Polcy Optons for Improvng Market Partcpaton and Sales of Smallholder Lvestock Producers: A case study of Ethopa Smeon Ehu 1, Samuel Benn 2 and Zelekawork Paulos 3 Abstract Market access plays an essental role n assurng better ncome and welfare levels for smallholder lvestock producers, and thus contrbutes to poverty allevaton. Ths s even more so n the Ethopan context where lvestock play an essental role n the economy. Makng use of the Heckman estmaton procedure, ths paper dentfes polcy and technology optons to ncrease partcpaton and sales of smallholder producers n lvestock markets n Ethopa, based on data from 934 household surveys conducted between 1999 and 2001 n the hghlands of Tgray and Amhara regons n northern Ethopa. The analyss demonstrates that physcal captal (ownershp of dfferent speces of lvestock and landholdng) and fnancal captal (crop ncome and non-farm ncome) are the man factors nfluencng market partcpaton and sales. Educaton was also found to postvely affect value of sales of dary products. Dstance to markets and towns were not found to be sgnfcant. We conclude that n the case of Ethopa, constrants to producton of lvestock and lvestock products (e.g. captal to purchase anmals, feed, and processng equpment) are the man factors lmtng partcpaton and sales n lvestock markets. Key words: market partcpaton and sales, lvestock markets, Ethopan smallholder farmers Ths draft s not meant to represent n any way the vews of the World Bank or any of ts Executve Drectors, the Internatonal Food Polcy Research Insttute. Draft prepared for presentaton at the 27 th Conference of the Internatonal Assocaton of Agrcultural Economsts (IAAE), 16-22 August 2009, Bejng, Chna. 1 Sector Leader, Sustanable Development World Bank, Abuja, Ngera. sehu@worldbank.org 2 Internatonal Food Polcy Reseearch Insttute, Washngton DC, USA 3 Internatonal Food Polcy Reseearch Insttute, Adds Ababa, Ethopa
Polcy Optons for Improvng Market Partcpaton and Sales of Smallholder Lvestock producers: A case study of Ethopa 1. Introducton In a recent study, Delgado et al. (1999) projected that per capta consumpton of lvestock products wll ncrease by about 50% from 1993 to 2020, wth most of the ncreases attrbuted to developng countres, as a result of populaton growth, urbanzaton, and rsng ncomes. In sub-saharan Afrca (SSA) total consumpton of meat and mlk s expected to double between 1997 and 2020 to reach 11.3 and 35.4 mllon tonnes (Ehu et al. 2002). Ths expected ncrease n demand for anmal products has profound mplcatons for food securty and poverty allevaton among rural people n SSA. In partcular, the expected demand for lvestock products presents expandng market opportuntes for poor smallholder lvestock producers. Therefore, mprovng access to markets of poor smallholder lvestock producers can help them beneft from the rapdly growng demand for lvestock products. The Internatonal Lvestock Research Insttute (ILRI) and ts partners have dentfed that encouragng market partcpaton of smallholder lvestock producers s a major pathway for gettng rural people out of poverty and mprovng ther food securty, as lvestock contrbute to the lvelhoods of more than two-thrds of the world s rural poor (ILRI, 2002). Market access plays an essental role n assurng better ncome and welfare for smallholder lvestock producers through varous channels. By rasng ncomes, markets ncrease purchasng power, whch creates demand for consumer goods and, thus, enhances welfare. In addton, by creatng demand for producton nputs and nvestment goods, markets promote economc growth. Markets also facltate the accumulaton of assets, provde the opportunty for mproved nutrton and balanced dets and, therefore, help allevate poverty. Wth lmted opportuntes for access to markets, lvestock are often kept beyond ther optmum productve levels (e.g., n terms of producng mlk, meat and draft power) where there s no value-added from feedng and grazng, resultng n overstockng and overgrazng. Furthermore, mprovng market access can also ncrease productvty and supply of lvestock, wthout whch the antcpated ncrease n demand for lvestock products may not be met. 1
Although mprovng access to markets has numerous benefts, there may be potental offsettng effects assocated wth removng the constrants lmtng market partcpaton and supply. For example, market access s restrcted prmarly by hgher transacton costs created by poor or lack of roads. Thus, whle buldng roads or mprovng road condtons may ncrease access to markets, ncreasng partcpaton and supply n the short run, t may also ncrease non-farm employment opportuntes and rase the opportunty cost of lvestock producton, reducng partcpaton and supply n the short run n the long run, complcatng polcy mplcatons of road development on market partcpaton and supply decsons. Smlar complcatons arse for other factors that lmt market access, ncludng lack of fnancal and physcal captal. Ths paper provdes an emprcal bass for dentfyng optons to ncrease partcpaton and sales of smallholder producers n lvestock markets n Ethopa, based on data from 934 household surveys conducted between 1999 and 2001 n the hghlands of Tgray and Amhara regons n northern Ethopa. Identfyng polcy optons for mprovng market partcpaton and sales of lvestock and lvestock products s partcularly mportant n the Ethopan context as lvestock play a very promnent role n the Ethopan economy. Wth a per capta gross natonal ncome (GNP) of US$ 100 and 31% of the populaton lvng wth less than US$ 1 a day (World Bank 2003 4 ), Ethopa s one of the poorest countres n the world. Ethopa has the largest lvestock populaton n Afrca where lvestock contrbute about 30-35% to agrcultural gross domestc product (GDP) and about 13-16% to overall GDP (Befekadu and Berhanu 2000). Moreover t s estmated that over 85% of farm cash ncome s generated from lvestock (bd.). Over the last decade (between the Ethopan fscal years1991/92 and 2000/01), the share of lvestock n agrcultural exports and total exports averaged 15% and 13% respectvely 5. Thus, dentfyng polcy optons for ncreasng market partcpaton and sales of lvestock and lvestock products s beleved to sgnfcantly mprove the ncome and welfare 4 The fgures for per capta GNP and the percentage of the populaton lvng below US$ 1 a day are reported for the years 2001 and 1995 respectvely. 5 Author s calculatons usng data obtaned from the Natonal Bank of Ethopa (the data orgnated from the Ethopan Customs Authorty). 2
of smallholder lvestock producers and help allevate poverty n Ethopa. The next secton presents the conceptual framework and hypotheses regardng market partcpaton and supply decsons. Secton 3 presents the data. Emprcal approach, results and dscusson are presented n Secton 4, and conclusons and polcy mplcatons n Secton 5. 2. Conceptual framework and hypotheses The econometrc approach used s the Heckman (1979) two-step procedure. Frst the factors that nfluence the decson to sell lve and/or anmal products are estmated. Second the factors that affect the amount of sales are dentfed. The decson to partcpate n the market (the selecton equaton), specfed n equaton (1), s estmated by maxmum lkelhood as an ndependent probt model from the entre sample of market partcpants and non-partcpants. z w u, (1) where u ~ N(0,1 ) and, z 1 f z 0, z 0 f z 0 The sample selecton bas, what Heckman (1979) refers to as the nverse of Mll s rato, s computed from the parameter estmates of the selecton equaton for each observaton n the selected sample (Greene 1993), and s represented by: ( w ) (2) ( w ) where and are, respectvely, the densty and dstrbuton functons. 3
The level of sales, y, specfed n equaton (3), s observed only f w + u >0, and s estmated by ordnary least squares where the vector of nverse Mll s ratos s ncluded as an addtonal regressor. y (3) x where ~ N(0, ) The error terms of the market partcpaton and the sales equaton are correlated as the Heckman procedure assumes that the decsons pertanng to market partcpaton and the amount of sales are nterdependent. The correlaton coeffcent for the error terms u and s represented by, where u and are bvarate and normally dstrbuted (Greene 1993). corr, (4) u By ncorporatng the nverse Mll s rato as an addtonal explanatory varable n the sales equaton, the second stage of the Heckman procedure corrects the sample selecton bas. To summarze, the conceptual framework for analysng market partcpaton and supply decsons are gven by: Partcpat on f ( X P ) ; Partcpaton [0,1] (5) Sales f X ) ; Sales>0 (6) ( S Where X P and X S are the factors affectng partcpaton and sales, respectvely, and X P X S. The factors nclude transacton costs (dstance to roads, markets and towns, transport avalablty, labour and populaton densty), human captal (age, educaton, gender, extenson tranng), physcal captal (number of lvestock producng stock, farmland), and fnancal captal (crop ncome, non-farm ncome, credt). In general, farmers frst decde to partcpate n the market when t s proftable to do so, and then decde on how much to sell. Thus, the factors mentoned above are those that affect proftablty by affectng marketng costs. Transacton costs. In general, we expect that farmers wth lower transacton costs wll be more lkely to partcpate n markets and sell more, as they are more lkely to recover ther 4
producton and marketng costs. Thus, we expect farmers lvng closer to roads, markets or towns, havng means of transportaton, or more labour to partcpate and sell more lvestock and lvestock products. Smlarly, farmers n more densely populated areas wll face greater demand for lvestock and lvestock products and have lower search costs, and thus expected to postvely affect partcpaton and sales. However, the mpacts may also be negatve. For example, better access to roads, markets or towns may ncrease the opportunty cost of labour and captal n lvestock producton and marketng, where alternatve employment opportuntes exst and the returns to labour and captal are hgher, reducng partcpaton and sales. Larger households also tend to have more dependents and ther producton actvtes may be more subsstence orented, dmnshng market partcpaton and sales. More dense areas also tend to be assocated wth greater demand for other servces, creatng opportuntes for other employment, ncreasng the opportunty cost of labour and captal nvestment n farmng. Human captal. We expect experence to be negatvely assocated wth market partcpaton, as older households tend to have more dependents and more subsstence producton actvtes. Gender s also expected to nfluence market partcpaton, and expect female-headed households and households wth more female members to have a postve effect, especally wth respect to dary products, snce females are prmarly nvolved n the producton actvtes. Educaton and extenson may have mxed mpacts. On one hand, educaton and extenson enhances the skll and ablty to better utlze market nformaton, whch may reduce marketng costs and make t more proftable to partcpate n the market. Ths s very mportant gven the general belef by farmers that markets are not favourable. Educaton, however, rases the opportunty cost of labour and may reduce the proftablty of lvestock producton and market partcpaton by farmers where alternatve employment opportuntes exst and more proftable to engage n. Smlarly, extenson tranng n croppng actvtes ncreases the proftablty of crop producton relatve to lvestock, reducng partcpaton n lvestock markets. Physcal captal. Ownershp of more lvestock producng stock s expected to have a postve effect on partcpaton and sales, although dfferent speces may have dfferent 5
effects, dependng on the roles and functons they play n the farmng system. For example, snce cows are prmarly kept for mlk producton, ownershp of cows s more lkely to have a postve effect on dary (mlk, butter and cheeses) market partcpaton and sales, whle ownershp of young cattle, small rumnants and poultry are more lkely to have a postve mpact on sales of lves anmals. Ownershp of bulls and oxen, on the other hand, may have non-postve effects, as they are kept for reproductve purposes and draught power, respectvely. Physcal assets such as land may have ndrect postve mpacts by enablng farmers to overcome credt constrants, where land can as collateral for credt, and allow them to adopt mproved technologes that ncrease productvty. Fnancal captal. Ths s expected to have mxed mpacts. For example, households wth hgher crop ncome may ndcate low proftablty of lvestock producton, dmnshng partcpaton and sales. On the other hand, households wth hgher crop ncome may be more lkely to dversfy ther producton actvtes nto lvestock and adopt mproved technologes that could potentally ncrease productvty. Smlarly, hgher non-farm ncome of households may ndcate low proftablty of lvestock producton, dmnshng partcpaton and sales, although exogenous ncome, such as remttances, may ncrease adopton of mproved technologes and ncrease lvestock productvty and, hence, partcpaton and sales. Smlar to extenson, credt for lvestock actvtes (such as fattenng or purchase of cow) s expected to have postve effects, whle those gven for croppng and other actvtes may rase the opportunty cost of lvestock producton, hence reducng partcpaton n lvestock markets and sales of lvestock and lvestock products. Factors affectng market partcpaton but not sales. It s clear to see why factors that affect the decson by farmers on how much to sell wll also affect the decson on whether or not to partcpate n the market. These factors prmarly affect productvty of lvestock and, hence, determne how much s produced, so that greater producton leads to greater marketable surplus. However, there are factors that may nfluence the decson on whether or not to sell, but not how much to sell. Such factors, ncludng many of the household characterstcs, do not have nfluence on the productvty of lvestock. Ths ssue s more 6
relevant when dscussng the econometrc model n Secton 4, wth respect to dentfcaton and estmaton of the sales equaton. 3. Survey and data The varables used n ths analyss were constructed from data collected n a sample survey conducted among 198 vllages (communtes) and 934 households n Tgray and Amhara regons of northern Ethopa between 1998 and 2001. A stratfed random sample of 99 Peasant Assocatons (PAs, usually consstng of 4 or 5 vllages) 6 was selected from hghland areas (above 1500 m.a.s.l.) of the two regons. Strata were defned accordng to varables assocated wth mosture avalablty (one major factor affectng agrcultural productvty), market access and populaton densty. In Amhara regon, secondary data was used to classfy the weredas (dstrcts) accordng to access to an all-weather road, the 1994 rural populaton densty (greater or less than 100 persons/km 2 ), and whether the area s drought-prone (followng the defnton of the Ethopan Dsaster Preventon and Preparedness Commttee). Two addtonal strata were defned for PAs where rrgaton projects are found. In each of the 10 strata, 4-5 PAs were randomly selected. From each sample PA, 2 vllages were randomly selected, for a total of 98 vllages. In each vllage, 4-5 households were randomly selected, for a total of 434 households. In Tgray regon, PAs were stratfed by whether an rrgaton project was present or not, and for those wthout rrgaton, by dstance to the wereda town (greater or less than 10 km.). A total of three strata were defned n Tgray, wth 54 PAs randomly selected per strata. PAs closer to towns and n rrgated areas were selected wth a hgher samplng fracton to assure adequate representaton. Four PAs n the northern part of Tgray could not be studed due to the war wth Ertrea. From each of the remanng PAs, 2 vllages were randomly 6 The Peasant Assocaton (PA) s the lowest admnstratve unt n Ethopa. 7
selected, and from each vllage, 5 households were randomly selected. A total of 50 PAs, 100 vllages, and 500 households were then surveyed. Informaton collected ncludes sales of lvestock and lvestock products, ownershp of varous speces of lvestock, and access to markets, nfrastructure and servces, household composton and assets. The data were supplemented by secondary geographc populaton nformaton. The farmng system n the northern Ethopan hghlands s characterzed by mxed crop-lvestock producton, where households n the study area hold an average of 1.2 hectares of farmland, wth about 82% of the households n the study ownng up to 17 (average of 3.3) tropcal lvestock unts of cattle and small rumnants. On average, the number of lvestock owned per households was 1.4 oxen, 0.8 cows, 0.5 TLUs of other cattle (bulls, young bulls, hefers and calves), and 2.5 sheep and goats. Ownershp of poultry and pack anmals was also common, averagng 2.4 and 0.3 per household, respectvely. Sales were restrcted to lve anmals and dary products (mlk, butter and cheese). Among the households that owned cattle and small rumnants, about 50% of them sold lve anmals, averagng 1 TLU or Ethopan Brr (EB) 582. 7 Wth respect to dary products, only 264 households produced mlk, butter or cheese, and about 50% of them dd sell some, averagng EB 141 per household. The total value of lve anmals and dary products sold was EB 558 per household. 4. Econometrc approach and results Econometrc approach and estmaton We estmated three sets of regressons where the dependent varables are: 1a) whether or not households sold lve anmals (cattle and small rumnants) 8, and 1b) value of lve anmals sold; 2a) whether or not households sold dary products (mlk, butter and cheese), and 2b) 7 Durng the survey perod, US$ 1 EB 8.5. 8 We only estmate partcpaton and sales wth respect to cattle and small rumnants, as they consttute the bulk of lve anmal sales by households. 8
value of dary products sold; and 3a) whether or not households sold lve anmals or dary products, and 3b) total value of lve anmals and dary products sold. As dscussed n Secton 2 we estmated the partcpaton equaton by probt, and used Heckman s approach (Heckman, 1979; Maddala, 1983) to estmate the sales equaton, snce sales are observed only for those households that partcpated n the market or sold lvestock or lvestock products. Therefore, sales of lvestock and lvestock products are censored at values greater than zero. In prncple, estmatng the sales equaton ndependently wll result n nconsstent estmates due to sample selecton bas. Thus, we use the nverse mlls rato (IMR) obtaned form the probt estmaton of partcpaton n the least squares estmaton of the sales. Snce the sales equaton uses IMR estmated from another estmaton, the standard errors are based and need to be corrected for. We use the Heckman procedure n STATA software (StataCorp, 2002), whch corrects for the bas n the standard errors. We also tested for ndependence of the partcpaton and sales equatons, whch we faled to reject n all three sets of estmatons. The explanatory varables used n the estmatons are those that nfluence transacton costs and assocated wth human captal, physcal captal, and fnancal captal. Descrpton of all the varables (dependent and explanatory) and summary statstcs are shown n Tables 1 and 2, respectvely. There are two econometrc ssues to deal wth here. Frst, s the dentfcaton of the sales equaton, whch we deal wth by excludng some of human captal varables (ncludng age and sex of household, household sze, proporton of female household members, and dependency rato) used n the probt estmaton of partcpaton from the sales regresson. These varables are expected to have effects on partcpaton but not on sales. The second econometrc ssue s censorng of the market partcpaton varable. Ths problem arses because of the nature of the data and how the effectve sample sze for partcpaton s defned. For example, households that dd not own cattle or small rumnants (or produced these anmals) could not have partcpated n the market for sellng them even f they would have otherwse. Smlarly, households that dd not produce mlk, butter or cheese could not have partcpated n the market for sellng these products. Therefore, the effectve 9
sample sze for partcpaton n the market for cattle, small rumnants, or dary products are restrcted to only those households that owned cattle or small rumnants or produced mlk, butter or cheese, meanng that the market partcpaton varable s also censored. Thus, we frst predct the probablty of ownng cattle or small rumnants or producng mlk, butter or cheese, obtan the IMRs assocated wth ownershp or producton, and use the IMRs n the respectve probt estmaton of the partcpaton equatons. Note that the dentfcaton ssue and correctng the standard errors also apples. Therefore, n addton to the explanatory varables used n the partcpaton equaton (except ownershp of cattle and small rumnants), we ncluded dstance to veternary clnc, dstance to nearest all-weather road, and alttude as nstruments, and then estmated the ownershp and partcpaton equatons usng the heckprob command n STATA (StataCorp, 2002). Results and dscusson The econometrc results for partcpaton and sales are shown n Tables 3 and 4, and the probt results for predctng the probablty of ownng cattle or small rumnants and producng mlk, butter and cheese are shown n the Appendx. Market partcpaton. The results show that partcpaton n lvestock (cattle and small rumnants) and lvestock products (mlk, butter and cheeses) markets s nfluenced to a great extent by physcal and fnancal captal of the household, and not so much by human captal (except dependency rato whch s postvely assocated wth partcpaton) or the physcal dstance from markets and access to nfrastructure. Landholdng unambguously reduces partcpaton n both markets, suggestng crop producton bas. However, ownershp of small rumnants ncreases partcpaton n the market for lve anmals, but reduces partcpaton n the market for dary products, whle ownershp of oxen also reduces partcpaton n the market for dary products. That ownershp of small rumnants ncreases partcpaton n markets s not surprsng because small rumnants requre mnmal nputs and are easy to dspose of. Crop ncome and non-farm ncome ncreases partcpaton n both lve anmals and dary products markets, although crop ncome s assocated wth a reducton n partcpaton n dary products markets. Households located n Tgray, compared to ther 10
Amhara counterparts, were assocated wth lower lkelhood to partcpate n dary products markets, although there was no sgnfcant dfference between the two regons n partcpaton n lve anmals markets. Sales. Here too physcal dstance to markets and towns has no sgnfcant mpact. However, populaton densty, educaton, and physcal and fnancal captal of households sgnfcantly affect sales (value) of lvestock and lvestock products. Areas wth greater populaton pressure are assocated wth lower sales of lve anmals, but greater sales of dary products, suggestng the proftablty of dary producton n densely populated areas. Households wth better educaton, more cows, and greater non-farm ncome are assocated wth greater sales of dary products, whle those wth more young cattle, pack anmals, and crop ncome are assocated wth greater sales of lve anmals. These fndngs renforce the fndngs of Holloway et al. (2000) where they fnd access to anmal stock and ntellectual captal such as extenson ncreases access to dary markets by small holder producers. In a smlar ILRI study conducted n the Phlppnes, Lapar et al. (2002) have also shown that smallholder partcpaton n lvestock markets and sales ncrease wth hgher levels of anmal stock, fnancal captal resources and lower levels of transacton costs. Ownershp of poultry reduces sales, whle locaton n Tgray (or Amhara) has mxed mpacts. In partcular, households located n Tgray are assocated wth greater sales of dary products, whle those located n Amhara are assocated wth greater sales of lve anmals. 5. Conclusons and mplcatons Ths paper examned the factors affectng lvestock market partcpaton and sales, usng the Heckman procedure and data from the mxed crop-lvestock farmng systems n the hghlands Amhara and Tgray regons of northern Ethopa. The results show that physcal captal (ownershp of dfferent speces of lvestock and landholdng) and fnancal captal (crop ncome and non-farm ncome) are the man factors nfluencng market partcpaton and sales, and not the dstance to markets and towns. These suggest that constrants to producton of 11
lvestock and lvestock products (e.g., captal to purchase anmals, feed, and processng equpment) may be the man factors lmtng market partcpaton and sales. Snce lvestock are trekked to markets, ownershp of pack anmals (mules, donkeys, horses, and camels) s crtcal n promotng market partcpaton. Pack anmals serve two purposes of transportng people to the market and commodtes purchased from the market to the home. Educaton was also found to postvely affect value of sales of dary products. Ths suggests that ncome from sale of mlk, butter and cheese can be ncreased through educaton and tranng (especally targetng females) to mprove the qualty of products, whch can n turn attract better prces. 12
References Befekadu Degefe and Berhanu Nega eds. 2000. Annual report on the Ethopan economy: Volume I, 1999/2000. The Ethopan Economc Assocaton, Adds Ababa, Ethopa. 429 pp. Delgado, C., M. Rosegrant, H. Stenfeld, S. Ehu, and C. Courbos. 1999. Lvestock to 2020: The Next Food Revoluton. Food, Agrculture, and the Envronment Dscusson Paper 28. Internatonal Food Polcy Research Insttute (IFPRI), Food and Agrculture Organzaton of the Unted Natons (FAO), and the Internatonal Lvestock Research Insttute (ILRI), Washngton, D. C., 72 pp. Ehu, S., S. Benn, T. Wllams, and S. Mejer. (2002). Food securty n sub-saharan Afrca to 2020. Soco-economcs and Polcy Research Workng Paper 49. Internatonal Lvestock Research Insttute (ILRI), Narob, Kenya, 52pp. Greene, W.H. 1993. Econometrc analyss. Macmllan Publshng Company, New York Cty, USA, 791 pp. Heckman, J.J. 1979. Sample selecton bas as a specfcaton error. Econometrca 47(1):153-161. Holloway, G. J., and S. Ehu. 2002. Expandng market partcpaton among smallholder lvestock producers: A collecton of studes employng Gbbs samplng and data from the Ethopan hghlands, 1998-2001. Soco-economcs and Polcy Research Workng Paper 48. Internatonal Lvestock Research Insttute (ILRI), Narob, Kenya, 77 pp. Holloway, G. J., C. Ncholson, C. Delgado, S. Ehu and S. Staal. 2000. Agrondustralzaton Through Insttutonal Innovaton: Transacton Costs, Cooperatves and Mlk-Market Development n the East Afrcan Hghlands. Agrcultural Economcs 23:279-88. Internatonal Fund for Agrcultural Development (IFAD). 2001. Rural poverty report 2001 The challenge of endng rural poverty. Oxford Unversty Press, New York, USA. 266 pp. Internatonal Lvestock Research Insttute (ILRI). 2002. Lvestock a pathway out of poverty: ILRI s strategy to 2010. Internatonal Lvestock Research Insttute, Narob, Kenya, 24pp. Lapar, Ma. L., G. Holloway, and S. Ehu. 2002. Polcy Optons for Promotng Smallholder Market Partcpaton: A Case Study from the Phlppnes. Soco-economcs and Polcy Research Workng Paper 47. Internatonal Lvestock Research Insttute (ILRI), Narob, Kenya, 21 pp. Maddala, G. S. 1983. Lmted-dependent and qualtatve varables n econometrcs. Cambrdge Unversty Press, New York Cty, USA. 401 pp. StataCorp. 2002. Stata statstcal software: verson 7.0. College Staton, Texas, USA: Stata Corporaton. World Bank. 2003. World Development Report. Oxford Unversty Press. 13
Table 1: Descrpton of varables Varable Dependent varables Whether sold lve anmals Value of lve anmals sold Whether sold dary products Value of dary products sold Whether sold lve anmals or dary products Value of lve anmals and dary products sold Explanatory varables Dstance to nearest market Dstance to dstrct town Populaton densty Educaton Extenson Oxen owned Cows owned Other cattle owned Small rumnants owned Poultry owned Pack anmals owned Landholdng Crop ncome Non-farm ncome Credt Regon Instruments Sex of household head Age of household head Household sze Proporton of females Dependency rato Dstance to nearest all-weather road Dstance to veternary clnc Alttude Descrpton Dummy varable of whether or not household sold lve anmals (1=yes; 0=no) Value of lve anmals sold by household (Ethopan Brr) Dummy varable of whether or not household sold dary products (1=yes; 0=no) Value of dary products sold by household (Ethopan Brr) Dummy varable of whether or not household sold lve anmals or dary products (1=yes; 0=no) Value of lve anmals and dary products sold by household (Ethopan Brr) Walkng tme from peasant assocaton to the nearest market (mnutes) Dstance from peasant assocaton to the dstrct town (klometres) Number of households n peasant assocaton per hectare Average number of years of formal educaton of household members Dummy varable of whether or not household receved extenson (1=yes; 0=no) Number of oxen owned by household Number of local and crossbred cows owned by household Number of other cattle (bulls, young bulls, hefers and calves) owned by household (TLU) Number of sheep and goats owned by household Number of chckens owned by household Number of pack anmals (donkeys, horses and camels) owned by household (TLU) Amount of farmland owned by the household (hectares) Amount of crop ncome of household (Ethopan Brr) Amount of non-farm ncome of household (Ethopan Brr) Dummy varable of whether or not household receved credt (1=yes; 0=no) Dummy varable of regonal locaton of household (1=Amhara regon; 0=Tgray regon) Sex of household head (1=female; 0=male) Age of household head (years) Number of household members Proporton of household members that are female Proporton of household members less than 15 years or greater than 59 years Walkng tme from peasant assocaton to the nearest all-weather road (mnutes) Walkng tme from peasant assocaton to the nearest veternary clnc (mnutes) Average elevaton of the peasant assocaton of the household (meters above sea level) 14
Table 2. Summary statstcs Varable N Mean Standard Mn Max Error Dependent varables Whether sold lve anmals 727 0.448 0.031 0.00 1.00 Value of lve anmals sold 326 582.108 46.010 7.00 2800.00 Whether sold dary products 264 0.472 0.056 0.00 1.00 Value of dary products sold 132 140.803 19.241 3.90 1900.00 Whether sold lve anmals or dary products 738 0.495 0.031 0.00 1.00 Value of lve anmals and dary products sold 379 558.358 47.632 7.00 3150.00 Explanatory varables Dstance to nearest market 911 91.391 4.064 0.00 480.00 Dstance to dstrct town 911 35.465 1.467 0.00 168.00 Populaton densty 891 0.425 0.013 0.04 1.37 Educaton 911 1.809 0.108 0.00 19.50 Extenson 911 0.504 0.027 0.00 1.00 Oxen owned 911 1.366 0.060 0.00 7.00 Cows owned 911 0.838 0.061 0.00 12.00 Other cattle owned 911 0.461 0.035 0.00 5.50 Small rumnants owned 911 2.523 0.203 0.00 49.00 Poultry owned 911 2.375 0.184 0.00 60.00 Pack anmals owned 911 0.340 0.032 0.00 5.6.00 Landholdng 870 1.141 0.048 0.00 7.90 Crop ncome 911 2094.155 121.362 0.00 14901.33 Non-farm ncome 908 409.897 29.087 0.00 12800.00 Credt 911 0.630 0.025 0.00 1.00 Regon 911 0.179 0.005 0.00 1.00 Instruments Sex of household head 911 0.133 0.018 0.00 1.00 Age of household head 911 43.548 0.685 16.00 86.00 Household sze 911 6.315 0.145 1.00 16.00 Proporton of females 911 0.486 0.009 0.00 1.00 Dependency rato 911 0.537 0.010 0.00 1.00 Dstance to nearest all-weather road 910 183.940 8.385 0.00 1440.00 Dstance to veternary clnc 909 127.036 4.435 0.00 840.00 Alttude 911 2428.560 19.862 1500.00 3635.00 Notes: Durng the perod of the survey, US$1 EB8.5. Means and standard errors are adjusted for weghtng, stratfcaton and clusterng of sample. 15
Table 3. Econometrc results of lvestock and lvestock products market partcpaton by households (probt corrected for sample selecton) Whether sold lve anmals or dary Explanatory varable Whether sold lve anmals Whether sold dary products products Dstance to nearest market 0.0015 0.0015 0.0015 Dstance to dstrct town -0.0031-0.0015-0.0033 Populaton densty 0.3168 0.0021 0.2036 Sex of household head -0.1458 0.0568-0.2226 Age of household head -0.0014-0.0104 0.0000 Household sze -0.0015-0.0026-0.0091 Proporton of females 0.5977 0.7563 0.4933 Dependency rato 0.7998 * 0.8406 0.8750 * Educaton 0.0278 0.0205 0.0247 Extenson 0.2589-0.2313 0.2183 Oxen owned -0.1256-0.1733 *** -0.1541 Cows owned 0.1095-0.0133 0.1830 ** Other cattle owned -0.1079 0.1664 ** -0.0271 Small rumnants owned 0.0490 *** -0.0144 *** 0.0321 * Poultry owned -0.0266 0.0071-0.0198 Pack anmals owned -0.0682-0.0985 0.0056 Landholdng -0.3080 *** -0.2642 ** -0.2882 *** Crop ncome 0.0001 ** -0.0002 *** 0.0001 * Non-farm ncome 0.0002 * -0.0000 0.0003 *** Credt 0.0356-0.3268 * 0.0325 Regon 0.0031-0.8785 *** 0.0764 Inverse mlls rato assocated wth ownng -2.2711-3.558 ** -1.9957 ** lvestock or producng mlk Intercept -0.7353 2.2540 *** -0.6607 Ch-squared statstc 41.4900 *** 145.4400 *** 37.9000 *** Number of observatons 683 253 693 Notes: Detaled descrpton of dependent and explanatory varables s gven n Table 1. The dependent varables are dummy varables of whether or not household sold a partcular tem. Coeffcents are corrected for weghtng, stratfcaton and clusterng of sample. *, ** and *** ndcate statstcal sgnfcance at the 10, 5 and 1% levels, respectvely. 16
Table 4. Econometrc results of sales of lvestock and lvestock products by households (Heckman second-stage regresson) Value of lve anmals and dary Explanatory varable Value of lve anmals sold Value of dary products sold products sold Dstance to nearest market -0.4494 0.3545-0.4393 Dstance to dstrct town -1.0475 0.2769-0.9096 Populaton densty -357.0670 *** 264.1013 *** -254.5551 * Educaton 1.1815 13.3306 ** 4.8732 Extenson 29.9347 3.0234 63.1691 Oxen owned 0.5757 13.9895-2.4621 Cows owned -16.6416 30.4656 * -26.8594 Other cattle owned 125.4479 ** -36.2912 108.4980 * Small rumnants owned -3.2148-2.4801 4.4802 Poultry owned -13.2304 * -10.6378 *** -15.7609 ** Pack anmals owned 92.3673 * -3.1906 72.8187 * Landholdng -66.7317 80.2008 *** -53.7187 Crop ncome 0.0404 * 0.0061 0.0389 * Non-farm ncome 0.0606 0.0425 *** 0.0490 Credt 71.7862-29.2254 60.8650 Regon -173.6522 ** 98.5587 * -172.9850 ** Inverse mlls rato assocated wth partcpaton 374.6619 *** 160.6887 *** 404.6642 *** Intercept 730.3359 *** -148.1872 * 664.7062 *** Ch-squared statstc 56.6500 *** 34.0100 *** 45.7300 *** Number of observatons 299 126 353 Notes: Detaled descrpton of dependent and explanatory varables s gven n Table 1. The dependent varables are the values sold by the household of a partcular tem. Coeffcents are corrected for weghtng, stratfcaton and clusterng of sample. *, ** and *** ndcate statstcal sgnfcance at the 10, 5 and 1% levels, respectvely. 17
Appendx: Econometrc results of sales of lvestock ownershp and producton of dary products by households (probt regresson) Whether owned cattle or small Explanatory varable Whether owned cattle or small rumnants Whether produced dary products rumnants or produced dary products Dstance to nearest market -0.0025-0.0014-0.0022 Dstance to dstrct town 0.0008-0.0062 ** 0.0008 Populaton densty 0.4482-0.4504 0.3397 Sex of household head 0.0020-0.7133 0.0019 Age of household head -0.0085 0.0014-0.0105 Household sze 0.0233 0.0112 0.0326 Proporton of females -0.7930 * -0.3565-0.8921 ** Dependency rato 0.1696-0.4717 0.2284 Educaton 0.0526 0.0191 0.0778 ** Extenson -0.1030-0.1555-0.1620 Poultry owned 0.0481 * 0.0329 0.0578 ** Pack anmals owned 1.0336 0.1756 0.9432 *** Landholdng 0.5770 *** 0.3187 *** 0.6153 *** Crop ncome 0.0002 * 0.0001 0.0002 *** Non-farm ncome -0.0002 0.0001-0.0002 * Credt -0.1530-0.1529-0.2330 Regon 0.2738 0.8008 *** 0.2259 Dstance to nearest all-weather road -0.0005 0.0006 ** -0.0003 Dstance to veternary clnc 0.0009-0.0010 0.0005 Alttude 0.0003 * 0.0004 *** 0.0003 * Intercept -0.2579-1.5396 *** -0.1244 Ch-squared statstc 41.4900 145.4400 37.9000 Number of observatons 845 845 845 Notes: Detaled descrpton of explanatory varables s gven n Table 1. The dependent varables are the values sold by the household of a partcular tem. Coeffcents are corrected for weghtng, stratfcaton and clusterng of sample. *, ** and *** ndcate statstcal sgnfcance at the 10, 5 and 1% levels, respectvely. 18