01)00033 - 7 CSA. (2020) A ea and p oduc ion o c ops, and a m managemen p ac ices. Addis Ababa, E hiopia: CSA. Dang, H.- A.H. & Dabalen, A.L. (2019) Is po e y in A ica mos ly ch onic o ansien ? E idence om syn- he ic panel da a. The Jou nal o De elopmen S udies, 55(7), 1527– 1547. h ps://doi.o g/10.1080/00220 388.2017.1417585 De con, S. & Gollin, D. (2014) Ag icul u e in A ican de elopmen . Theo ies and s a egies. Annual Re iew o Resou ce Economics, 6(1), 471– 492. h ps://doi.o g/10.1146/annu e - esou ce- 10091 3- 012706 Di Falco, S., Doku, A. & Mahajan, A. (2020) Pee e ec s and he choice o adap a ion s a egies. Ag icul u al Economics, 51(1), 17– 30. h ps://doi.o g/10.1111/agec.12538 Di en, M. (1999) Dai y clus e s in La in Ame ica in he con ex o globaliza ion. The In e na ional Food and Ag ibusiness Managemen Re iew, 2(3– 4), 301– 313. h ps://doi.o g/10.1016/S1096 - 7508(01)00045 - 3 Funk, M.J., Wes eich, D., Wiesen, C., S ü me , T., B ookha , M.A. & Da idian, M. (2011) (2011): doubly obus es ima ion o causal e ec s. Ame ican Jou nal o Epidemiology, 173(7), 761– 767. h ps://doi.o g/10.1093/aje/ kwq439 Gal ez- Nogales, E. (2010) Ag o- based clus e s in de eloping coun ies: s aying compe i i e in a globalized economy. In Ag icul u al Managemen , Ma ke ing and Finance Occasional Pape (FAO) (25). Gal ez- Nogales, E. & Webbe , M. (Eds.). (2017) Te i o ial ools o ag o- indus y de elopmen . A sou cebook. Ag obased clus e s. Wi h assis ance o K. mu phy. Rome, I aly: FAO. Goe z, S.J., Shields, M. & (Cindy) Wang., Q. (2004) Ag icul u al and ood indus y clus e s in he no heas U.S.: echnical epo . Regional u al de elopmen pape No. 26. In: The no heas egional Cen e o Ru al De elopmen , College o Ag icul u al Sciences. Uni e si y Pa k, PA: The Pennsyl ania S a e Uni e si y. Jo e, O.M., de V ies, J.R., Kle kx, L. & Poo lie , P.M. (2020) Why a e clus e a me s adop ing mo e aquacul u e echnologies and p ac ices? The ole o us and in e ac ion wi hin sh imp a me s' ne wo ks in he Mekong Del a, Vie nam. Aquacul u e, 523(July), 735181. h ps://doi.o g/10.1016/j.aquac ul u e.2020.735181 Jo e, O.M., Poo lie , P.M. & Kle kx, L. (2019) To clus e o no o clus e a me s? In luences on ne wo k in e - ac ions, isk pe cep ions, and adop ion o aquacul u e p ac ices. Ag icul u al Sys ems, 173(1), 151– 160. h ps:// doi.o g/10.1016/j.agsy.2019.02.011 Louhichi, K., Temu sho, U., Colen, L., & Paloma, S.G. (2019) Upscaling he p oduc i i y pe o mance o he ag icul- u al comme cializa ion clus e ini ia i e in E hiopia. Luxembou g: Publica ions O ice o he Eu opean Union. Min en, B., Mohammed, B. & Tam u, S. (2020) Eme ging medium- scale enan a ming, gig economies, and he COVID- 19 dis up ion: he case o comme cial ege able clus e s in E hiopia. The Eu opean Jou nal o De elopmen Resea ch, 32(5), 1402– 1429. Mon i lo , M. O., Ba , P. J. and Mu ay- P io , R. (2015) Socio- economic impac o clus e a ming o small- holde a me s in sou he n Philippines. Cu in Uni e si y o Technology, GPO box U1987, Pe h WA 6845, Aus alia. Mwabu, G. & Tho becke, E. (2004) Ru al de elopmen , g ow h and po e y in A ica. Jou nal o A ican Economies, 13(suppl_1), i16– i65. h ps://doi.o g/10.1093/jae/ejh015 Na ional Bank o E hiopia. (2019) Annual epo s 2019– 2020. Addis Ababa, E hiopia: Na ional Bank o E hiopia. Os e , E. (2019). Unobse able selec ion and coe icien s abili y. Theo y and E idence. Jou nal o Business & Economic S a is ics, 37(2), 187– 204. h ps://doi.o g/10.1080/07350 015.2016.1227711 Po e , M.E. (1998) Clus e s and he new economics o compe i ion. Ha a d Business Re iew, 76(6), 77– 90. Poul on, C., Do wa d, A. & Kydd, J. (2010) The u u e o small a ms. New di ec ions o se ices, ins i u ions, and in e media ion. Wo ld De elopmen , 38(10), 1413– 1428. h ps://doi.o g/10.1016/j.wo ld de .2009.06.009 Ruml, A. & Qaim, M. (2021) New e idence ega ding he e ec s o con ac a ming on ag icul u al labo use. Ag icul u al Economics, 52(1), 51– 66. S e ens, V.C. (2016) Po en ials o ag a ian clus e de elopmen o imp o ing smallholde 's income– a case s udy om he SAGCOT ini ia i e in Tanzania. Bachelo . Uni e si y o Cologne, Cologne. A ailable online a h ps://geog aphie.uni- koeln.de/si es/ geog aphie/ use _uploa d/Bache lo a bei _S e ens.pd . Tabe- Ojong, M.P., Hause , M. & Mausch, K. (2022) Does ag icul u al comme cialisa ion inc ease asse and li e- s ock accumula ion on smallholde a ms in E hiopia? Jou nal o De elopmen S udies, 5(4), 79– 544. h ps://doi. o g/10.1080/00220 388.2021.1983170 Tabe- Ojong, M.P., Mausch, K., Woldeyohanes, T.B. & Heckelei, T. (2021) Th ee hu dles owa ds comme cializa ion: in eg a ing subsis ence chickpea p oduce s in he ma ke economy. Eu opean Re iew o Ag icul u al Economics, 5(4), 79. h ps://doi.o g/10.1093/e ae/jbab023 Va awa, J.T, Picke ing, A.S. and Singh, S. (2014) Small a me g oups ( a m clus e s) as a s a egy o up- scale ilapia ish a m p oduc ion in Fiji Islands. Sec e a ia o he Paci ic Communi y, Fiji and Fiji Minis y o Fishe ies and Fo es s. Wa dhana, D., Ihle, R. & Heijman, W. (2017) Ag o- clus e s and u al po e y. A spa ial pe spec i e o Wes Ja a. Bulle in o Indonesian Economic S udies, 53(2), 161– 186. h ps://doi.o g/10.1080/00074 918.2017.1298722 | 115 AGRO- CLUSTERS AND POVERTY Wa dhana, D., Ihle, R. & Heijman, W. (2020) Fa me coope a ion in ag o- clus e s. E idence om Indonesia. Ag ibusiness, 36(4), 725– 750. h ps://doi.o g/10.1002/ag .21637 Woold idge, J.M. (1995) Sco e diagnos ics o linea models es ima ed by wo- s age leas squa es. In: Maddala, G.S., Phillips, P.C.B. & S ini asan, T.N. (Eds.) Ad ances in Econome ics and Quan i a i e Economics: Essays in Hono o P o esso C. R. Rao. Ox o d: Blackwell, pp. 66– 87. Woold idge, J.M. (2007) In e se p obabili y weigh ed es ima ion o gene al missing da a p oblems. Jou nal o Econome ics, 141(2), 1281– 1301. h ps://doi.o g/10.1016/j.jecon om.2007.02.002 Zhang, X. & Hu, D. (2014) O e coming successi e bo lenecks: he e olu ion o a po a o clus e in China. Wo ld De elopmen , 63, 102– 112. SUPPORTING INFORMATION Addi ional suppo ing in o ma ion may be ound in he online e sion o he a icle a he publishe ’s websi e. How o ci e his a icle: J Tabe- Ojong, M.P. & Du e i, G.G. (2023) A e ag o- clus e s p o- poo ? E idence om E hiopia. Jou nal o Ag icul u al Economics, 74, 100– 115. A ailable om: h ps://doi.o g/10.1111/1477- 9552.12497