Mekelle University College of Business and Economics Department of Economics

Transcription

1 Mekelle Unversty College of Busness and Economcs Department of Economcs The Impact of Camel Transportaton on the Lvelhood of Pastoralsts: In Berahle Woreda, Afar Regonal State of Ethopa BY: SELAMAWIT TEKLU ARAYA A Thess Submtted n Partal Fulfllment of the Requrements for the Degree of MASTER OF SCIENCE IN ECONOMICS Wth Specalzaton n DEVELOPMENT POLICY ANALYSIS Prncpal Advsor: Kdanemaram G. Egzabher (PhD.) Co-advsor: Abrha Megos (MSc) June, 2014 Mekelle, Ethopa

2 DECLARATION I, the undersgned, declare that ths thess enttled The Impact of Camel Transportaton on the Lvelhood of Pastoralsts: In Berahle Woreda, Afar Regonal State of Ethopa, submtted to department of economcs, college of busness and economcs, Mekelle Unversty s my authentc work. All the sources that I have used n ths project work have been acknowledged and ndcated as a reference. I solemnly declare that ths thess s my own work and has not submtted earler to any other nsttuton anywhere for the award of any other degree, dploma or certfcate. SELAMAWIT TEKLU ARAYA Sgnature Date Place: Department of Economcs, College of Busness and Economcs, Mekelle Unversty, Mekelle, Tgray, Ethopa

3 CERTIFICATION Ths s to certfy that ths thess enttled The Impact of Camel Transportaton on the Lvelhood of Pastoralsts: In Berahle Woreda, Afar Regonal State of Ethopa s an authentc work of Ms. Selamawt Teklu who carred out the research under my gudance. Furthermore, to the best of my knowledge the work reported here does not form part of any project report or thess on the bass of whch a degree or award was conferred on an earler occason on ths or any canddate. Hence, I recommend that t be submtted as fulfllng a thess requrement. PRINCIPAL ADVISOR: - KIDANEMARIAM G.EGZIABHER (PhD.) Sgnature Date CO-ADVISOR: - ABRHA MEGOS (MSc) Sgnature Date Mekelle Unversty College of Busness and Economcs Department of Economcs

4 ABSTRACT The man objectve of ths study was to assess the mpact of camel salt transportaton servce on the lvelhood of pastoralsts n Berahle woreda, Afar regonal state of Ethopa. A cross sectonal prmary data of 250 sample households (ncludng 115 camel owners and 135 camel non-owners) was used for analyss. Propensty score matchng (PSM) method was mplemented to capture the lvelhood contrbuton of camel salt transportaton servce for the camel owner pastoral households n the study area. Four matchng algorthms (Nearest- Neghbour, Kernel, Radus and Stratfcaton) were used for estmaton and ther result was consstent whch shows ther robustness. Income, lvestock ownershp and fxed asset formaton of the households were used as lvelhood ndcators to compare the lvelhood dfferences between the camel owner and non-owner households. Accordngly, a postve and sgnfcant dfference was found n two of the lvelhood ndcators.e., ncome and lvestock ownershp, but not fxed asset holdngs between the two groups of households. The paper has also tred to dentfy the man determnant factors nfluencng the camel rent decson of the camel owner pastoral households usng ordered probt regresson model. The result reveals that, lteracy of the household head and number of camels owned by the household are the factors affectng household head s decson to rent out hs/her camels for other salt transporters or not. Another focus of the study was, dentfyng the man benefcares of the salt mne among the salt trade value chan partcpants and calculatng the proft percentage share of the salt transporters, usng descrptve statstcs. Based on the analyss, the salt transporters were found to be the man benefcares among the salt trade value chan partcpants. Keywords:- Camel, Transportaton, Lvelhood, Propensty Score Matchng (PSM), Camel Rent Decson, Ordered Probt, Proft

5 ACKNOWLEDGEMENTS Thanks to almghty GOD, the lord of the worlds, for all the prvleges he has been gvng me and for pushng me forward to where I am standng today. I would lke to extend my heartfelt grattude to my advsor Dr. Kdanemaram G.Egzabher for hs wse gudance and ndspensable advce provded me n wrtng ths thess. Furthermore, I really apprecate the kndness and overall support that he gave me wthout hestaton throughout my study. Fensten Internatonal Center, Tufts Unversty s hghly acknowledged for ts generous fnancal assstance for the successful completon of the study. It s my pleasure to express my thanks to Dr. Dawt Abebe and Dr. Berhanu Admassu (senor polcy advsor) for ther gudance and nformaton provson on how to work. I apprecate Sheck Momn Mohammed Beshr (the deputy admnstrator of Klbet Rasu or zone 2 of Afar regonal state), Ato Muftah Adem Ga as (the deputy admnstrator of Berahle Woreda), all the staffs of the zone and woreda admnstraton and Ato Abdela Yasn (admnstrator of Assaele Salt Trade Cooperatve) for ther cooperaton n the tme of data collecton. I would lke to express my deepest grattude to my frends Yonas G/Egzabher and Halefom Ygzaw for the generous assstance and care they gave me throughout my study. I am ndebted to Mr. Tefera Kebede, head of Department of Economcs, for hs all-round support durng my study perod. My specal thanks go to my famly: Meselech Kassa, Genet Abreha, Tgst G/Egzabher, Bnam G/Egzabher and Shewt Teklu for ther encouragement, support, care, love and advce whch helped me to make my dreams come true. The utmost apprecaton goes to my mother, Meselech Kassa, wthout whom my effort would have been worth nothng. Lastly, I am ndebted to ndvduals and nsttutons whose valuable contrbutons have enabled the completon of ths study. Especally, grattude goes to the ntervews and focus groups met durng the study and the communty of the area for ther hosptalty and provson of crucal nformaton. v

6 ACRONYMS AND ABREVIATIONS AE ATT CIA COMESA CSA E.C. EDH ETB FAO FDRE FGD GDP HH Kg KM LPPS MLM NNM PS PSM RM SM TLU Adult Equvalence Average Treatment Effect for the Treated Condtonal Independence Assumpton Common Market for Eastern and Southern Afrca Central Statstcal Authorty Ethopan Calendar Ethopan Demography and Health Ethopan Brr Food and Agrculture Organzaton of Unted Natons Federal Democratc Republc of Ethopa Focus Group Dscusson Gross Domestc Product Household Klogram Kernel Matchng Lokht Pashu-Palak Sansthan (s Pastoralst Welfare Insttute ) Maxmum Lkelhood Estmaton Method Nearest Neghbour Matchng Propensty Score Propensty Score Matchng Radus Matchng Stratfcaton Matchng Tropcal Lvestock Unt v

7 Table of contents page DECLARATION CERTIFICATION..... ABSTRACT ACKNOWLEDGEMENTS v ACRONYMS AND ABREVIATIONS...v Table of Contents v Lst of Tables......x Lst of Fgures......x Lst of Annexes.....x CHAPTER ONE INTRODUCTION Background of the Study Statement of the Problem Research Questons Objectves of the Study General Objectve Specfc Objectves Sgnfcance of the Study Scope of the Study Lmtaton of the Study Organzaton of the Study...7 CHAPTER TWO LITERATURE REVIEW Theoretcal Revew Termnology and Defntons Anmal Tracton and Its Benefts Camels and Ther Transportaton Servce Challenges to Camel Producton and Causes of Camel Loss Emprcal Revew v

8 CHAPTER THREE METHODOLOGY OF THE STUDY Descrpton of the Study Area Data Source and Samplng Procedure Data Source and Data Type Samplng Procedure and Sample Sze Questonnare Desgn Methods of Data Analyss Descrptve Method of Data Analyss Econometrc Method of Data Analyss Econometrc Framework and Estmaton Strategy Propensty Score Matchng Ordered Probt Model Descrpton of Varables Used n the Analyss CHAPTER FOUR RESULTS AND DISCUSSIONS Descrptve Analyss Households Soco-Economc and Demographc Characterstcs Households Salt Transportaton and Camel Rent Experences Income of Camel Renters and Rsk Sharng Between the Two Partes The Proft Proporton of the Salt Transporters Out of the Total Proft of the Salt Trade Value Chan Partcpants Returns from Camels Salt Transportaton Servce Employment Contrbuton of Camel Salt Transportaton Servce The Contrbuton of Camel Salt Transportaton Servce for the Natonal Economy Econometrc Analyss Propensty Score Estmates Impact Estmates Estmaton Results of the Senstvty Analyss Estmaton Results and Dscussons of the Ordered Probt Model v

9 CHAPTER FIVE CONCLUSIONS AND POLICY RECOMMENDATIONS Conclusons Polcy Recommendatons Suggestons for Further Research References v

10 Lst of Tables Table 3.1 Descrpton of Dependent and Independent Varables Used for Analyss n the Study 40 Table 4.1 Descrpton of Households Soco-Economc and Demographc Characterstcs (Dummy and Categorcal Varables) Table 4.2 Summary of Basc Households Soco-Economc and Demographc Characterstcs 47 Table 4.3 Salt Transportaton and Camel Rent Experences of the Camel Owner Pastoral Households Table 4.4 Revenue, Cost, Proft and Proft Percentage Share of Salt Trade Value Chan Partcpants Table 4.5 Logstc Estmaton Results for the Lkelhood of Households Camel Ownershp...62 Table 4.6 ATT Results for Income of Households Table 4.7 ATT Results for TLU of Households...64 Table 4.8 ATT Results for Fxed Asset Value of Households...65 Table 4.9 Smulaton-Based Senstvty Analyss Results Table 4.10 Ordered Probt Estmaton Results for the Households Camel Rent Decson...67 x

11 Lst of Fgures Fgure 3.1 Locaton Map of the Study Area (Berahle Woreda) Fgure 4.1 The Proporton of Camel Owner Pastoral Households Under the Three Categores of Camel Rent Decson...49 x

12 Lst of Annexes Annexes..77 Annex I. Descrpton of Varables Used for Analyss n the Study Annex II. Adult Equvalence (AE) Converson Factors...80 Annex III. Converson Factors Used to Estmate Tropcal Lvestock Unt (TLU)...80 Annex IV. Summary Statstcs of Varables Used for Analyss n the Study...81 Annex V. Correlaton Matrx.83 Annex VI. Logstc Estmaton Results for Households Camel Ownershp 84 Annex VII. Average Treatment Effect for the Treated Result of Income of Households...88 Annex VIII. Average Treatment Effect for the Treated Result of Lvestock Ownershp (TLU) of Households..90 Annex IX. Average Treatment Effect for the Treated Result of Fxed Asset Formaton of Households...92 Annex X. Smulaton-Based Senstvty Analyss Results.94 Annex XI. Ordered Probt Estmaton Results for Households Camel Rent Decson 99 Annex XII. Questonnare Used for Survey. 101 Annex XIII. Sample Pcture of an Adult Loaded Camel.118 Annex XIV. Sample Pcture of the Salt Store Found n Berahle Woreda that Serve as a Market n Whch the Salt Transported by the Salt Transporters s Exchanged and Stored x

13 CHAPTER ONE 1. INTRODUCTION 1.1 Background of the Study Pastoralsm, also known as anmal husbandry, s a socal and economc system practced by people dwellng n ard and sem-ard envronments n whch rasng and herdng of lvestock such as camels, goats, cattle, and sheep usng manly tradtonal knowledge s the prmary economc actvty of the socety. Pastoral producton systems are those n whch at least 50 percent of the gross ncomes of households (.e. the value of market producton and the estmated value of subsstence producton consumed by households) come from pastoralsm or ts related actvtes, or else, where more than 15 percent of households food energy consumpton nvolves the mlk or dary products they produce (Swft, 1988). In Afrca, partcularly n the ard and sem-ard lands of the Horn and East Afrca, pastoralsm s one of the oldest, most reslent and most adaptve lvelhood system practced by mllons of peoples. The pastoral lands, n whch these pastoralsts resde, cannot support sustaned and relable agrculture because of the extreme varablty n tme and place of weather patterns.e. ran. These dffcultes make pastoralsm to have a moble aspect, movng the herds n search of fresh pasture and water (n contrast to pastoral farmng, n whch non-nomadc farmers grow crops and mprove pastures for ther lvestock). Pastoral households use clmatc varablty to ther advantage to maxmze the productvty of ther herds by lvestock moblty and selectve breedng of anmals wthout degradng the envronment n the rangelands (COMESA, 2009). In Ethopa, pastoral and agro pastoral areas cover about 61 percent of the country s land mass on whch over 12 mllon people are lvng. These areas are characterzed by ther poor nfrastructural development and the hghest ncdence of poverty and household food nsecurty. The number of lvestock, whch s the man resource base for the pastoral households, contnues to declne due to recurrent droughts, anmal dseases and conflcts (FDRE Mnstry of Federal Affars, 2008). 1

14 Ethopa has the largest number of domestc lvestock n Afrca and much of t orgnates from the country s pastoral zones. These areas contan approxmately 30 percent of the natonal populaton or 9.3 mllon cattle, 52 percent or 12.4 mllon sheep, 45 percent or 8.1 mllon goats and close to 100 percent or 1.8 mllon camels (Catley, 2009). Ths lvestock populaton contrbutes 15 to17 percent of the country s GDP and 35 to 49 percent of agrcultural GDP, and 37 to 87 percent of the household ncomes (Sntayehu et al., 2010). In these pastoral zones, camel s a prmary stock and s hghly valued. Camel s a large and strong anmal of the desert and t s one of the powerful transportaton anmals. It can go n such area where transportaton vehcle can t go and carry people and heavy loads to the places where there are no roads. Furthermore, t s the camel's ablty to wthstand extreme temperatures and harsh condtons. It can floursh where no other domestc anmal can survve whch make t so valuable n ard and hot condtons durng the day and cold temperatures at nght. In tmes of water scarcty, t can endure wthout water for more than two weeks. It s generally used as pack anmal and not so much as a free rde. Camel s unversally hghly valued and provdes socal standng for ts owner. Afar regon 1 s one of the four major pastoral regonal states n Ethopa located n north eastern part of the country. It s a regon n whch ts all woredas are pastoral woredas. The Afar pastoralsts rase mxed speces of prmary lvestock, usually camels and cattle, and keep supplementary herds of goats and sheep. Smlar to the country s general context, camel s also best suted to the ard desert areas of Afar regon. It s a prmary stock and status ndcator and represent the nomadc captal wealth of the Afar socety and s essentally rased and kept for ths reason (Gunand, 2000). Accordng to Sntayehu et al. (2010), Afar regon consttutes about 22.5 percent of the total camel populaton of the country. 1 The short notaton Afar regon s used to represent Afar Regonal State of Ethopa throughout the paper. 2

15 1.2 Statement of the Problem Lvestock has an economc and socal mportance both at natonal and nternatonal levels partcularly n pastoral areas. Anmals assst humans n carryng out dfferent tasks n agrculture, transport, rrgaton, buldng ndustry, etc. In transportaton, they are used for pullng carts and loads over a surface, loggng and carryng loads as pack anmals (Smalenga and Joubert, 1997). Accordng to FAO report (n.d), n many parts of the world, rural trade s asssted through anmal-powered transport (on farm, marketng, rdng, pack transport). Anmals save household (partcularly women and chldren) tme and effort by carryng water and fuel wood. Anmal power can also be used for water-lftng, mllng, loggng and land excavaton and road constructon. To carry out these tasks, many dfferent types of anmals are employed, partcularly cattle (oxen, bulls and cows), camels, buffaloes, mules, donkeys and horses. Farmers wth anmal transport (carts or pack anmals) have wder contacts wth traders. The resultng enhanced market access allows them to ncrease ther producton and also ther proft. Wth anmal transport, greater use s made of manure and crop resdues, whch ncreases overall farm producton. Anmal power can provde mportant local `feeder' transport between farms and roads, to complement motorzed road transport systems. The development of effcent anmal-based transport s often constraned by lmted supples of carts and captal or credt for acquston. However, anmal-based transport s usually very proftable (Ibd). Camels, known as "shps of the desert", are the only desert anmals those can carry heavy loads of goods and travel for a long perod of tme wthout food or water. They can carry loads weghng kg over long dstances and 450 kg over short dstances and are convenent n tmes of water scarcty to travel n dry lands (Heanvng and Zhongmn, n.d). Camels have socal and economc beneft for ther owners. Mllons of people who lve n Afrca depend on camels to supply most of ther needs. In lands at the edge of the deserts, camels pull ploughs, turn water wheels to rrgate felds and carry gran to market places. Deep n the deserts, camels are almost the only source of transportaton, food, clothng, and shelter (Farah and Fscher, 2004). In addton, accordng to Schwartz (1988), durng the recent famnes n the vast dry areas n Afrca, camels were frequently used to carry relef food to remote, otherwse naccessble locatons. 3

16 Nowadays, tradtonal modes of transportaton are dmnshng ther use and ther mportance due to technologcal advancement and the role of pack anmals transportaton s substtutng by the motorzed means of transportaton. But, even f pack anmals transportaton s substtutng by modern means of transportaton, n many countres of the world partcularly n the LDCs stll t s mportant means of transportaton. Partcularly n Ethopa, despte the ntroducton of a motorzed transportaton, anmal power partcularly camels reman an mportant means of transportaton by provdng a cheap and relable alternatve for short, medum and long dstance transport of many knds of goods such as gran, salt, domestc water and fuel, household tools and anmal feed n the ard and semard regons of the country. Camel n Ethopa has multple uses. It serves as a: Source of ncome Means of transportaton Source of food by provdng mlk and meat for household consumpton Lftng of water and Powerng of ol mlls In Afar regon, camel has a unque functon than n any other regons of the country.e., t s useful to transport salt, whch s one of the major mnerals found n the regon, from ts source where t s extracted to the market where t s sold. Ths salt trade usng camels s one of the means of ncome generaton for the camel owner pastoralsts n the area and others partcpatng n the trade chan of the salt. Hence, among the transportaton servces that a camel provdes to the pastoralsts, the researcher s man concern s ts servce n transportng salt from the mnng area to market place. Even f camel s the man means of transportng salt that the salt transporters use n the study area, stll there s no as such well documented emprcal study whch explores the economc contrbuton of the camel salt transportaton servce n general and ts mpact on the lvelhoods of the camel owners n partcular, though t s an ssue whch demands a due concern. Ths gap motvates the researcher to put some effort and fll ths nformaton gap by nvestgatng the factors those determne camel ownershp of households, the mpact of camel salt transportaton servce on the lvelhoods of the camel owner pastoral households and the proft proporton of the salt transporters out of the total proft of the salt trade value chan partcpants n Berahle woreda of Afar regon whch s untouched area. 4

17 1.3 Research Questons The research was attempted to answer the followng questons: What are the determnant factors those affect households camel ownershp n the study area? Does camel salt transportaton servce has an mpact on the lvelhoods of the camel owner pastoral households? What are the factors behnd, whch affect the camel owner pastoral households camel rent decson (whether to rent out ther camels for other salt transporters or to transport salt by themselves)? How much s the proft proporton of the salt transporters from the salt mne among the salt trade value chan partcpants? 1.4 Objectves of the Study General Objectve The overall objectve of the study was to assess the mpact of camel salt transportaton servce on the lvelhoods of pastoralsts n Berahle woreda, Afar regonal state of Ethopa Specfc Objectves Ths study has tred to address the followng specfc objectves. To dentfy the determnant factors those affect households camel ownershp n the study area. To assess the mpact of camel salt transportaton servce on the lvelhood of pastoralsts by comparng the lvelhood of camel owner pastoral households wth nonowner households usng propensty score matchng. To nvestgate the factors behnd camel owner pastoral households camel rent decson (whether to rent out ther camels for other salt transporters or transport salt by themselves). To calculate the proft percentage share of the salt transporters out of the total proft of the salt trade value chan partcpants. 5

18 1.5 Sgnfcance of the Study Most of pastoral households n Afar regon are hghly dependent on lvestock producton, partcularly camel, for ther lvelhoods. For these pastoralsts, camel has multple uses: t serves as a means of ncome generaton, means of transportaton, and source of mlk and meat for household consumpton. In the regon, partcularly n Berahle woreda, camels are almost the only means of transportng salt, food, clothng and shelter. But stll there s a mspercepton and a lack of understandng among people about the economc contrbuton and sgnfcance of the camel transportaton servce. Snce one of the major objectves of the country s to acheve an economc growth to be one of the mddle ncome countres and pastoralsm s part of the country s economy, the government s gvng a heavy emphass to promote a postve vew of pastoral economes and development to counter the very strong under-apprecaton of the economc contrbutons of pastoralsm. Thus, t s beleved that, valuaton of the mpact of camel salt transportaton servce on the lvelhoods of the camel owner pastoralsts s mportant for better understandng of economc contrbuton and mportance of camel transportaton servce. Ths study s expected to be an mportant nput for polcy makers on how to make pastoralsm n general and camel transportaton servce n partcular as a postve contrbutor to economc growth and development. Confdently, t wll also be an mportant ngredent n pavng the way for researchers who have the nterest to conduct a research n ths area. Generally, ths research s sgnfcant for polcy makers, researchers and government organzatons. 1.6 Scope of the Study Out of the dfferent transportaton servces that a camel provdes to ts owners n the study area, only ts role n transportng salt s analyzed n ths study. However, t was also probably better to value ts transportaton servce n transportng other goods such as domestc water and fuel, household tools when the nomadc resdents around the study area move from one area of resdence to other, and so on. 6

19 In addton, even f pastoral households of dfferent parts of the regon and non-pastoral ndvduals who came from dfferent regons of the country are benefcares of ths camel salt transportaton servce, the researcher tred to nvestgate ts mpact on the lvelhoods of pastoralsts of one woreda of the regon, Berahle woreda, only whch s selected as a sample representatve of the regon. However, t s supposed that, t was mportant to value ts transportaton servce n other woredas of the regon and n other regons and even at natonal level. To avod that the study could become too general and so as to make t manageable and feasble, t was decded to conduct an ntensve nvestgaton wthn these delmtatons. 1.7 Lmtatons of the Study Bascally, the survey was conducted at one sngle pont n tme provdng a cross sectonal data; hence, ndvdual changes through tme are not avalable. Moreover, the survey was undertaken takng one woreda as a representatve n whch t has to be recognzed that the fndngs of the study regardng the economc contrbuton of camel transportaton servce may not be generalzed beyond the boundares of the study area even for the other woredas of the regon and other pastoral regons of the country wth smlar camel owner pastoral households. The response of the sample households and FGDs may not also be free from personal bas. However, gven all these lmtatons, consderable care was taken n makng the study as objectve and systematc as possble. 1.8 Organzaton of the Study The paper s framed n fve chapters. The frst chapter ncorporates background of the study, statement of the problem, objectves of the study, scope of the study, sgnfcance of the study and lmtatons of the study. The second chapter presents the revew of related lteratures that ncludes both theoretcal and emprcal lteratures. The thrd and the fourth chapters also present methodology of the study and data analyss and presentaton of results respectvely. In the fourth chapter, the raw data s analyzed va both descrptve and econometrc methods of data analyss. The ffth chapter presents the conclusons and polcy recommendatons. 7

20 CHAPTER TWO 2. LITERATURE REVIEW Ths chapter has two parts: the theoretcal revew and the revew of emprcal lteratures. The frst part brefly dscusses about the theoretcal perspectves wth a subject matter related to the benefts of anmal tracton partcularly camel transportaton servce, challenges to camel producton and causes of camel loss. The second part emphaszes on gvng general nformaton on how to use PSM method on lvelhood mpact assessment of lvestock producton and what factors determne lvestock ownershp by revewng related lteratures undertaken on lvelhood contrbuton of lvestock. 2.1 Theoretcal Revew In ths part, revew of lteratures carred out n areas related to the general benefts of anmal tracton, partcularly the benefts of camel and ts transportaton servces are presented Termnology and Defntons Lvelhoods: - Lvelhoods nvolve the use of assets n actvtes to produce outputs to enable them meet consumpton requrements and aspratons and also to nvest assets and actvtes for the future. Ths commonly takes place n the context of an uncertan envronment. For nstance, for the pastoral households, keepng lvestock may help them to meet ther consumpton requrements not only by drectly provdng them wth food, fuel, transport, or wth har or wool for clothng, but also by generatng sales ncome that helps them to purchase these and other consumpton goods and servces (LPPS, 2005). Pastoralsm: - Is a lvelhood and producton system practced by peoples who lve n ard and sem-ard envronments. The man actvty of pastoral peoples s rasng and herdng of lvestock such as camels, sheep, goats, cows, etc. Most of the land n whch pastoralsts resde cannot support sustaned and relable agrculture because of the extreme varablty of weather patterns that t faces. Due to ths factor, pastoralsm has a moble aspect, movng the herds n search of fresh pasture and water. Lvestock: - Is an umbrella term used for domestcated anmals rased usually n an agrcultural envronment wth the ntent of provdng food, textles, labor, or fertlzer to ther 8

21 owners. Anmals lke horses, pgs, goats, cows, sheep, camels and poultry are consdered lvestock (McMahon, 2013). Tropcal Lvestock Unts (TLU): - The equvalence between dfferent speces s often expressed n terms of TLU. A TLU s 250 klograms lve weght of any domestc herbvore (Sandford, 2006). Anmal tracton: - s the human use of non-human workng anmals [cattle (bulls, oxen and cows), donkeys, mules, horses, goats, camels, etc.], to assst humans n carryng out dfferent tasks n agrculture, transport, rrgaton, buldng ndustry, etc. In transportaton, they are used for pullng carts and loads over a surface, loggng and carryng loads as pack anmals (Smalenga and Joubert, 1997). Pack transportaton: - s transportng loads on the backs of anmals whch provde a reasonably rapd, quet, and relable moblty even n mountans, jungles, and other terran whch are unsutable for vehcular transportaton. Camel: - Camels are ungulates and are herbvores. There are two speces of camels: Dromedary camels and Bactran camels. A Bactran camel has two humps and a Dromedary Camel has one vsble hump whch s a dstnctve fatty depost on ther back. The dromedary camel s natve to the Mddle East and the Horn of Afrca and the Bactran camel s nhabt n Central Asa. Camel s a large and strong hard workng anmal of the desert. It can travel a great dstance wth lttle food and water across hot and dry desert where transportaton vehcle can t go and carry people and heavy loads to the places where there are no roads. The average lfe expectancy of a camel s 40 to 50 years. A full-grown adult camel stands 1.85 m at the shoulder and 2.15 m at the hump. Camels can run at speeds of up to 65 km/h (40 mph) n short bursts and sustan speeds of up to 40 km/h (25 mph) n long dstances. Camels are able to wthstand changes n body temperature and water consumpton that would kll most other anmals. Camel tran: - Is a seres of camels carryng goods or passengers or both n a group as part of a regular or sem-regular servce between two ponts. 9

22 2.1.2 Anmal Tracton and Its Benefts Dfferent authors have wrtten about the general economc and socal benefts of anmal tracton to the natonal economy n general and to pastoral households n partcular. Some of the related lteratures revewed n ths study are presented below. Accordng to Heffernan et al. (2001), for many poor households, lvestock s the prmary form of savngs. As an nvestment, few other resources can match lvestock as a means of captal growth. Anmal sales may allow poor households to generate cash quckly durng tmes of need. Moreover, lvestock, ncludng manure, s often a key source of ncome. Lvestock s also socal captal. It s mportant n supportng socal relatonshps. Loans and gfts of lvestock contrbute to bondng, brdgng and lnkng n socal captal relatonshps, and lvestock s one means by whch famly and household socal captal may be measured. Smlar study conducted by Nkala (2012) n central Mozambque has also stated the wde range contrbuton of anmals to socal and economc wellbeng of peoples. Nkala reported that, lvestock together wth crop producton, comprses the man source of ncome for the agropastoralsts n the study area. The agro-pastoral households n the area dd rase dfferent types of lvestock n dfferent combnatons - cattle, sheep and goats or camels, goats and sheep or all. Cattle and sheep are prmarly grazers, whle camel and goats are normally browsers. Cattle producton domnated n agro-pastoral lvestock producton. As pasture condton deterorated over the years, agro-pastoral communtes shfted from fewer cattle to more camel producton wth shoats to sustan subsstent households ncome. Anmals assst n elmnatng poverty, reducng drudgery and creaton of wealth. Anmal tracton s partcularly mportant for food securty n smallholder farmng systems. Anmals save household (especally women and chldren) tme and effort by carryng water and fuel wood (FAO, n.d). Anmal power s a renewable energy source that s partcularly suted to famly-level farmng and to local transport. Anmal power s generally affordable and accessble to the smallholder farmers, who are responsble for much of the world's food producton. The avalablty of anmal power allows women and men to ncrease ther effcency and reduce ther drudgery, compared wth manual alternatves. The transport role of anmals s mportant for carryng farm nputs and outputs. Pack anmals and carts facltate the marketng of produce, stmulatng local trade and producton. Furthermore, anmals can be very mportant for carryng domestc 10

23 water and fuel, releasng tme that can be used n other productve or socally mportant tasks. Even f motorzed power also brngs many benefts, anmal power s normally more avalable and affordable to people n rural areas and fragle envronments (Ibd). The dea of the study conducted by Starkey and Fernando (1998) s also n lne wth ths. Transport systems based on anmal energy can have several socal and economc benefts for women and communtes. The trng and tme consumng women s tasks of transportng water, fuel wood and grans for mllng can be greatly releved usng anmal energy. If anmals are avalable, women may be able to delegate water collecton to chldren. The woman gans mportant tme, whle chldren often fnd drvng an anmal-drawn cart as much a recreaton as a household duty. Women's trade and marketng, whch s a very mportant for the rural economy as a whole and ther domestc economy n partcular, has been restrcted by ther capacty to head-load goods n many crcumstances. Moreover, Starkey and Fernando (1998), has also reported that, farmers wth anmal transport (carts or pack anmals) have larger crcles of contacts and trade whch allows them to ncrease ther producton and ther proft. Usng anmal energy for transport, greater use s made of manure and crop resdues, whch also ncreases overall farm producton. As women farmers and traders are freed from the lmtatons of head-loadng, more s produced and traded, ncreasng profts and overall economc actvty. Ths ndcates that, pack anmals offer a sgnfcant payload advantage over human carrage, especally f one person can command the use of several anmals. Even wth a sngle anmal, the potental cost reducton from substtuton of pack for human carrage s of the order of 50 percent, whch would sgnfcantly mprove the effcency of transport work by farmers (Tesfahunegn, 1986). Importance of anmals as a means of transportaton has to be consdered startng from the tme n whch there was no any developed modern means of transportaton. Because transportaton s mportant for human exstence and t s obvous that, n those tmes, peoples do use ther anmals where no other alternatve modern means of transportaton was avalable. Regardng ths pre modernzaton contrbuton of anmals, McMahon (2013) has wrtten that: Before the tme when we could hop on a plane for a few hours to get to the other sde of the world or take a quck drve to the store, humans reled on anmal-based transportaton. For all of human exstence, 11

24 transportaton has been of the utmost mportance. From the transport of a nomadc communty across barren lands to the current mport of ol to Amerca, human socetes have always been forced to devse ways to move tems from one place to another. It seems only natural that n a landscape so full of bodversty, early humans would have turned to anmals as a form of transportaton. However, even f anmal transportaton servce has all the sad economc and socal mportance, relyng on such knd of forms of transportaton has also ts own negatve consequences. Relance on tradtonal forms of transportaton poses a consderable barrer to the development of an exchange economy and locks the peasant farmer nto a subsstence mode of exstence and low qualty of lfe from whch t s dffcult to escape (Howe and Garba, n.d). It s obvous that, anmal power s an old technology. However, snce anmals are convenent to be used complementarly wth other power sources, the ncreased use of tractors and motorzed vehcles for transport can even be assocated wth an ncreased use of anmals, as the overall rural economy grows and dversfes (FAO, n.d). Regardng the contnuty of anmal tracton, Smalenga and Joubert (1997) has predcted that, anmal tracton contnues to ncrease n many parts of the world, partcularly those where there are sgnfcant numbers of smallholder farmers. It wll contnue to be mportant for food securty, self-relance and poverty allevaton. All countres, whatever ther degree of ndustralzaton and urbanzaton, can beneft from ecologcally sustanable power sources. Domestc anmals can play a valuable role n assstng human endeavors and mprovng the qualty of lfe of women, men and chldren Camels and Ther Transportaton Servce Camels are known as "shps of the desert" and have been used for transportng goods across deserts for thousands of years. In fact, camels are the only desert anmals those can carry heavy loads of goods and travel for a long perod of tme wthout food or water. Transportaton, however, s not the only beneft that camels can offer us. Desert peoples also rely on camels for ther mlk, meat, and fur. Even camels' droppngs are useful; desert peoples use camels' manure as fuel. 12

25 In pastoral producton systems of east Afrca, camels are used as multfunctonal anmals and are kept for the am of producng: mlk, meat, blood, hdes and skns, provson of transport, barter trade of sale and exchange, and socal and cultural functons (Kaufman and Bnder, 2002). Camels provde a cheap and relable alternatve for short dstance transport. Many knds of goods are regularly transported over short and medum dstances whatever the volume or value of trade s too low to make motorzed transport feasble, and where roads are bad or nonexstent. Durng the recent famnes n the vast dry areas n Afrca, camels were frequently used to carry relef food to remote, otherwse naccessble locatons (Schwartz, 1988). Heanvng and Zhongmn (n.d) n ther study about Camel trans n the desert have wrtten that, n the Thar Desert, camel carts are stll popular and reman a frequent means of transportaton. They fulfll ths functon not only n remote rural areas, but also n the major ctes. They move goods of all knds, especally wood, fodder, gas cylnders, fabrcs, brcks, etc. Ownershp of a camel and a cart s a sold source of ncome, suffcent to support a famly. Camel can carry loads weghng kg over long dstances and 450 kg over short dstances. Other chores performed by camel nclude threshng, lftng of water and powerng of ol mlls. In the past two thousand years, caravans on the slk Road transported slk, tea, pottery and lacquer ware from Chna to the western regons; and pearls, jade, herbal medcnes and perfume from Central and West Asa and Europe to Chna usng camels. Peasants n Mnqn County rase camels n ther spare tme. In summer, they mlk camels and collect camel har; and n autumn, they earn money by usng camels to transport goods. Camels are loyal companons and gudes to desert travellers. Therefore, those who rase camels look after them well as they depend on them for survval at tmes (Ibd). Study by Srvastava (1991) n Rajasthan shows the value that s gven to camel by the socety. He states that, camel s often regarded as emblematc of Rajasthan. In Rajasthan folklore t symbolzes love, and ownershp of a camel once sgnaled status and wealth. It was used for warfare by the Maharajahs and played an mportant role n desert communcaton, transportaton and trade. Hstorcally, camels were thus a valuable commodty used by the rulng classes and by the busness communty. 13

26 Comng to Ethopan case, Eyassu (2009) n hs Analyss of the Contrbutons of the Dromedary Camel and Constrants to Camel Producton n Jjga and Shnle Zones of Eastern Ethopa reported that, dromedary camels play an mportant role to the lvelhood and survval of nomadc pastoralsts n the study areas. He found the man contrbutons rendered by dromedary camels n the study areas to be mlk producton and transportaton. The major transportaton servces, that a camel render to the pastoralsts, s transportng of people, goods and moble houses durng ther seasonal mgraton. In the study areas, camels are often hred as a cargo to transport goods; partcularly male camels are used as a draft anmal for transportaton of goods and people. But camels n these areas feed manly on poor-qualty natural vegetaton. Other related studes such as The camel n Ertrea: an all-purpose anmal by Gebrehwet (1998) revew the overall role of camel n general and camel as a pack anmal n partcular. He reported that, n vew of lack of roads and transportaton facltes and the naccessble terran n most pastoral areas, the role of camel as a pack anmal s crucal and determnes the survval of the nomads n the hostle envronments of the desert. Then, due to all these valuable contrbutons that camels render, promotng camel husbandry s expected to mprove lvelhood of camel producng pastoral households. For nstance, accordng to Houten (2002), promotng camel husbandry has mproved food securty for many pastoralsts. One of the most sgnfcant aspects s how camel husbandry and producton have now been taken up by many hundreds of pastoralsts n hs study areas, Samburu and Maasa. Although they herd ther camels n a dfferent manner, ther herd productvty s often much hgher than that of tradtonal camel owners. Nowadays, modern transportaton s expandng whch seems to replace the tradtonal means of transportaton. However, regardless of the ntroducton of motorzed transportaton, camels reman an mportant means of transportaton n the ard and sem-ard regons. Apart from use by nomads who are contnuously seekng grazng and water, these anmals are employed n transportng farm produce to local markets (Katsna, 1990). Hassan and Ibtoye (n.d) have also made smlar conclusons about contnuty of camels as a means of transportaton. They predcted that, camels wll contnue to play actve roles n the transportaton of farm nputs and produce for short dstance travels due to the rsng costs of motor vehcles and spare parts. 14

27 2.1.4 Challenges to Camel Producton and Causes of Camel Loss Despte the benefts that Afrcan pastoralsts get by producng camels, ther camels are vulnerable to dfferent challenges such as camel dseases, drought, rustlng and predaton whch expose the pastoralsts to rsks of losng ther camels whch are source of ther lvelhoods. For nstance, Njuk et al., (2011) has stated the major challenges assocated wth camel producton n the pastoral areas of east Afrca to be predaton, drought and camel dseases. Drought, as a major challenge to camel producton, can be attrbuted to loss of pasture and dryng up of water sources durng dry perods whch results n the camels sufferng from dehydraton and starvaton and eventually they would de. It may also lead to encroachment of wldlfe protected areas leadng to ncreased camel predaton especally at the grazng felds and waterng ponts (Onono et al., 2010). Smlarly, Eyassu (2009) has also found the major constrants assocated wth camel producton n hs study areas to be feed shortage and prevalence of dsease. Furthermore, he stated that, as a consequence of these problems, camels n the study areas feed exclusvely on unmproved perennal natural vegetaton of low nutrtve value and they are not gven supplementary feed. 15

28 2.2 Emprcal Revew Propensty score matchng (PSM) method s used to assess the mpact of a partcular treatment or nterventon on outcome varables of nterest. Partcularly, whle analyzng the lvelhood mpact of lvestock producton usng cross sectonal data, PSM s the approprate method that has to be employed. Ths part was ntended to present revew of an mpact assessments undertaken by dfferent authors on the lvelhood contrbuton of lvestock producton n general and anmal tracton n partcular usng PSM method of data analyss. It would be better, f lteratures whch explore the lvelhood contrbuton of lvestock, analyzed usng PSM method of analyss were revewed. However, for a varety of reasons, there are no as such documented lteratures n ths area. Therefore, due to the absence of such emprcal lteratures, the only lvestock related lterature revewed n ths study s presented below. Brol et al. (2010) has tred to nvestgate the role of poultry on the lvelhoods portfolos of households and the mpact of supply and demand shocks that may be caused by Hghly Pathogenc Avan Influenza (HPAI) on varous lvelhoods outcomes of households n four Sub-Saharan Afrcan (SSA) countres; Ethopa, Kenya, Ghana and Ngera. They use probt model and zero nflated negatve bnomal model to profle the household, farm and regonal characterstcs of those households who are most lkely to keep poultry, and those who are most lkely to be engaged n ntensve poultry producton,.e., keep larger household flocks and estmate the mpact of the dsease outbreaks and threats on lvelhood outcomes by usng propensty score matchng. From ther estmaton results they revealed that, across the four SSA countres, the profles of households who are predcted to be poultry keepers and those who are predcted to keep larger small-scale flocks have older and less educated household heads, and are larger wth more chldren and wth more adult women. In terms of asset ownershp, households who are predcted to be poultry keepers and those who are predcted to keep larger flocks have hgher average values of lvestock wealth and other assets (e.g., land). Moreover, the result of the mpact assessment reveals that, across the all four study countres, households whch are larger small-scale producers seem to be most vulnerable to HPAI related shocks n terms of lvestock ncome and/or wealth (asset value) loss. 16

29 CHAPTER THREE 3. METHODOLOGY OF THE STUDY 3.1 Descrpton of the Study Area Afar regon s one of the regonal states of Ethopa located n the north eastern part of the country. The regon s subdvded nto fve admnstratve zones and one specal woreda and t has a total of 5 zones and 29 woredas. The Ethopan Central Statstcal Authorty (CSA) has estmated the 2008 populaton of the Afar Admnstratve Regon at 1,449,000, of whch 137,000 are urban resdents. The breakdown by gender (803,000 males and 646,000 females) ether stands out as a glarng example of data errors that produced hghly lopsded numbers by sex, or suggests a troublng scenaro of a harsh survval envronment for female members of the populaton. The numbers suggest a sex rato of 124 males per 100 females, often found among populatons who have suffered a level of socetal dsrupton such as excessve gender-specfc mgraton, or excessve gender-specfc mortalty. Out of the total populaton of the regon, 90 percent are pastoralsts and the remanng 10 percent are agro-pastoralsts. Those peoples n the regon therefore depend manly on lvestock producton for ther lvelhoods. The total fertlty rate of the regon s 4.9 whch s below the natonal average. The natve language n Afar regon s Afargna, whch s of Cushtc orgn. Much of the land of the regon s dry and rocky, whch s unsutable for crop cultvaton. Out of the total area of the regon (estmated at 97,250 km 2 ) cultvated and arable land consttutes 5.24 percent, forest 1.54 percent, bush and shrub percent, grassland 1.56 percent, marshy land 2.74 percent, water bodes 0.63 percent, and degraded and rocky land 63.7 percent. The regon s alttude ranges from a maxmum of 1500m above sea level to a mnmum of 166m below sea level. Temperature vares from 25ºC durng the wet season to 48ºC durng the dry season. Ranfall s erratc and scarce, and annual precptaton ranges from 200mm to 600mm. The regon s frequently exposed to persstent droughts and s classfed as one of the droughtprone regons n Ethopa. 17

30 Fgure 3.1: Locaton map of the study area (Berahle woreda) Berahle Source: Aynalem (n.d) The ste n whch the research has been conducted s Berahle woreda, one of the woredas n the regon, located n the north eastern part of the regon. It s part of Admnstratve Zone 2, and ts terrtory ncludes part of the Afar Depresson. Ths woreda s bordered on the south by Afdera and Abala, on the southwest by Tgray Regon, on the west by Koneba, on the north by Dallol, and on the northeast by Ertrea. Towns n Berahle nclude Berhale and Tyarabora. The average elevaton n ths woreda s 233 meters above sea level. The major water body n ths woreda s the salne Lake Karum (also known as Lake Assela). As of 2008, Berahle has 236 klometers of all-weather gravel road and about 13 percent of the total populaton has access to drnkng water. 18

31 Based on the 2007 Census conducted by the Central Statstcal Agency of Ethopa (CSA), ths woreda has a total populaton of 78,881, of whom 45,501 are men and 33,380 are women. Wth an area of 2, square klometers, Berahle has a populaton densty of A total of 11,402 households were counted n ths woreda, whch results n an average of 6.9 persons to a household, and 11,653 housng unts percent of the populaton sad they were Muslm, and 1.03 percent were Orthodox Chrstans. The woreda s found at a dstance of 120 klometers far from Mekelle (captal cty of Tgray regonal state) and around 993 klometers from Adds Ababa (captal cty of Ethopa). Mnng s the prncpal ndustry n ths woreda and the best known resource extracted s salt. 3.2 Data Source and Samplng Procedure Data Source and Data Type In order to successfully acheve the stated objectves, the study employed both prmary data (cross-sectonal data) and secondary data. The prmary data was collected through dspersng of structured questonnare wth face to face ntervew from selected sample households of the target area, drect observatons and focus group dscussons. Enumerators, those who collect data from the sample households, were selected accordng to ther level of educaton, data collecton experence and ther profcency on the local language Afargna and then were traned on the basc contents of the questonnare and data collecton technques. Before the man data collecton, a plot survey was conducted usng some randomly selected households from the two groups of households (camel owners and camel non-owners) n order to check the effcency of the questonnare and to make some correctons f necessary. The man survey was conducted on November Samplng Procedure and Sample Sze A multstage samplng technque (ncludng both probablty and purposve) was employed n order to reach at the selecton of sample households used for the analyss. At the frst stage, Berahle woreda was purposely selected, out of the total woredas of the zone due to the fact that the salt mnng area s found wthn the boundary of ths woreda and s under admnstraton of the woreda. In the second stage, 1 kebelle (Berhale kebelle) out of the total 9 kebelles of the 19

32 woreda was selected; because t consttutes the hghest number of salt transporters and camel renters even f there are few partcpants n the other 8 kebelles too. Then after, a total of 250 sample households, ncludng 115 treated (camel owners) and 135 controls (camel non-owners) were randomly selected and ntervewed from all kushets of the kebelle. The kebelle has a total of 5 kushets. Takng camel ownng as a treatment, those households who own camels and use ther camels for salt transportaton ether by rentng out them for other salt transporters or transportng salt by themselves are referred to as treated and those households who do not have ther own camels are referred to untreated or control group. Households those who own camels but do not ether transport salt by themselves or rent out them for other salt transporters and households those who do not own camels but transport salt by rentng n camels from camel owners are excluded from ths study. And the sample sze does not consttute these groups. Ths s because: The formers, even f they are camel owners, they are gettng nothng from salt trade whle ther camels are not nvolved n salt transportaton. Hence, they cannot be consdered as treated groups. The latters on the other hand, even f they don t own camels, whle they are benefcares of camel salt transportaton servce by rentng n camels from other camel owners, they cannot be consdered as control groups too Questonnare Desgn In order to attan the stated objectves of the study, a structured questonnare consstng of dfferent sectons, each secton havng dfferent types of questons was developed. The developed questonnare used for survey s presented n Annex XII. The frst secton of the questonnare asks respondents about the demographc characterstcs of household members. The second secton consttutes questons about the total asset ownershp and value of the households. Basc assets ncluded were land, lvestock and fxed assets. But no one of the sample households was found wth agrcultural land ownershp. 20

33 The questons n the thrd secton of the questonnare are related to ncomes of the households n 2012/13 (2005 E.C.). The subsectons under ths ncome secton are: the ncome that the households ganed from employment for wage, own busness actvtes, mgratng to other places, sale of lvestock and lvestock by-products. The fourth secton consttutes questons related to camel rent and salt transportaton experences of the households. The ffth part s about the fodder and water that the households supply to ther camels. And n the last secton, the respondents were asked general questons about ther camels and questons related to ther lfe condton. 3.3 Methods of Data Analyss For analyss purpose, ths paper has utlzed both descrptve and econometrc data analyss methods Descrptve Method of Data Analyss In the descrptve method of data analyss, the data collected from the sample households and FGDs was analyzed usng statstcal technques lke, tables, means, percentages, ratos, frequences and charts. In order to acheve the fourth objectve of the study,.e. n order to dentfy the man benefcares of the salt mne among the salt trade value chan partcpants; and calculate the proft percentage share of the salt transporters out of the total proft of the salt trade value chan partcpants, descrptve method of data analyss was employed. In addton, the same method of data analyss was used to calculate returns from camel salt transportaton servce, the contrbuton of camel salt transportaton servce for employment and for the natonal economy Econometrc Method of Data Analyss For the proper estmaton of the mpact of camel transportaton servce on the lvelhoods of the camel owner pastoral households, the paper has also employed econometrc estmaton models. 21

34 Econometrc Framework and Estmaton Strategy When nvestgatng the mpact of camel transportaton servce on the lvelhood of camel owner pastoral households (treated group) by comparng ther lvelhood wth those who are non-owners (control or untreated group), the researcher faces a problem of estmaton selecton bas. Ths s due to the fact that, to evaluate the mpact of a treatment on performance ndcators, t s necessary to draw a counterfactual scenaro about the performance ndcators of the treated group. The counterfactual ndcators would then be compared wth the performance level of the treated when they become treated n order to evaluate the mpact of the treatment on the performance ndcators (Heckman et al., 1997). For the camel owners (treated group) ther counterfactual would be ther lvelhood condton n the absence of the gan from ther camel ownershp. Whle for the non-owners (control or untreated group), ther counterfactual would be the level of ther lvelhoods f they have moved to camel owner status. However, n realty a household cannot hold both actually treated and control group status and therefore t s only by constructng a counter factual group that the proper comparson can be made. Thus, to elmnate selecton bas, there s the need to compare the performance levels of treated and control groups whch are statstcally comparable (Rosenbaum and Rubn, 1983; Khandker et al., 2010). Rosenbaum and Rubn (1983) suggest the use of Propensty Score Matchng (PSM) approach to deal wth selecton bas Propensty Score Matchng (PSM) Matchng subjects on an n-dmensonal vector of characterstcs s typcally unfeasble for large n; hence, ths method proposes to summarze pre-treatment characterstcs of each subject nto a sngle-ndex varable (the propensty score) that makes the matchng feasble. 22

35 In ths study, t s possble to estmate the lvelhood effect of camel salt transportaton servce by matchng how lvelhood ndcators dffer for households who own camel and use them for salt transportaton relatvely to observatonally smlar camel non-owners. In analyzng usng PSM approach, the followng assumptons have to be held. (A) Condtonal Independence Assumpton (CIA): states that gven a set of observable covarates X whch are not affected by treatment T (camel ownershp n ths case), potental outcomes Y (where Y s the set of lvelhood ndcators lke ncome, fxed asset formaton and lvestock ownershp) are ndependent of the treatment assgnment T. If Y T represents outcome ndcators of the treated group (camel owners) and Y C represents outcome ndcators of the control group (camel nonowners), then, the condtonal ndependence assumpton mples that: (Y T, Y C ) T X for all X Where represents ndependence. Ths assumpton s also called unconfoundedness (Rosenbaum and Rubn 1983), and t mples that uptake of the treatment s based entrely on observed characterstcs. Selecton s solely based on observable characterstcs and that all varables that nfluence treatment assgnment and potental outcomes are smultaneously observed. Condtonng on all relevant covarates s lmted n case of a hgh dmensonal vector X. Accordng to Rosenbaum and Rubn (1983), to overcome the problem of dmensonalty for non-randomzed observatons, treatment and outcome are ndependent of the propensty to score P(X), whch s a condtonal probablty of an ndvdual to partcpate n a treatment gven hs observed covarates X whch s so called balancng score. It shows that, f potental outcomes are ndependent of treatment condtonal on covarates X, they are also ndependent of treatment condtonal on a balancng score. The condtonal ndependence assumpton (CIA) based on the propensty score (PS) can be wrtten as: (Y T, Y C ) T P(X) for all X (B) Common Support or overlap condton: states the phenomenon of perfect predctablty of T gven X:.e., 0 < P (T= 1 X) < 1 where T s camel ownershp 23

36 It ensures that persons wth the same X values have a postve probablty of beng both camel owners and non-owners (Heckman et al., 1999). There are two steps that have to be followed n order to conduct an assessment on the mpact of camel salt transportaton servce on the lvelhood of camel owner pastoral households n the study area usng PSM approach. Estmatng the probablty of households camel ownershp and Estmatng the lvelhood contrbuton of camel salt transportaton servce for the camel owner pastoral households I. Model for the analyss of the determnant factors those affect households camel ownershp Bnary model s used to estmate households probablty of beng treated (beng camel owner) on observable characterstcs. Snce propensty score s a condtonal probablty estmator, any dscrete choce model such as logt or probt can be used as they yeld smlar results (Calendo and Kopeng, 2008). The two models are bascally the same except the dfference they have n ther dstrbuton: Logt Cumulatve standard logstc dstrbuton (F) Probt Cumulatve standard normal dstrbuton (Φ) In ths study, logt model s used as a bnary model n order to estmate households probablty of camel ownershp. Logt model s a non-lnear regresson model and s approprate when the dependent varable s bnary (dummy) whch takes values of ether 0 or 1. It estmates the probablty of the dependent varable to be 1.e., P(X ) = 1 (Gujarat, 2004). The form of the logt model s: P( X ) P( camel_ ownp = 1/ X ) = F( β 0 + β X + ε) = k Where: P(X ) = s the probablty of households camel ownershp X = s a set of all observable characterstcs those affect the probablty of households camel ownershp β 0 = s the constant term 24

37 25 β k = s a set of parameters of nterest to be estmated ԑ = s the dsturbance term The logt model s then specfed as, ) ( ) 1 _ ( ) ( β β β β X X X e e e X F X ownp camel P X P + = + = = = = Where, k k X X X X X β β β β β β = P(X ) s nonlnear not only n X but also n the β s. The probablty of households camel non-ownershp s gven by: ) 0 _ ( ) ( 1 β β β X X X e e e X ownp camel P X P + = + = = = Therefore, we can smply derve the odds rato (the rato of the households probablty of camel ownershp to the probablty of camel non-ownershp). It can be wrtten as: β β β X X X e e e X P X P = + + = ) ( 1 ) ( X β e = s the odds rato If we take the natural logarthm of the above equaton, we obtan: k k X X X X X X e X P X P L β β β β β β β... ln ) ) ( 1 ) ( ln( = = = = L, the log of the odds rato, s not only lnear n X, but also lnear n the parameters (Rosenbaum and Robn (1983), Bryson et al. (2002)). L s called the logt model. Then, we have the logstc estmaton as: β X X P X P L ownp camel = = = = ) ) ( 1 ) ( ln( 1) _ ( Prob

38 Where X 6 β = β 0 + β 1 sexhh + β 2 agehh + β 3 educhh + β 4 acc _ credt + β 5 fam exp_ co + β adulthh + β adul _ rato + ε 7 Maxmum Lkelhood Method (MLM) of estmaton s used to estmate these parameters of nterest. Fnally, propensty scores of each ndvdual household are estmated from the logstc regresson. Propensty scores close to 1 ndcate treatment characterstcs assocated wth a hgh probablty of camel ownershp, whch are calculated for each treated by applyng ther background values to the logstc model. II. Model to analyze lvelhood contrbuton of camel salt transportaton servce for the camel owner pastoral households The propensty score matchng produce vald matches whch can be used to estmate mpact of camel salt transportaton servce on lvelhood at the second stage of the analyss after the propensty score P X ) s known n the frst stage. Ths s done by matchng the two groups of ( respondents on the bass of the predcted propensty score (Backer and Ichno, 2002). Accordng to Roy-Rubn model (Roy (1951) and Rubn (1974)), the mpact of a treatment on the outcome of a household nvolves speculaton about how ths household would have performed had t not receved the treatment. In the case of a bnary treatment, the treatment ndcator T equals 1 f household s treated (camel owner n ths case) and zero otherwse. The potental outcomes (ncome, lvestock ownershp and fxed asset formaton) are then defned as Y (T ) for each household, where = 1,,N and N denotes the total populaton. The treatment effect.e. the effect of the gan from camel salt transportaton servce for a household can be wrtten as: TE = Y (1) Y (0) Where, Y (1) s the outcome of household after beng camel owner and Y (0) s the outcome of the household before havng camels But a problem arses because only one of the potental outcomes s observed for each household. The unobserved outcome s called counterfactual outcome. Hence, estmatng the ndvdual household treatment effect TE s not possble; we have to shft to the populaton average treatment effect. 26

39 `Average treatment effect on the treated' (ATT) s a parameter of nterest wth a great attenton n most lteratures and s estmated as follows: ATT = E( ATT T = 1) = E[ Y (1) T = 1] E[ Y (0) T = 1] However, E [ Y (0) T = 1], whch s a mean counterfactual for those beng treated s not observable. Hence, n order to estmate the ATT, t has to be replaced by proper substtute. Usng the mean outcome of untreated ndvduals E [ Y (0) T = 0] as a counterfactual for the treated may result n based estmates. Ths s because the outcomes of households from treatment and comparson groups would dffer even n the absence of treatment leadng to a `self-selecton bas' (Calendo and Kopeng, 2005). However, n socal experments where assgnment to treatment s random, E[ Y (0) T = 1] - E [ Y (0) T = 0] =0 and the treatment effect s dentfed. Therefore, ATT s the mean dfference n outcomes of the two groups of households over the common support approprately weghted by the propensty score dstrbuton of partcpants. Four dfferent matchng algorthms whch nvolve trade-offs n terms of bas and effcency are used n the study to match treated (camel owner) and control (camel non-owner) households. These are: 1. Nearest Neghbour Matchng (NNM) - Under ths method, a control unt can be a best match for more than one treated unts (usng replacement). It conssts of takng each treated unt and searchng for the control unt wth the closest propensty score. In NNM, all treated unts fnd a match (Backer and Ichno, 2002). - Once each treated unt s matched wth a control unt, the dfference between the outcome of the treated unts and the outcome of the matched control unts s computed. The ATT s then obtaned by averagng these dfferences. 2. Radus Matchng (RM) - In ths algorthm, a household from the control group s chosen as a matchng partner for a treated household that les wthn the specfed radus n terms of 27

40 propensty score. I.e. each treated unt s matched only wth the control unts whose propensty score falls nto a predefned neghborhood of the propensty score of the treated unt. 3. Kernel Matchng (KM) - In KM, each treated s matched wth a weghted average of all controls wth weghts that are nversely proportonal to the dstance between the propensty scores of treated and controls. - All respondents of the underlyng sample of control group are ncluded; and weght more dstant observed characterstcs among both groups of households down (Heckman et al., 1997). The Kernel based estmator of the ATT descrbes the mean dfference n outcome whle the matched outcome s gven by Kernel-weghted average of the outcome of control group of respondents. 4. Stratfcaton Matchng (SM) - Conssts of dvdng the range of varaton of the propensty score n ntervals such that wthn each nterval, treated and control unts have on average the same propensty score. Then, wthn each nterval n whch both treated and control unts are present, the dfference between the average outcomes of the treated and the controls s computed. - The ATT of nterest s fnally obtaned as an average of the ATT of each block wth weghts gven by the dstrbuton of treated unts across blocks. The above four matchng algorthms reach dfferent ponts on the fronter of the trade-off between qualty and quantty of the matches, and none of them s a pror superor to the others. Ther jont consderaton, however, offers a way to assess the robustness of the estmates. It should also be noted that wth all these methods, the qualty of the matches may be mproved by mposng the common support restrcton (Lechner, 2001). For further detals of these matchng algorthms and ther STATA software commands, we refer to Becker and Ichno (2002). In general, whle estmatng the average treatment effect of camel salt transportaton servce on the lvelhood of camel owner pastoral households usng the stated matchng algorthms, the common support restrcton s mposed and ther jont estmaton s consdered. 28

41 III. Senstvty Analyss PSM cannot control for unobservable characterstcs. It can only control the observed varables whch are ncluded n the propensty score used to match the two groups of households (Calendo and Kopeng, 2005 and 2008; Rosenbaum and Slber, 2001). Thus, a queston arses whether the obtaned superor lvelhood level of camel owners s due to ther camel ownershp or ther other unobserved characterstcs. Hence, before nterpretng the baselne estmate as evdence of a true causal effect of the treatment, testng the presence of unobserved varable becomes a greater mportance whch can be done by usng senstvty analyss (Rosenbaum, 2002). Senstvty analyss s used to assess whether and to what extent the estmated average treatment effects are robust to possble devatons from the condtonal ndependence assumpton (CIA) (Nanncn, 2007). It s assumed that treatment assgnment does not hold the CIA assumpton gven the set of covarates X but holds gven X and an unobserved bnary varable U. As long as U s not observed, the outcome of the controls cannot be credbly used to estmate the counterfactual outcome of the treated. On the contrary, knowng U (together wth the observable covarates X) would be enough to consstently estmate the ATT. In order to make the smulaton of the potental confounder feasble, U s assumed to be bnary and condtonally ndependent wth respect to X. The dstrbuton of the bnary confoundng factor U s fully characterzed by the choce of four parameters: p j Pr( U = 1T =, Y = j) = Pr( U = 1T =, Y = j, X ) wth, j {0,1 } whch gve the probablty that U = 1 n each of the four groups defned by the treatment status and the outcome value and then a value of U s attrbuted to each unt. In case of contnuous outcomes, the smulaton parameters p j are defned on the bass of T and a bnary transformaton of Y nstead of the outcome tself. Thus, p j Pr( U = 1T =, I( Y > y*) = j) wth, j {0,1 }, where I s the ndcator functon and y* s a chosen typcal value of the dstrbuton of Y. The smulated U s then treated as any other observed covarate and s ncluded n the set of matchng varables used to estmate the propensty score and to compute the ATT. Usng a gven set of values of the senstvty parameters, the matchng estmaton s repeated 50 tmes n ths case and a smulated estmate of the ATT s retreved as an average of the ATTs over the dstrbuton of U, whch s robust to the falure of the CIA mpled by that partcular confguraton. 29

42 Of the three alternatve varances: wthn-mputaton, between-mputaton and total, total varance s used n ths analyss whch leads to conservatve nferental conclusons, snce t s always greater than the other two alternatves. Whle computng a standard error for the smulated ATT, the mputaton of U s consdered as a normal problem of mssng data, whch can be solved by repeatedly mputng the mssng values of U. Then, the total varance assocated to AT ˆ T can be expressed as: se 2 T = m m (1 )( ( ˆ ˆ 2 se k + + ATTk ATT ) m k = 1 m m 1 k = 1 Where m s the number of mputatons of the mssng U, and ATT ˆ k and se 2 k are the pont estmate and the estmated varance of the ATT estmator at the k-th mputed data set (wth k = 1, 2,...,m). The smulated ATT, the ATT ˆ k over the m replcatons. AT ˆ T ), s obtaned by the average of A grd of dfferent p j s bult, n order to capture the characterstcs of the potental confounders that would drve the ATT estmates to zero or far away from the baselne result. Accordng to Ichno, Meall and Nanncn (2007), by smply choosng the parameters pj t s possble to smulate a dangerous confounder (a confounder whose exstence mght gve rse to a postve and sgnfcant ATT estmate even n the absence of a true causal effect) f followng mplcatons hold: p 01 > p00 Pr( Y0 = 1T = 0, U = 1, X ) > Pr( Y0 = 1T = 0, U = 0, X ) p 1. > p0. Pr( T = 1U = 1, X ) > Pr( T = 1, U = 0, X ) Therefore, by smply assumng that p 01 > p00 and p 1. > p0. one can smulate a confoundng factor that has a postve effect on the untreated outcome Y 0 and on the treatment assgnment (condtonng on X) respectvely. Besdes, the dfferences d = p 01 p00 and s = p 1. p0. are a measure of the effect of U on the untreated outcome and on the selecton nto treatment respectvely. However, by settng the senstvty parameters p j, we can control the sgn but not the magntude of the condtonal assocaton of U wth Y 0 and T. To avod ths lmtaton, we can measure how each chosen confguraton of the p j translates n terms of the effect of U on Y 0 and T (condtonng on X). At every teraton, a logt model of Pr( Y = 1T = 0, U, X ) s 30

43 estmated and the average odds rato of U s reported as the outcome effect of the smulated confounder: Γ = Pr( Y Pr( Y = 0T Pr( Y = 1T Pr( Y = 1T = 0, U = 0, U = 0, U = 0T = 0, U = 1, X ) = 1, X ) = 0, X ) = 0, X ) Smlarly, the logt model of Pr( T = 1U, X ) s estmated at every teraton and the average odds rato of U s reported as the selecton effect of the smulated confounder: Λ = Pr( T Pr( T Pr( T Pr( T = 1U = 0U = 1U = 0U = 1, X ) = 1, X ) = 0, X ) = 0, X ) By smulatng U under the assumptons that d > 0 and s > 0, both the outcome and selecton effects must be postve (.e., Γ > 1and Λ > 1). Г and Λ as an addtonal output of the senstvty analyss, provde us the magntude of these two effects, whch end up characterzng the smulated confounder U. In general, n ths study, senstvty analyss s used to dentfy whether the nference taken about the effect of camel ownershp on the lvelhoods of the camel owner pastoral households s relable. The analyss s undertaken based on Nanncn (2007) Model to dentfy the determnant factors affectng camel owner pastoral households camel rent decson (Ordered Probt Model) The decson of all of the camel owner pastoral households regardng way of ther partcpaton n salt trade s not the same. Some of them may decde to always rent out ther camels for other salt transporters and beng benefcary from the salt trade n the form of camel rent; some of them may decde to always transport and sale salt by themselves; and some of them may decde to be nvolved n both of them.e. sometmes to transport salt by oneself and sometmes to rent out camels for other salt transporters. These dfferences n the camel rent decson of the camel owner pastoral households, attracts some nterest to understand the determnants of these alternatve decsons. Hence, to deal wth the major pushng determnant factors of camel rent decson of the camel owner pastoral 31

44 households, ordered multnomal choce model s used. Furthermore, among the alternatve two multnomal choce models.e. ordered logt and ordered probt, the ordered probt model s used to nvestgate these determnant factors. Multnomal model s used for a data n whch the choce varable takes more than or equal to three values. Sometmes the categores of such dscrete varables can be naturally ordered. If values of the choce varable can be ordered, we call t ordered multnomal choce model. Ordnal outcomes represent categorcal outcomes where there s clear natural rankng or order from low to hgh among the outcomes but the dstance between the adjacent categores s unknown. Whle modelng these types of outcomes, numercal values are assgned to the outcomes, but the numercal values are ordnal and reflect only the rankng of the outcomes and are no longer arbtrary. When orderng the responses of the choce varable, larger values are assumed to correspond to hgher outcomes (Schmdheny, 2007). In ths study, the choce varable s camel rent decson of the 115 camel owner pastoral households wth three outcome categores. 1 Never 2 Sometmes 3 Always The never outcome category represents response of those camel owner pastoral households who always transport salt by themselves and have never been rented out ther camels for other salt transporters; the sometmes represents response of households those who have been sometmes rentng out ther camels because they sometmes transport salt by themselves; and the always outcome category s assgned for the response of households those who were always rentng out ther camels for other salt transporters and have never been transported salt by themselves due to dfferent factors to be dentfed and presented n the econometrc analyss part of the paper. The average number of months per year n whch the camel owner pastoral households have been rentng out ther camels for other salt transporters s 0, 3.5 and 8 for the households wth camel rent responses of never, sometmes and always respectvely. An ordered probt model, used n ths analyss, s a latent varable model that offers a data generatng process for categorcal dependent varables (camel rent decson n ths case) (Schmdheny, 2007). 32

45 Let y be an ordered response takng the values (1, 2, 3) for (never, sometmes, always): The ordered probt model for y, condtonal on explanatory varables x, can be derved from a latent varable model y *,.e. y * = β + x β + x β x β + ε k k Or y * = x β + ε Where y * = latent ndex of camel rent decson and s a functon of observed and unobserved varables - Once y * crosses a certan value, frst never s reported, then sometmes and fnally the always category outcome follows. x = vector of varables that explans the varaton n the observed dependent varable β = vector of coeffcents and ε = the error term wth cumulatve dstrbuton functon F (.) n whch the model s determned by ts assumed dstrbuton In ordered probt model, the error term s ndependently and normally dstrbuted wth mean 0 and varance σ 2 If the latent varable.e.. F( ε ) ε = observed dependent varable can form: 1 1 exp ε 2π 2 2 * y denotes a natural orderng among the possble outcomes, then the 1 y = 2 3 = Never = Sometmes = Always f y * µ 1 1 µ < y * µ y * > µ 2 2 Where µ 1 < µ 2 are the two unknown cut ponts or threshold parameters to be estmated along wth β n the model. In theory, the threshold values are dfferent for everyone and ther average across households s estmated. These threshold parameters determne how the values of * y to get translated nto the three possble values of y. Then, we have the ordered probt regresson estmaton as: 33

46 y * = x β + ε Where X β = β + β agehh + β educhh + β adumale + β acc credt + β ncomehh + β camels _ hh + ε _ 5 6 Fnally, gven the standard normal assumpton forε, the condtonal dstrbuton of y gven x can be derved by computng each response probablty. The probablty that a household chooses alternatve 1 s: Pr * ( y = 1) = Pr( y µ ) = Pr( x β + ε µ ) = Pr ( ε µ x β ) = Φ( µ x β ) = 1 Φ[ x β µ ] The probablty that a household chooses alternatve 2 s: Pr * ( = 2) = Pr( µ < µ ) = Pr( µ < x β + ε ) Pr( x β + ε µ ) y 1 y = Pr = Φ ( ε < x β µ 1 ) Pr( ε µ 2 xβ ) ( x β µ ) Φ( x β µ ) The probablty that a household chooses alternatve 3 s: Pr 1 * ( y = 3) = Pr( y > µ ) = Pr( x β + ε > µ ) = Pr 2 ( ε > µ x β ) = Φ[ x β µ ] Maxmum lkelhood method (MLM) s used to estmate the model and to do ths; a loglkelhood functon s needed at the frst stage. Lkelhood functon Addng up the three probablty outcomes lsted above gves us the lkelhood functon for the maxmum lkelhood estmaton..e., = [ y = ] L Pr( ) Where m goes from 1 to 3 m And the log lkelhood functon s ln L = ln[ Pr( y = )] m L = Pr( y = 1) Pr( y = 2) Pr( y = 3) [ Pr( y = 1) ] + ln[ Pr( y = 2) ] + ln[ Pr( y 3) ] ln L ln = = 34

47 Or ln L = ln[ 1 Φ[ x β µ ]] + ln[ Φ( x β µ ) Φ( x β µ )] + ln[ Φ[ x β µ ] y = ] y = 2 y = 3 From ths, only the sgn of the estmated parameters β can be drectly nterpreted. A postve sgn tells whether the choce probabltes shft to hgher categores when the ndependent varable ncreases. Note, however that, the absolute magntude of the parameters s meanngless as t s arbtrarly scaled by the assumpton σ = 1. One can therefore e.g. not drectly compare parameter estmates for the same varable n dfferent subgroups. In order to be able to nterpret the varables n terms of magntude, one can obtan the category specfc margnal effects from these (Schmdheny, 2007). Margnal Effects Margnal effects show changes n the choce probabltes due to change n the ndependent varables assumng µ = 0 and σ 2 = 1 and are a functon of: Pont of expanson (x s) Frame of reference for outcome (y) The margnal effects for each of the three choce probabltes are presented as follows. [ 1 Φ[ xβ µ ]] x / Pr( y = 1) / x = 1 = Φ( x β ) β [ Φ( x β µ 1 ) Φ( xβ µ )] x / Pr( y = 2) / x = 2 ( x β µ 1 ) β Φ( β µ 2 ) = Φ x [ Φ[ xβ µ ]] x / Pr( y = 3) / x = 2 = Φ ( x µ 2 β ) β β 3.4 Descrpton of Varables Used n the Analyss Revew of lteratures, dea of experts and knowledge of the researcher were used to dentfy the potental determnant factors of the assgned dependent varables used n ths study. The assgned dependent varables to be analyzed are: households camel ownershp, lvelhood ndcators of the households and camel rent decson of the camel owner pastoral households. The ndependent varables whch are expected to affect these dependent varables are 35

48 categorzed nto two groups; 1) Households demographc characterstcs and 2) Soconsttutonal and economc factors. The descrpton of the varables s presented below. 1. Treatment varable: - the treatment varable used n ths study s households camel ownershp (camel_ownp); whether a household own camels or not, wth a bnary value of 1 f the household s camel owner and 0 f not. 2. Lvelhood ndcators: - the followng three varables are lvelhood ndcators of the households used for the mpact assessment. ) Income (ncomehh):- s a contnuous varable whch measures the total annual ncome of a household. It s the summaton of the ncomes that the household members earn from employment for wage, own busness actvtes, ncome collected by mgratng to other places, sale of lvestock and lvestock by-products. Income from agrcultural output s not ncluded whle no one of the sample households was found to be wth land ownershp. Of all the prevously mentoned sources of ncome, ncome from own busness actvtes s expected to contrbute more for the dfference n the ncome level between the camel owners and nonowners, because ncome from salt trade s ncluded under ths category. Income from farm output s not ncluded as a source of ncome for the sample households; because no one of the households was found wth agrcultural landholdng. Moreover, snce ncome may produce naccurate data, we also consder the followng two alternatve measures to evaluate the contrbuton of camel salt transportaton servce on the lvelhood of the camel owner pastoral households. ) Fxed asset formaton (ass_value):- The total fxed asset formaton of a partcular household s measured by summng up the monetary value of the fxed assets of the household such as house, rado, tape recorder, televson, moble phone, bed, refrgerator etc. It s a contnuous varable and s used as an alternatve lvelhood ndcator n ths study. ) Lvestock ownershp (tlu):- s a contnuous varable quantfyng the number of lvestock owned by a household measured nterms of tropcal lvestock unt (TLU). Lvestock are source of ncome and are consdered as a wealth ndcators for ther owners. Lvestock owners can convert ther lvestock nto cash at any tme that they want; by sellng them or sellng ther by-products such as ther meat, mlk, 36

49 butter, skns and hdes. Ths sales value s one source of ncome for the households and s expected to create them a capacty to buy fxed assets. 3. Camel rent decson (cmlrent_dec):- Almost all of the camel owner pastoral households n the study area are ganers from ther camels salt transportaton servce; but not n the same way. These households do have ther own respectve ndependent decson makng power. As a result, some of them may decde to always rent out ther camels for other salt transporters and receve camel rent; some of them may also decde to sometmes rent out ther camels and sometmes transport salt by themselves; and some fracton of them may decde to never rent out ther camels for others but to always transport salt by ther own. Ths mples that, a partcular household do have three alternatves about ts camels rent decson; to never or sometmes or always rent out them. Hence, the varable camel rent decson s an ordered dscrete varable wth values of 1, 2 and 3 n whch 1 represents the camel owner pastoral households those who never, 2 those who sometmes and 3 those who always rent out ther camels for other salt transporters. 4. Independent varables:- ) Households demographc characterstcs Sex of the household head (sexhh):- s one of the ndependent varables of the analyss. It s a dummy varable n whch 1 s assgned for male household heads and 0 for female household heads. Male-headed households are expected to have a hgher probablty of ownng camels than ther counterparts due to ther better capablty of herdng camels whch s a dffcult task. In addton, the salt mnng area s somehow far from the study center, Berhale kebelle. If an ndvdual resdence of the kebelle wants to transport salt usng camels, he s expected to walk for days on foot (walkng long hours per day) to arrve there. Due to the effort exerton that these tasks demand, t s dffcult for females to do these actvtes by themselves. Hence, they are expected to have a lower lkelhood of ownng camels than males. As a result of ther hgher lkelhood of camel ownershp and the expected economc contrbuton of camel salt transportaton servce, male-headed households are expected to be more lkely capable of havng hgher ncome, much number of lvestock, more asset value and better lvelhoods than the female headed households. Age of the household head (agehh):- s a contnuous varable whch s expected to affect all of the outcome varables. It s supposed that, an ncrease n age of the household head causes an ncrease n ncome of the household untl the stage n whch the ncrease n age s wthn the productve labor force. However, as the household head gets older and older, hs manageral 37

50 ablty and physcal capablty s expected to decrease whch results n decrease n hs ncome and lvestock ownershp. Smlarly, as far as ther camel rent decson s concerned, elder household heads are expected to more lkely rent out ther camels for other salt transporters snce they can t transport salt by themselves due to the dffculty of the task. In addton, the elder household heads are more lkely expected to have less probablty of camel ownershp because they do dstrbute ther camels among ther marred chldren. Lteracy of the household head (educhh):- s a dummy varable n whch 1 s assgned to lterate household heads and 0 s assgned for the llterate ones. Lterate household heads are expected to have better opportunty to be employed for wage n organzatons n whch sklled labor and professonal s demanded. Due to ths fact, ther lkelhood of partcpatng n salt transportaton and camel ownershp n general s expected to be low. Moreover, household heads wth relatvely hgher level of educaton are supposed to have better lvelhood due to ther ablty n money and tme management and ther opportunty to have good job wth better ncome than those who are uneducated. In general, beng lterate s expected to decrease the probablty of camel ownershp, ncrease the lkelhood of rentng out camels and have a postve mpact on lvelhoods. Number of adult members n the household (adulthh):- s a contnuous varable measured n number of household members of age between 15 and 64 years. It s the summaton of both male and female adult members n the household. Households wth large number of adult members are supposed to have hgher lkelhood of ownng camels than those wth few adult members, due to ther better capablty of purchasng camels and undertakng the heavy tasks of herdng camels. Hence, the varable s expected to have a postve mpact on the households probablty of camel ownershp and on the lvelhood ndcator outcome varables. Number of adult males n the household (adumale):- s a contnuous varable whch quantfes the number of male adult members n the household. Large number of male adult members s supposed to ncrease the lkelhood of self-salt transportaton decson and decrease the lkelhood of rentng out camels of the household. Adult rato for the household (adul_rato):- s the rato of male adult members to female adult members n the household. Even f havng large adult members n a household s supposed to have a postve mpact on camel ownershp of that partcular household, dentfyng ether 38

51 havng large number of male adult members per female adult member or vce versa ncreases the lkelhood of ownng camels s mportant. Therefore, t s hypotheszed that, an ncrease n the rato have a postve mpact on the lkelhood of camel ownershp and the outcome varables of the household. ) Soco-nsttutonal and economc factors Access to credt (acc_credt):- s a dummy varable takng a value of 1 f the household head has ever had an access to credt from any type of credt provder and 0 f the household head has never been borrowed from any lender. Credt creates a capacty for a household to purchase a fxed asset or lvestock such as camel at a tme whch may be mpossble or take long tme f t was thought to be purchased by tmely savng from ones ncome. Hence, access to credt s expected to ncrease ones lkelhood of ownng camels. In addton, to transport salt, a partcular salt transporter s expected to pay tax for the salt that he s gong to transport before gong to the salt mnng area. Ths mples, some amount of ntal captal s needed to transport salt. Thus, households wth an access to credt can have better capacty to transport salt by themselves than ther counter parts and the varable s expected to have a negatve effect on camel owner households lkelhood of always rentng out ther camels. Camel ownershp of parents of the household head (famexp_co):- s a dummy varable where 1 s assgned to those household heads wth camel owner parents and 0 for those wth camelnon-owner parents. It s expected that, the probablty of camel ownershp of those household heads wth camel owner parents be hgher than ther counterparts. Ths s because; they have a hgher lkelhood of beng experenced on how to herd camels and gettng camels from ther parents n the form of gft or nhertance. Number of camels owned by the household (camels_hh):- n the study area, almost all of the camel owner pastoral households do partcpate n salt trade ether drectly by transportng salt or ndrectly by rentng out ther camels. Ths ndcates that, ths salt trade s one of the sources of ther ncome n whch ts amount depends on the number of camels that they own. The number of camels s a contnuous varable and s hypotheszed to have a negatve effect on the lkelhood of rentng out camels. As the number of camels that a household owns ncreases, ts probablty of rentng out them for other salt transporters s expected to decrease as far as ther number s large enough to make the household head proftable by salt self- transportng. 39

52 In general, ndependent varables whch are expected to affect the lkelhood of households camel ownershp are: sex, age and lteracy of the household head, access to credt, camel ownershp of parents of the household head, number of adult members n the household and adult rato for the household. On the other hand, determnant varables whch are expected to nfluence households camel rent decson are: age and lteracy of the household head, number of adult male members n the household, access to credt, ncome of the household and number of camels owned by the household. Table 3.1 Descrpton of dependent and ndependent varables used for analyss n the study Varable Name Treatment varable camel_ownp Lvelhood ndcators ncomehh tlu ass_value Decson varable cmlrent_dec Independent varables sexhh agehh educhh acc_credt famexp_co camels_hh adulthh adumale adul_rato Descrpton Households camel ownershp (1=Ownershp, 0=Non-ownershp) Annual ncome of a household (Contnuous) Tropcal lvestock unt of a household (Contnuous) Value of fxed assets for a household (Contnuous) Camel owner pastoral households camel rent decson (1=Never, 2=Sometmes, 3=Always) Sex of the household head (1=Male, 0=Female) Age of the household head (Contnuous) Lteracy of the household head (1=Lterate, 0=Illterate) Access to credt (1=Yes, 0=No) Camel ownershp of household head s parents (1=Yes, 0=No) Number of camels owned by the household (Contnuous) Number of adult members n the household (Contnuous) Number of adult male members n the household (Contnuous) Adult rato for the household (Contnuous) 40

53 CHAPTER FOUR 4. RESULTS AND DISCUSSIONS The raw data whch s collected from 250 sample households through dspersng of structured questonnare and the data collected from FGDs s analyzed usng descrptve and econometrc methods of data analyss. In ths chapter, results of the descrptve and econometrcs estmatons wll be presented. The descrptve analyss presents, the analyss of soco-economc and demographc characterstcs of the sample households; salt transportaton and camel rent experences of the camel owner pastoral households; the proft proporton of the salt transporters out of the total proft of the salt trade value chan partcpants; the returns from camel salt transportaton servce, the contrbuton of camel salt transportaton servce for employment and for the natonal economy usng statstcal technques lke: tables, means, percentages, ratos, frequences and charts. In the econometrc analyss, factors those determne the households camel ownershp are dentfed by logstc regresson of the treatment over some ndependent varables. Moreover, the mpact of camel ownershp as well as salt trade value chan partcpaton on the lvelhoods of the camel owner pastoral households s analyzed by comparng the lvelhoods of the camel owners wth the non-owners usng propensty score matchng. Secondly, ordered probt regresson model s employed to dentfy the determnant factors behnd households camel rent decson. 4.1 Descrptve Analyss The detaled descrpton and summary statstcs of all the varables used for analyss s presented n Annex I and IV Households Soco-Economc and Demographc Characterstcs Demographc Characterstcs Out of the total 250 sample households from whch the survey s conducted, the 115 (46 percent) are camel owners and the remanng 135 (54 percent) camel non-owners. Regardng sex of the household heads, 230 (92 percent) of the households are male headed and the remanng 20 (8 percent) are female headed. The female headed households are most lkely camel non-owners n whch only 6 (30 percent) of them are camel owners. All of the respondent household heads were Muslms n whch 228 (91.2 percent) of them are pastoralsts 41

54 and the remanng 22 (8.8 percent) are non-pastoralsts. The pastoralsts nclude all of the camel owner pastoral households and most of the camel non-owner households whch herd other types of lvestock. The mean household sze for the total sample households s 4.8 whch range from a mnmum of 1 to a maxmum of 10 members n a household. 239 (95.6 percent) of the respondent household heads are marred and the remanng 11 (4.4 percent) are unmarred n whch 3 sngle, 2 separated and 6 wdowed household heads are merged together. The mean age of the household heads s 37.3 wth a mnmum and maxmum of 20 and 68 years old respectvely. As far as lteracy of the household heads s concerned, t ncludes two categores; lterate and llterate. The lterate household heads are those who have ever been attended formal school whatever ther level of schoolng s (ther hghest educatonal achevement could be prmary, junor, secondary or post-secondary). On the other sde, the llterates category ncludes those household heads that have never been attended formal school even f they could have the ablty of wrtng and readng. Accordngly, 190 (76 percent) and 60 (24 percent) household heads are found to be llterate and lterate respectvely. Out of the total 115 camel owner household heads, only 20 (17.4 percent) of them are observed that they are lterate. The presence of a number of adult members n household, partcularly male adult members, whch s supposed to contrbute for lvelhood postvely, was consdered for each household. The mean number of male adult members s 1.59 for each household whch ranges from havng no adult male member at all up to havng 6 male adult members n a household. However, the camel owner pastoral households are found havng at least 1 male adult member. Ths mples that, camel owner households are more lkely wth much number of adult male members than the camel non-owners. The adult rato whch s the rato of adult male members to adult female members of a household for the entre sample s around 1.30, wth mnmum and maxmum values of 0 and 6 respectvely. The mean number of dependent members per ndependent member.e. dependency rato s 0.36 for the total sample households rangng from households wth no dependent members to households wth 0.75 dependents per ndependent. Soco-Insttutonal Characterstcs The sample households were asked whether they have an access to credt or not; and for those who sad yes, for what purpose most of the tme they borrow. The data of ther responses 42

55 shows that, only 17.2 percent or 43 household heads have credt experence, but the remanng 207 (82.8 percent) household heads sad, no! I have never been borrowed money from any one before. The major potental reasons for the borrowng of the household heads wth an access to credt are: to buy lvestock, to buy house propertes, to pay salt tax, to fnance trade, to buy food and to cover health expenses of household members wth respectve percentage shares of 20.9 percent, 16.3 percent, 16.3 percent, 16.3 percent, 13.9 percent and 4.7 percent. The remanng 11.6 percent households do borrow money for dfferent purposes ncludng purchasng camel feed, house and buldng materals. Regardng extenson servce partcpaton of households, 159 (63.6 percent) of the total sample households do partcpate n health extenson servce. Of whch, the 66 (41.5 percent) households are camel owners but the remanng 93 (58.5 percent) are camel non-owners. There are dfferent local nsttutons n the kebelle n whch ths survey s conducted. The number of partcpant households n these local nsttutons out of the total 250 sample households s referred to be 105 (42 percent). Of whch, the 37 (35.2 percent) households are from the camel owner group of households. Based on the number of member households, the major local nsttutons n whch those 105 households do manly partcpate are: Assaele salt trade cooperatve, equb, both equb and Assaele salt trade assocaton and Idr wth 55 (52.4 percent), 34 (32.4 percent), 6 (5.7 percent) and 4 (3.8 percent) respectve number of member households. The remanng 6 (5.7 percent) households have stated that they partcpate n other small neghborhood level assocatons. Households resdng n dfferent kushets of Berhale kebelle do not have an equal access to basc socal servces such as electrcty and drnkng water. Bascally, drnkng water, whch s supposed to be basc for the healthness of a socety, s not as such avalable for all households. Of the total 250 sample households, only 185 (74 percent) households do have an access to clean water and only 181 (72.4 percent) households do have an access to electrcty. However, t has to be noted that, ths naccessblty of the households to drnkng water and electrcty s not only due to unavalablty of the servces, but also due to the nfluence of the resdence type of the households: whether t s temporary or permanent resdence. Those households wth permanent settlement are more lkely wth an access to clean water and electrcty than the nomadc households. The random samplng method used n ths survey to select the respondent households has made the proporton of the households to be 4.1 settled households per 1 movable nomadc 43

56 household. I.e. the 201 (80.4 percent) households found to be permanently resdent households and the remanng 49 (19.6 percent) nomadc pastoral households. Economc Characterstcs The mean annual ncome of the 250 sample households s Brr wth mnmum and maxmum amounts of 0.00 Brr and Brr respectvely. The 0.00 Brr annual ncome concerns the households those who lve by recevng remttance from ther non-resdent household members and ad from the government. They don t partcpate n ncome earnng actvtes and some proporton of ther consumpton consttutes ther lvestock by-products. The result from the raw data, concernng the number of partcpant households n each of the ncome earnng actvtes wthn the year 2012/13 shows us that, out of the total sample households, members of the 118 (47.2 percent), 131 (52.4 percent), 7 (2.8 percent) and 76 (30.5 percent) were employed for wage, engaged n own busness actvtes (such as handcraft, trade n lvestock, sellng salt, transportng goods by pack anmal, and sellng wood and charcoal), mgrated to dfferent places seekng job out of the kebelle and sellng lvestock byproducts (manly goat mlk, butter and meat) respectvely. Sellng lvestock by-products s common for the pastoral households n Berahle woreda and s one of the sources of ther ncome and consumpton. However, camel mlk s not marketable there, snce sellng and purchasng of camel mlk s forbdden n the area. Out of the 118 households wth employed for wage household members, only 40 households are from the camel owner group of households. Ths ndcates that, the members of camel nonowner households do more lkely partcpate n employment for wage than members of camel owner households. Households lvestock ownershp, measured n TLU converson factor, s wth a wde gap between the treated and untreated groups. The mean TLU for the total sample households s 4.97 wth a mnmum of 0 and maxmum of 29.6 values per household. Dfferently, the mean TLU value for the camel owner pastoral households s 8.6, whch s approxmately twce of the mean TLU value for the total sample households. The mnmum TLU value for the camel owner pastoral households s The converson factors used to estmate tropcal lvestock unt (TLU) are depcted n Annex III. Regardng fxed asset formaton of the households, the wealthy household has an asset value of 791,400 brr, and on the other extreme, there are households wth 0 asset values ndcatng that 44

57 they are not owners of any knd of fxed asset. Mergng these two extreme ponts together, the mean asset value of the total sample households s averaged to Brr per household. As far as the number of camels owned by the camel owner pastoral households s concerned, the least number of camels owned by those households s 1 and the households wth the largest number of camels do own upto 20 camels. The average number of camels for the households s 5.7. The number of camel holdngs of a partcular camel owner household may reman constant, decrease or ncrease over tme due to dfferent factors. Of the 115 camel owner pastoral households of the survey, 53 (46.1 percent), 41 (35.6 percent) and 21 (18.3 percent) households has reported that, the number of ther camels s decreasng, constant and ncreasng over tme respectvely. The man reasons mentoned by the households for the ncrement of ther camels are reproducton due to ther female camel ownershp, purchase due to the ncrease n ther ncome and mprovement n ther lvelhood, nhertance and gft wth a respectve percentage shares of 52.4 percent, 33.3 percent, 9.5 percent and 4.8 percent. On the other sde, the man factors for the decrement of the number of camels over tme are observed to be recurrent drought (shortage of food and water), dsease that causes death, rustlng and sales wth a percentage shares of 45.3 percent, 35.9 percent, 5.7 percent and 3.8 percent respectvely. The average number of years that the camel owner households have spent n camel producton s 17.9 wth a mnmum and maxmum of 1 and 50 years respectvely. Camel ownershp of parents of the household head whch s expected to be one of the major determnant factors affectng household s camel ownershp was surveyed for each of the sample households. Accordng to the responses of the household heads, only parents of the 80 (32 percent) household heads were camel owners. But the majorty 170 (68 percent) household heads are from the camel non-owner famles. Of the 80 household heads wth camel owner parents, the 67 (83.75 percent) are currently camel owners too. Ths ndcates that, household heads wth camel owner parents do more lkely contnue to be camel owners than those wth camel non-owner parents. The short summary of the sample households major soco-economc and demographc characterstcs s presented as follows. 45

58 Table 4.1 Descrpton of Households Soco-Economc and Demographc Characterstcs (Dummy and Categorcal Varables) Varable Categores Frequency (% of the total sample) sexhh male female 20 8 educhh lterate llterate relgon Muslm martalhh marred unmarred acc_credt yes no resd_type permanent temporary camel_ownp yes no pastor_hh pastoral non-pastoral famexp_co yes no cmls_overtme decreasng (for camel owners) constant ncreasng empt_wage yes no Source: Own Survey Data,

59 Source, Own Survey Data, 2013 Table 4.2 Summary of Basc Households Soco-Economc and Demographc Characterstcs Varable Combned Camel Owners Camel Non-owners Dfference t-test Pearson Mean Standard Error Mean Standard Error Mean Standard Error Mean Standard Error ch2(1) sexhh agehh hhsze *** martalhh educhh ** adulthh adumale adul_rato deprato * empt_wage *** acc_credt famexp_co *** resdence pastor_hh *** camels_hh *** yrs_cmlprod *** ncomehh ** tlu *** ass_value Dfference = mean (camel non-owners) - mean (camel owners) 47

60 4.1.2 Households Salt Transportaton and Camel Rent Experences More than 99 percent of the camel owner pastoral households of Berahle woreda do engage ther camels for salt transportaton. Ths ndcates that, almost all of the camels wth carryng capacty found n the woreda are used to transport salt from the mnng area to the market. The camels may undertake ths task ether under the leadershp of ther owners or the grooms who rented n them from ther owners. Majortes, but not all of the camel owner pastoral households, do drectly partcpate n salt transportaton. Some of them do rent out ther camels for other salt transporters and gan from the salt trade n the form of camel rent. Accordngly, we can classfy the 115 camel owner pastoral households under three categores; pure salt transporters (those who always transport salt by themselves and Never rent out ther camels for other salt transporters), pure renters (those who Always rent out ther camels for other salt transporters and never transport salt by themselves) and those households who undertake both actvtes ( Sometmes rent out ther camels and sometmes transport salt by themselves). The majortes or 67 (58.3 percent) households are categorzed under the pure salt transporters and contrarly, the 29 (25.2 percent) households are under the category of pure camel renters. Only 19 (16.5 percent) of the households do have an experence of nvolvng n both actvtes. The short summary of the number and proporton of households under each category s presented n the table and fgure below. Table 4.3 Salt transportaton and camel rent experences of the camel owner pastoral households Salt transportaton Camel rent experence experence No Yes Total No Yes Total Source: Own Survey,

61 Fgure 4.1 The proporton of the camel owner pastoral households under the three categores of camel rent decson Camel Owner Pastoral Households' Salt Transportaton and Camel Rent Experences 17% 25% 58% Households those who have "Never" been rented out ther camels Households those who have been "Sometmes" rentng out ther camels Households those who were "Always" rentng out ther camels Source: Own Survey, 2013 For the 86 households wth salt transportaton experence, the mean number of years that they have spent n salt transportaton s 12 years whch ranges from 1 to 42. Smlarly, for the 48 households who have an experence of rentng out camels, the mean number of years that they have spent n camel rent s 9 years wth a mnmum and maxmum of 1 and 35 years respectvely. Ths dfference shows the camel owner s longer experence of transportng salt rather than rentng out camels. A partcular household can have ts own reasons those motvate t to rent out ts camels for other salt transporters. The man motve that was observed from the response of the sample households wth more lkely camel rentng experence s that, the satsfacton that I get from my lesure tme by rentng out my camels s better than the ncome I can get by transportng salt myself. On the contrary, majorty of the households wth better experence of transportng salt rather than rentng out camels have stated the man factors that hnder them not to rent out ther camels for other salt transporters as I want to use my camels by myself because I don t beleve that anybody can take care of them lke that of me and the ncome that I get by transportng salt myself s better than the ncome that I could get from camel rent. 49