Analysis of Live Goats Market Chain: The Case of Pastoralists in Moyale District, Ethiopia

Goat production and marketing provide vast socio-economic benefits. However, the subsector has encountered with lack of information about what determines goat market participation, supply, and outlet choice. Therefore, the study was intended to analyze the live goat market chain in the Moyale district. Both primary and secondary data were used. Primary data were collected from a randomly selected sample of 196 goat producers and were analyzed using econometric methods. To investigate the determinants of household market participation and supply, Heckman's two-stage procedure was used. Heckman's 1 stage result showed that different variables have affected live goats' market participation decisions either positively and negatively. Heckman's 2 stage result indicated that the number of live goats supplied to the market is also affected either positively or negatively by different variables. To identify factors affecting outlet choice multivariate probit model was used. Multivariate probit model results showed that distance to market and selling price significantly affected the entire outlets (namely; large scale traders, small-scale traders, cooperatives, collectors, and consumers) either positively or negatively. The study recommended the need to develop market infrastructures, improve the capacity of productive labor, promote credit institutions, and provision of awareness creation training towards goat trading business.


Data Types, Sources and Methods of Collection
The study employed both primary and secondary data. Secondary data were gathered from different records of district office of agriculture, CSA, web sites and other published and unpublished relevant sources. Primary data were gathered using structured questionnaire. The questionnaire were pre-tested and amended based on feedbacks from the pre-test. Personal observations and group discussions with community members were also done for triangulation. Trained enumerators were hired to gather data under the supervision of the principal researcher.

Sampling Technique and Sample Size
Goats producing kebeles and sample households were selected by utilizing a two-stage sampling procedure. From 19 kebeles in the district 10 goat producing kebeles were purposively selected.
At the first stage, three representative kebeles namely Kabanawa, Mado Migo and Chamuk were selected randomly from selected 10 goat producing kebeles. In the second stage, from three randomly selected kebeles, 196 sample households head were selected randomly based on the probability proportional to size. Sample size of goat producers were determined by using Cochran (1977) formula which is developed to calculate a representative sample for proportions by considering the 95% confidence level and 7% level of precision. The formula is: = / (1) Where, ′ ′ is the sample size of goat producer households ′ ′ is the selected critical value of desired confidence level ( = 1.96) and e is the desired level of precision( = 0.07), p is the expected proportion of market participants from population of the goat producers ( = 0.5)and = − 1 = 0.5. Hence, the sample size is: = 1.96 × 0.5(0.5) 0.07 = 196 The distribution of goat producers and sample households are given in the table 1.

Methods of Data Analysis 2.4.1. Determinants of household's market participation and supply of live goats
Heckman two-steps choice model was found to be applicable to investigate factors touching goat market participation decision and factors touching number of goats brought to the market at the same time since it controls selectivity bias. Specification of the Heckman two-step technique is given in terms of the probability of goat market participation (GMP), and goat market supply (GMS) is: The participation Equation/the binary probit equation The latent regressed variable that is not observed.
: Factors that are assumed to have effect on the chance of sampled household market participation : Vector of unknown parameter in participation equation : Residuals that are independently and normally distributed with zero mean and constant variance The observation equation/the supply equation observed. The error terms, and , are assumed to be bivariate, normally distributed with correlation coefficient. and are the parameter vectors. By employing ordinary least squares method , is regressed on the independent variables , and the vector of inverse Mills ratios (' ) from the selection equation. Where: : The observed regressand variable : Factors expected to affect number of goat supplied : Vector of unidentified parameter in the supply equation : Residuals in the supply equation that are independently and normally distributed with zero mean and constant variance. ' = (()*) +,-()*) (4) .( ) is density function and 1 − /( ) is distribution function

Goats' market outlets and factors affecting outlets choice
Multivariate probit model was found to be appropriate to analyze factors affecting goat producers' market outlets choice decisions since it capture the influence of the set of explanatory variables on each of the five outlet choices in the study area, while error terms are allowed to be freely correlated.
The observed outcome of choosing market outlets of live goats can be modeled following random utility formulation for individual choice (Greene, 2012). Assume the i th live goat supplier household (i=1, 2…... N), encountering with a problem to decide on whether to choose existing market outlets or not. Let U0 represent the benefits to the goat producing households who chooses large scale traders, and let Uk, be the advantage of goat producing household to select the K th , live goat market outlet: As K symbolizes choice of large scale traders (Y1), small scale traders (Y2), cooperatives (Y3) collectors (Y4) and consumers (Y5). If Y *ik = U *k −U0 > 0, the goat supplier household decides to select the K th market outlets. The latent variable is the net earnings (Y *ik) which the goat supplier household gains from choosing a market outlets is determined by witnessed independent factors (Xi) and the error terms (εi): * 1 = 2 1 3 (4 = , , 5 , 6 , 7 ) (5) Using the indicator function, the unobserved preferences in equation (9) translates into the observed binary outcome equation for each choice as follows: 1 = 8 1 9. 1 * > 0 0 :;ℎ =>9? (@ = , , 5 , 6 , 7 ) In multivariate model, where the choice of several market outlets is possible, the error terms jointly follow a multivariate normal distribution (MVN) with zero conditional mean and variance normalized to unity (for identification of the parameters) where (µy1, µy2, µy3, µy4) ~ MVN (0,Ω) and the symmetric covariance matrix Ω is given by:- Off-diagonal elements crossing the covariance matrix are of particular interest, it denotes unnoticed association among the stochastic components of various types of outlets. This supposition is that equation (11) makes a MVP model which jointly signifies decision to select specific goat market outlets. The specification using off-diagonal elements which are non-zero, permits for association across error terms of numerous latent equations, that denotes unobserved features that influence the choice of different live goat market outlets. Following the form applied by Cappellarri and Jenkins (2003), the log-likelihood function correlated with a sample outcome is provided by: Where ωi is an optional weight for observation i, and Φ is the multivariate standard normal distribution with arguments µi and Ω, where µi can be denoted as:-O = (4 2 , 4 2 , 4 5 5 2 5 ), While Ω 1 = 1 .:= P = 4 Q R (9) Ω S1 = Ω 1S = 4 S 4 1 ρ S1 .:= P ≠ 4, 4 = 1,2,3 … . >9;ℎ 4 1 = 2X 1 − 1 (10)

Determinants of Live Goats Market Participation and Supply 3.1.1. Determinants of Live Goats Market Participation Decision (Hechman 1 st step)
Among hypothesized twelve explanatory variables, five of them were identified to significantly influenced goat market participation decision. Among hypothesized independent variables family size, credit use and lagged price affected the goat market participation decision positively and significantly. Whereas distance to the nearest livestock market and other income from off/non-goat activities reduces live goat market participation decision (Table 2). Family size (manequi): This variable influenced the live goat market participation decision positively and Journal of Natural Sciences Research www.iiste.org ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online) Vol.12, No.11, 2021 significantly at 5% level of significance. As households labor force increase by one man equivalent, the probability of producer household participation decision in live goat market increases by 4.6%. The reason behind might be that those households who have higher family labor force can manage to drive the goat flock to the market easily and improves level of participation. The result is similar to the findings of Gezehagn (2015) who found that household size positively and significantly affected beef cattle market participation decision. Distance to market (dismkt): This variable influenced household's goat market participation decision negatively and significantly at 5% level of significance. As distance from the closest livestock market increase by one km the likelihood of the household to participate in goat market reduced by 2.09%. This might be due to the far distance the household resides from the nearest livestock market; the less likely the household involved in selling live goat due to long trekking time and relatively high marketing costs. This is associated with the findings of Gebremedhin et al. (2015) and Zemeda (2016) who found that distance to nearest livestock market negatively and significantly influenced small ruminants' market participation decision. Credit use (credit): Credit use was found to determine the probability of participation in goat markets significantly and positively at 1% probability level. As compared to credit non-users, those households with credit use participate in live goat markets by 8.7% more. The reason behind this might be that credit helps household in covering marketing costs and pave way for better market participation. This is in line with Gezehagn (2015) who found that credit access in beef cattle production rises the likelihood of producers' participation in the markets. 768 Note: Dependent variable is mktprt. ***, ** and *are statistically significant at 1%, 5% and 10% significance levels respectively. Source: Own computation from survey result, 2017 Lagged price (lgprice): The seasonal average lagged price of goat in the previous year was found to determine the probability of participation in live goat markets positively and significantly at a 5% level of significance. As lagged price for goat raises by one birr the likelihood of live goat market participation rises by 0.003%. The intuition behind this might be that the relatively high market price in the previous year may reassure the households to participate in the live goat market. The study by Shambel (2013) also reported the same result. Other incomes excluding income from goat (othrinc): This variable influenced goat market participation decision negatively and significantly at 1% significance level. The study indicated that involvement in other activities excluding goat marketing decreases the probability of participation of producers in goat marketing by 0.02%. This is mainly due to the fact that, households participating in other non-goat activities are gaining the income from on/off farm activities rather than goat market, this makes the household to abstain from goat market participation and decreases pastoralists goat market participation. This result is in line with Jemal (2017) who reported the same results.

Determinants of Number of Live Goats Market Supply (Heckman 2 nd step)
The Heckman second step estimation procedures is used to identify determinants of number of goats supplied to the market and it includes those variables used in Heckman first step procedure except one or more exclusion restriction variables. This study used lagged price as selection variable in participation equation which was found to affect live goat market participation decision but has no significant effect on number of goat supplied to the market in order to predict inverse of the Mills' ratio correctly. The coefficient of Mills ratio (Lambda) in the Heckman second-step estimation is significant at 5% probability level. This indicates sample selection bias, which is the existence of some unobservable household characteristics determining probability of participation in live goat market and thereby affecting the total number of live goats brought to the market. The Heckman selection model's overall goodness of fit was indicated by the wald chi2 (11) = 352.11 which is significant at 1% level of significance. This shows that the participation level can be explained by the independent variables jointly included in the selection model. Therefore, Heckman two step sample selection model was relevant model for this study.  (Table 3) Among the hypothesized variables, five of them influenced number of goat supplied to the market significantly. Sex of household head, number of goat owned and credit use affected the live goat supplied to market positively and significantly whereas distance to the nearest livestock market and other livestock owned negatively and significantly affected the level of live goat sales. Sex of household head: As hypothesized this variable influenced the live goat market supply positively and significantly at 1% level of significance. Holding other explanatory variables constant, as compared to female headed household, male headed household's goat market supply increases by 0.54TLU of goats. The reason behind might be that those households who are female headed household are busy with house work and as well in taking care of their children. This result is in line with Habtamu (2015) who found that being male headed household had positively and significantly affected the level of supply. Distance to market (dstmkt): This variable influenced number of goat market supplied negatively and significantly at 10% significance level. Holding other explanatory variables constant, the result shows that as distance away from the nearest livestock market increased by one kilometer, the level of goats supplied to market decreased by 0.06TLU of goat. This might be due to the fact that the far the market place the higher would be the trekking time, marketing costs, less access to market information and facilities. This is in line with Zemeda (2016) who indicated that distance to market caused the quantity of small ruminants supplied to market decrease in Tahtay adyabo district. Number of goats owned (goatown): As expected, this variable influenced households' number of goats supplied to market positively and significantly at 1% significance level. As number of goat owned increase by one TLU the number of goat supplied to the market increases by 0.16TLU. This is mainly due to the fact that pastoralists with more number of goats tend to favor the supply of goat and generate cash from goat sale. This result agrees with the finding of Zekerias (2017) who found that number of cattle owned by pastoralists increase the number of live cattle supplied to market. Livestock owned other than goats (othrtlu): This variable determined number of goats supplied to market negatively and significantly at 1% significant level. As other livestock owned increased by one TLU the number of goat supplied to market reduced by 0.07TLU. The intuition behind this might be that those households with more number of livestock other than goat tend to mainly disregard the benefits to be generated from supplying goat to the market. Shambel (2013) also reported the same result. Holding other explanatory variables constant, as compared to credit non-users, the level of goat supplied for those households with credit use increased by 0.41TLU. This suggests that credit use would enhance the financial capacity of the farmer to cover possible marketing costs to supply more goats to market. This is in line with Gebremedhin et al. (2015) who found that credit use increases the number of small ruminants supplied to the market. Inverse mills ratio (LAMBDA): According to the model output, the inverse Mill's Ratio or selectivity bias correction factor affected the number of goat supplied positively and significantly at 5% significance level and this result suggests that there appears to be unobserved factors that might affect both probability of households goat Journal of Natural Sciences Research www.iiste.org ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online) Vol.12, No.11, 2021 market participation decision and number of goat supplied, justifying the suitability of the Heckman two stage model for identifying the determinants of number of goats supplied to market.

Live Goats Market Outlets and Factors Affecting Their Choices
Goat producers in the study area have five major market alternatives to sell their goat. Multivariate probit model was used to analyze the determinants of households decision on market outlet choices as households are more likely chooses different outlets simultaneously. The model estimated jointly for five categorical dependent variables namely; (1) large-scale traders, (2) small-scale traders, (3) cooperatives, (4) collectors and (5)  The correlation between outlets as shown in the likelihood ratio test indicated that there are negative and significant interdependency between large-scale traders and cooperatives outlets, large-scale traders and collectors outlets, and large-scale traders and ultimate consumers' outlets. On the other hand small-scale traders' outlet had a positive and significant relationship with cooperatives, collectors and ultimate consumers market outlets, and cooperatives outlet had positive and significant relationships with both collectors and ultimate consumers' outlets. While there were positive and significant interdependency between collectors and ultimate consumers outlets. The model result also shows that the probability that goat producers choose large-scale traders, small-scale traders, cooperatives, collectors and consumers market outlets were 35.36%, 58.63%, 57.68%, 62.48% and 34.12%, respectively. The joint probability of choosing all market outlets was 3.04% and the joint probability of failure to choose all market outlets was 0.05%. The multivariate probit model analysis indicated that out of ten explanatory variables included in the model two variables significantly affected the choice of entire outlets at different magnitude and probability level. Distance to market (dstmkt): Distance to livestock market negatively and significantly affected large scale traders, small-scale traders and consumers' market outlets choice at 5%, 5% and 10% level of significance respectively. While it is positively and significantly affect cooperatives and collectors outlets at 5% and 10% significance level. This result showed that as the distance from the livestock market increases the households tend to sell their goat at farm gate and use outlets other than large and small-scale traders' market outlets. This is due to high trekking and other marketing costs to supply goat to the market. This is coincided with Nuri (2016) who found that increase in distance to the market center would decrease the probability of choosing wholesalers outlet, but increase the likelihood of choosing collectors outlet. The result is again supported by the findings of Addisu (2016) and Bezabih et al. (2015) who reported that distance to the nearest market positively influenced the likelihood of choosing retailer outlet.
Journal of Natural Sciences Research www.iiste.org ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online) Vol.12, No.11, 2021  Note: ***, ** and *are statistically significant at 1%, 5% and 10% significance levels respectively. Source: Own computation from survey result, 2017 Selling price (price): Selling price of goat positively and significantly affected the large-scale traders, small-scale traders and consumer market outlets choices at 1%, 5% and 1% level of significance respectively. In contrary, the variable affected the selection of cooperatives and collectors market outlets negatively and significantly at 5% and 1% level of significance respectively. The findings showed that the goat suppliers receive a better selling price from large-scale traders, small-scale traders and consumers as compared to collectors and cooperatives. Therefore households prefer to choose large-scale traders, small-scale traders, and consumer outlets than choosing cooperatives and collector market outlets as they pay less. This finding is related with Melkamu (2016) who reported that selling price significantly and positively affected wholesalers and consumer outlets and negatively affected collector outlet choice. Sultan (2017) found that price offered by market outlets affected cooperative market outlet negatively and significantly.

CONCLUSION AND RECOMMENDATIONS 4.1. Conclusion
Ethiopia is country with the largest small-ruminants population in Africa. This sub-sector plays significant role in improving income and livelihood of large portion of rural households in the district in terms of nutrition, income and intangible benefits (savings, insurance against emergencies, cultural and ceremonial purposes). However, the subsector faces various marketing challenges in the district. The study was aimed at analyzing goat market chain in Moyale district, Borena zone, Oromia region, Ethiopia. The study objectives are: (1) to analyze the determinants of goat market participation decision and level of supply; and (2) to analyze live goats market outlets and factors affecting their choices.
To meet the objectives of the study primary data were collected from 196 goat producers using pre-tested structured questionnaires and secondary data from different district offices, CSA, published and non-published sources. Econometric models (namely; Heckman two stages and multivariate probit model) were used to analyze the data collected from sample goat producer households. Heckman two stages model was used to identify the determinants of participation decision and level of participations. The result showed that; family size, credit use and lagged price were positively and significantly determined participation decision, while distance to market and other income negatively and significantly determined participation decision. Moreover the result indicated that the number of goat supplied was affected positively and significantly by sex of household head, number of goat owned and credit use. While on the other hand it is negatively and significantly affected by distance to market, other livestock owned and other income. The Multivariate probit model (MVP) was employed to identify factors affecting the decision of goat suppliers' to choose among alternatives market outlets. The model result showed that distance to livestock market negatively and significantly affected large scale traders, small-scale traders and consumer market outlets choice, while it affected cooperatives and collectors market outlets positively. Selling price of goat positively and significantly affected the small-scale traders', large-scale traders' and ultimate consumers' market outlets choices, while the variable affected cooperatives' and collectors' market outlets choice negatively and significantly.
In general, the study concludes that demographic, socio-economic and institutional factors play a vital role in determining live goat market participation decision, number of goats supplied and producers' market outlet choice decision. However, the general situation of goats' market in the district needs to be improved. Hence, efforts ought to be made to enhance the goat market participation, improve the number and quality of goats supplied to the market as well as market outlet choice decision in the study area.

Recommendations
Depending up on the findings of the study the succeeding recommendations have been made to be considered by responsible bodies in promoting goat marketing in the study area for the betterment of generation to come.
Distance to the nearest livestock market significantly and negatively affected the participation decision and number of goats market supply. Therefore, developing market infrastructure such as building market places, repairing roads and improving road networks to production areas reduce trekking time, minimize transportation and other marketing costs which in turn improve goat market participation and increase goat supply to the market. The development of transport infrastructures also enables households (particularly aged households) to supply their animal to better offering outlet rather than being exploited by nearby collectors for whom distance to market is an asset.
Family size (active labor force) positively and significantly influenced participation of households to the market. The variable again affected the choice of large and small-scale traders' market outlets positively and significantly. It is important for raring and also trekking the animals to the market. Lack of such labor force has a great impact on households' decision to participate and supply their animal to a better offering market outlets. Therefore, improving the capacity of available productive labor of the households is important to increase goat production, market participation and supply to the well-paying market outlets such as large-scale traders.
Credit use by producers positively and significantly affected goat market participation and number of goat supplied. Therefore, increasing and facilitating more financial institutions are critical in helping goat producer households' capacity to cover the marketing costs it takes to participate in the market and in turn supply more to the market. Therefore Credit providing institutions need to attract goat producers to the institution by linking them with the main urban and nearby supply of the service providing institutions.
Other livestock holding have negative and significant effect on the number of goat supplied to the market and the other income generated form non-goat business activity was negatively and significantly affected households decision to participate on goat market. This is mainly due to an inadequate recognition of the contributions goats make to the livelihoods of the poor. Therefore, market oriented training focused on awareness creation about the advantage of having alternative income sources and the real benefit to be obtained from goat marketing should be given.
Number of goats owned positively and significantly determines number of goats supplied to the market. Similarly it affected the choice of large-scale traders' market outlet positively and significantly and in contrary, it negatively and significantly affected the choice of collectors and ultimate consumer market outlets. Therefore, the concerned body needs to focus on increasing goat production which could be achieved through providing improved feed and other agricultural technologies. This in turn will enhance the choice of large-scale traders' outlet which buys large number of animal at once.
One year lagged price positively and significantly influenced goat market participation decision of households. High price in the previous production season results in improvement in the goat market participation by goat producer households. Therefore, information about previous year price can improve market participation decision of producer households.