Valuing the Relative Importance of Goat Farmers’ Trait Preferences: Implication for Designing Community Based Breeding Program

The success of a breed improvement programme depends on farmers' participation in the identification of breeding objective traits relevant to particular production environments. This study compared and contrasted two methodological approaches weighted rank vs. exploded logit modelin an effort to better understand and value the relative importance of farmers' trait preferences. For this purpose, two districts (Erer and Shinnile), were selected purposively based on their agro-ecologies, goat population size, the relative significance of goats to the livelihood of the communities, and the willingness of the farmers to participate. A survey with 120 farmers (60 from each district) owning goat flocks was carried out to investigate the purpose of keeping goats, breeding buck and doe selection criteria, and goat production constraints. The results found for the purpose of keeping goats, breeding buck and doe selection criteria, and goat production constraints reflected the multiple objectives of the farmers. Milk, income and meat were the most important traits for the purpose of keeping goats; body conformation, fast growth and pedigree were the most important traits for the selection of bucks; milk yield, body conformation and mothering ability were the most important traits for the selection of does; and drought, disease and feed were the most important traits of production constraints in the study areas. The results of this study form the basis for better designing of breeding programs for sustainable utilisation that address farmers’ preferences, speed up genetic progress and conserve indigenous goats, thereby meeting the needs of farmers.

Goats play a crucial role in the socioeconomic well-being of people in developing countries. They provide nutrition, income, and intangible benefits such as savings and insurance against emergencies. They also have cultural and ceremonial significance [1]. Furthermore, goats can efficiently utilize locally available feed resources and are well-suited for grazing on natural grasslands where crop cultivation is not feasible [2].

To meet the increasing demands of the human population, genetic improvement programs are needed to enhance and sustain goat production. However, such programs have not been very successful in developing countries [1,3]. This is partly due to the fact that many programs have been implemented without considering farmers' production goals, breeding preferences, and selection criteria [1,4,5]. For genetic improvement programs to be effective, they must be developed with a deep understanding of farmers' production objectives and trait preferences and must address constraints such as feeding, management, health control, and access to credit and marketing infrastructure [6,7].

Community-Based Breeding Programs (CBBPs) are new approaches to develop breeding programs in developing countries. These programs involve farmers in the decision-making process from the beginning and take into account their needs, visions, and active participation [8,9]. The success of CBBPs depends on properly considering breeding objectives, infrastructure, and farmer involvement [3,10]. Community-based breeding programs have recently been implemented for some breeds in different parts of Ethiopia [11,12], but there are still many other breeds and areas in the country where they need to be introduced.

Several studies have examined the production objectives, traits, and breed preferences of farmers for different livestock breeds in the tropics. For example, [13] used conjoint analysis to study trait preferences in Burkina Faso, while [14] used hedonic pricing to analyze breed preferences in Nigeria [15,16] assessed production objectives and selection criteria in the Gambia using a rating matrix. However, few studies have considered the diversity of farmers and the factors that influence their production goals and trait preferences [13,14,17,18]. This study aims to improve our understanding of farmers' trait preferences by comparing two methodological approaches: weighted rank and exploded logit. The hypothesis is that there is no heterogeneity among farmers' trait preferences.

Study area location

This study was carried out in two districts, namely Erer and Shinnile districts of Sitti Zone, Eastern Ethiopia, representing agro-pastoral and pastoral production systems. These districts were selected purposively based on their agro-ecologies, goat population size, the relative significance of goats to the livelihood of the communities, and the willingness of the farmers to participate. A brief description of the study areas is given below.

Shinnile: is located in the hot arid lowland agro-ecological zone at about 517 km East of Addis Ababa (Figure 1). The altitude of the district ranges from 500 to 1045 m.a.s.l with 9°40′N to 41°50′E. The mean annual rainfall ranges from 300-600 mm, falling mainly from July to mid-October). The rainfall pattern is very erratic and uneven, as a result of which, frequent droughts are common phenomena. Temperature ranges from a minimum of 15.1°C to a maximum of 35.0°C. The production system is a dominantly pastoral production system while some of the households practice an agro-pastoral production system.

Figure 1: Map of the study areas.

Erer: is located in the Sitti Zone of the Somali Region. Erer district is located in a semi-arid lowland agro-ecological zone at about 51 km east of Shinnile (Figure 1) [19]. The altitude ranges from 824 to 1107 m.a.s.l with 9°34′N to 41°23′E. The rainfall pattern is bi-modal (two rainy seasons) which is a small rainy season between mid-March to late May and the long rainy season from mid-July to mid-September. The aggregate average annual rainfall that the region gets ranges from 350 to 660 mm. Temperature ranges from a minimum of 14.7°C to a maximum of 34.1°C, respectively [20].

Sampling strategies and data collection

In each study district, randomly selected goat owners were approached and interviewed to collect data by interviewing. For this purpose, a total of 120 smallholder goat owners(60 from each district) were asked to rank the different levels of a trait (Table 1) in order of importance with the intention of assessing farmers’ trait preferences and the relative importance of the traits chosen by farmers. Ranks given were based on priority characteristics of a trait by farmers in selecting a trait as first, second and third.

Statistical analyses

For all statistical analyses in this study, SAS-programme version 9.4 [21] was used.

Descriptive statistics: Ranked data traits (i.e., purpose of keeping goats; selection criteria for breeding bucks and does; and production constraints of goats) were descriptively summarised and frequencies of the responses for the different rankings computed. Analyses and comparisons of ranked data on the collected traits were performed by two methodical approaches. In the first approach, indices were calculated for all ranked data to represent weighted averages of rankings as follows:

Rank-index = sum of [3 for rank 1 + 2 for rank 2 + 1 for rank 3] given for an individual reason (attribute) divided by the sum [3 for rank 1+ 2 for rank 2 + 1 for rank 3] for overall reasons (attributes).

In the second approach, ranked data were analysed to detect significant differences among the different levels of a trait using an exploded logit model, as described below.

Inferential Statistics: Ranked data were analysed using an exploded logit model [22]. The exploded logit model can be implemented using Proportional Hazards Regression (PHREG) &/or GLIMMIX procedures of SAS. In this model, a respondent i derives a certain utility Uij from each item (i.e., purpose of keeping goats; selection criteria for breeding bucks and does; and production constraints of goats) j, which is modelled as the sum of a systematic component μij and a random component eij, that is, Uij = μij + eij. The error term eij is assumed to be independently and identically distributed with an extreme value distribution. The systematic component can be modelled as: μij = βjxi, where βj is a row vector of coefficients, and xi is a column vector of explanatory variables that describe the ith respondent. Each of the βj vectors then describes how characteristics of the respondent affect the log-odds of choosing item j rather than the reference item. The SAS macro %MULT was used to display significant mean differences by a letter display [23].

Purpose of keeping goats

The weighted rank results for the purposes of keeping goats in Erer and Shinnile districts are presented in table 2a. According to the findings by [24], the experiences and knowledge for the purpose of keeping goats is primary and prerequisite for deriving operational breeding goals [25] have also reported that an accurate definition of breeding objectives is the most crucial step for designing breeding programmes as it determines the magnitude of the improvement directions.

In both Erer and Shinnile, the primary purpose for keeping goats was milk production, with index values of 0.41 and 0.43, respectively. This was followed by income generation, with index values of 0.28 and 0.29, respectively. Meat production was the third most important purpose, with index values of 0.19 in both districts.

Manure production had the lowest index value in Erer (0.03), while it had an index value of 0 in Shinnile. Saving had an index value of 0.07 in Erer and 0.05 in Shinnile, while wealth status had an index value of 0.02 in Erer and 0.04 in Shinnile.

These results indicate that milk production is the most important reason for keeping goats in both districts, followed by income generation and meat production. Manure production, saving, and wealth status were less important reasons for keeping goats.

The results found for both districts indicate that pastoralists had originally selected goats for their high milk production ability, income and meat and put more weight on these traits for breeding goats. This finding is in agreement with that reported by[26] who reported that milk production, cash income, and meat for home consumption were the primary reasons for keeping goats in Sitti Zone. According to the focus group discussion, goat milk is the common food item for most pastoralists at home and is used as fresh milk for childrens’ diets. In contrast to the present study, the report of [27] indicated that income generation followed by saving, manure, milk, and meat for home consumption was the primary purpose of eeping Hararghe highland goats. The difference can be explained by the type of production system where Hararghe highland goats are reared under mixed crop-livestock production while short-eared Somali goats are reared in pastoral and agro-pastoral production systems. In the study area, goats were also serving as a measure of wealth status and saving for emergency cases. Even though manure was ranked 5th important purpose of keeping goats in Erer goat owners, this was not mentioned by Shinnile goat keepers. This may be due to different production systems that existed between districts such that Erer goat owners were mostly agro-pastoralists in which manure was used as compost for farmers.

While the numerical differences in utilities shown in table 2a are evident, it is not possible to determine if these differences are statistically significant using the weighted rank approach, as it is descriptive rather than inferential. For example, in the Erer district, the relative importance of purposes for keeping goats, in decreasing order, was milk (0.41), income (0.28), meat (0.19), saving (0.07), manure (0.03), and wealth-status (0.02). However, pairwise comparisons, such as milk vs. income or milk vs. meat, are not accompanied by a significance letter, so it is not possible to determine if these differences are significant. This is where the exploded logit model has an advantage over the weighted rank approach. The inferential statistics result of the exploded logit model used is presented in table 2b. The table shows the likelihood ratio statistics and differences in estimates of Least Squares Means (LSM) for the different choices in both districts.

In both Erer and Shinnile, the sample rank effect and the choice effect were highly significant, with p-values less than 0.0001. This indicates that there were significant differences in the utilities of the different purposes for keeping goats.

From table 2b, it can be seen that in Erer, milk production had the highest LSM estimate (0.97a), followed by income (0.65b) and meat (0.55c). Saving had an LSM estimate of 0.34d, while manure and wealth status had the lowest LSM estimates (0.27e).

In Shinnile, milk production also had the highest LSM estimate (1.24a), followed by income (0.67b) and meat (0.52c). Saving and wealth status had LSM estimates of 0.32d and 0.31d, respectively, while manure had the lowest LSM estimate (0.25e).

Unlike the results presented in table 2a, the numerical differences in utilities shown in table 2b are accompanied by significance letters, which allow for the determination of significant differences between the utilities. This makes the exploded logit model a more preferable approach than the weighted rank, particularly when selecting or prioritizing between two attributes is in question.

Breeding buck and doe selection criteria

Breeding buck selection criteria: The weighted rank results for breeding buck selection criteria in the two districts are presented in table 3a. The table shows the rank and index values for each selection criterion.

In both districts, body conformation was the most important selection criterion, with index values of 0.33 and 0.32, respectively. This was followed by fast growth, with index values of 0.28 in both districts. Pedigree was the third most important selection criterion, with index values of 0.18 and 0.19, respectively. Coat colour had an index value of 0.11 in Erer and 0.10 in Shinnile. Libido had an index value of 0.05 in Erer and 0.06 in Shinnile, while polledness had an index value of 0.06 in Erer and 0.05 in Shinnile.

These results indicate that body conformation, fast growth, and pedigree were the most important selection criteria for breeding bucks in both districts; while coat colour, libido, and polledness were less important selection criteria. These findings indicated that goat owners in the study area are focused on bucks that had large size and fast growth for breeding purposes as well as for market supply.

As goat owners interviewed, goats with white plain and a combination of white with brown colours were perceived to be the most important colours for better adaptability, and tolerance to heat in harsh environments than black colour goats. While plain black and black-spot are unwanted coat colours in the study area as the market price for goats with black colour is very low and such goats are not preferred in markets. This finding is in general agreement with the finding of [28] for Aba’ala goats, who noted that white or light brown goats are better than black and dark-brown or black-grey goats at tolerating heat. Therefore, white plain coat colour is the most preferred coat colour, because such colours have the ability to reflect heat radiation to minimise increased body temperature and stress due to severe droughts than black colours. Selection criteria mentioned by goat owners were in agreement with the findings of [26] and [11], in which body size/conformation was identified as a first criterion to retain breeding bucks in the flocks for Sitti goats, and goats around Dire Dawa, respectively.

In the study area, physical traits like polledness in male goats were also preferred as a criterion for selecting male goats. Pastoralists prefer polled bucks for breeding and higher market prices due to their perceived qualities such as being more active breeders, having good body condition, higher dressing percentage for meat quality, tenderness, and higher market value compared to horned bucks [29]. Similar to this study[30] indicated that polled bucks are more preferred than horned bucks in the local markets and at Dire Dawa. Thus, it is important to include adaptation traits like coat colours and physical traits like polledness in the improvement programmes for good survival and performance of goats in the study areas.

The exploded logit model results for breeding buck selection criteria in Erer and Shinnile districts are presented in table 3b. The table shows the likelihood ratio statistics and differences in estimates of Least Squares Means (LSM) for the different choices in both districts.

In both Erer and Shinnile, the sample*rank effect and the choice effect were highly significant, with p-values less than 0.0001. This indicates that there were significant differences in the utilities of the different selection criteria for breeding bucks.

In Erer, body conformation had the highest LSM estimate (0.68a), followed by fast growth with LSM estimate of 0.57b. Pedigree had LSM estimate of 0.42c, while coat colour had LSM estimate of 0.37cd. Libido and polledness had the lowest LSM estimates (0.31d). In Shinnile district, body conformation and fast growth had the highest LSM estimates (0.63a and 0.60a, respectively), followed by pedigree with LSM estimate of 0.47b. Coat colour, libido, and polledness had lower LSM estimates (0.32c, 0.31c, and 0.29c, respectively).

These results indicate that body conformation and fast growth were the most important selection criteria for selecting breeding bucks in both districts, followed by pedigree. Coat colour, libido, and polledness were less important selection criteria.

Breeding doe selection criteria: The weighted rank results for breeding doe selection criteria in Erer and Shinnile districts are presented in table 4a. The table shows the rank and index values for each selection criterion in both districts.

In both Erer and Shinnile, milk yield was the most important selection criterion, with index values of 0.39 and 0.40, respectively. This was followed by body conformation, with index values of 0.23 in both districts. Mothering ability was the third most important selection criterion, with index values of 0.13 and 0.15, respectively. Kidding interval had an index value of 0.11 in Erer and 0.09 in Shinnile. Coat colour had an index value of 0.06 in both districts, while disease resistance had an index value of 0.08 in Erer and 0.07 in Shinnile.

These results indicate that milk yield, body conformation, and mothering ability were the most important selection criteria for breeding does in both districts; while kidding interval, coat colour, and disease resistance were less important selection criteria.

Although, goat owners stated different sets of breeding buck selection criteria, the top criteria for breeding doe selection were similar in both districts. This implies that farmers breeding objectives were mainly the breeding of animals to increase their milk production. Also, goat owners were interested in animals with higher genetic potential for milk production to complete their production objectives.

The survey result has also indicated that milk production was the main breeding goal of Short-eared Somali goat owners. Pastoralists believe that goat milk has medicinal value for children and contribute much more for household consumption while the sale of live animal to generate cash income was the next (second) breeding goal of goat keepers. This study is in agreement with the report of [30], in which goat milk has medicinal value for children and contributes as a diet for household consumption around Dire Dawa. [31] had also noted as goats browse different variety of trees and shrubs, goat owners believed that goat milk has medicinal value and contributes much more to the well-being of infants. Therefore, the reasons behind assigning relatively high values for milk traits in the present study could be due to the significant role of goats in sustaining the livelihood of farmers through milk production. This implies that there was an agreement between the purpose of keeping goats and the selection criteria of farmers for traits of interest.

In the study area, disease resistance as an adaptation trait was ranked 5th important trait preferred as compared to the other adaptational traits like heat tolerance, drought-tolerance, and walking ability. The reason behind assigning high values for disease resistance traits rather than other adaptational traits is that the prevalence of diseases is one of the major constraints of goat production in the study area. Goat owners frequently suffered from the loss of many animals and also encountered high costs for drugs to treat their animals. Due to this goat owners are interested in those animals that had better performance and animals that have tolerance or resistance to any diseases. The current result was similar to that reported by [26] and [32].

[11] and [33] have also noticed that a higher preference for milk yield, body size/conformation, and reproduction traits for Abergelle goats, and goats around Dire Dawa, respectively. Thus, devising good mechanisms like the practice of retaining best does and culling older, less productive and disease susceptible does from the flock is one of the major intervention programs that need to be implemented continuously to achieve desired goals of genetic improvement.

The exploded logit model results for goat production constraints in the two districts are presented in table 5b. The table shows the likelihood ratio statistics and differences in estimates of Least Squares Means (LSM) for the different choices in both districts.

For both Erer and Shinnile districts, the exploded logit model results indicate that ‘breeding doe selection criteria’ have utility order in decreasing order of milk yield > body-conformation > mothering-ability = kidding-interval = coat-colour = disease-resistance. In contrast to the results found in table 4a, in table 4b, the numerical differences for the different utilities are not only clear to see, but also accompanied by corresponding significant letters for separating the differences. This makes the exploded logit modelling approach to be preferred than the weighted rank.

Goat production constraints

The weighted rank results for goat production constraints in the two districts are presented in table 5a. The table shows the rank and index values for each production constraint in both districts.

Participatory identification of the major constraints of goat production is the first step to design the appropriate genetic improvement programmes. In both districts, drought was the most important production constraint, with index values of 0.35 and 0.34, respectively. This was followed by disease, with index values of 0.29 and 0.26, respectively. Feed availability was the third most important production constraint, with index values of 0.21 and 0.22, respectively. Water availability had an index value of 0.07 in Erer and 0.11 in Shinnile. Predator attacks had an index value of 0.05 in both districts, while market access had the lowest index value (0.03 in Erer and 0.01 in Shinnile).

These results indicate that drought, disease, and feed availability were the most important production constraints for goat farming in both districts; while water availability, predator attacks, and market access were less important production constraints.

The results of goat production constraints reported in this study are in line with that reported by [27], who reported drought, disease, and feed problems in Ziquala and Tanqua-Abergelle districts. Pastoralists in the study area do not mention water shortage as the main constraint. This is different from [30], who indicated that water shortage was the second-mentioned constraint in Jaldessa and Mudianeno rural villages of Dire Dawa.

As key informant groups mentioned, the flock size of their goats has been decreased in the last ten years mainly due to indices of the severe droughts that caused feed shortage and low productivity of rangeland due to bush encroachment. Predator and lack of markets were also goat production constraints that goat owners mentioned during the survey time. Therefore, establishing strong animal health services and improving the rangeland management to overcome recurrent droughts that lead to feed shortage by resorting and increasing the productivity of degraded natural browsing area need to be considered for setting-up of the appropriate breeding program in the study areas.

The exploded logit model results for goat production constraints in the two districts are presented in table 5b. The table shows the likelihood ratio statistics and differences in estimates of Least Squares Means (LSM) for the different choices in both districts.

In both districts, the sample*rank effect and the choice effect were highly significant, with p-values less than 0.0001. This indicates that there were significant differences in the utilities of the different production constraints for goat farming.

In Erer, drought had the highest LSM estimate (0.74a), followed by disease with LSM estimate of 0.62b. Feed availability had LSM estimate of 0.49c, while market access, predator attacks, and water availability had lower LSM estimates (0.29d, 0.28d, and 0.33d, respectively). In Shinnile, drought also had the highest LSM estimate (0.74a), followed by disease with LSM estimate of 0.68ab. Feed availability had LSM estimate of 0.58bc, while water availability had LSM estimate of 0.47cd. Predator attacks and market access had the lowest LSM estimates (0.35de and 0.28e, respectively).

These results indicate that drought and disease were the most important production constraints for goat farming in both districts, followed by feed availability; while water availability, predator attacks, and market access were less important production constraints.

In contrast to the results found in table 5a, the numerical differences in utilities shown in table 5b are accompanied by significance letters, which allow for the determination of significant differences between the utilities. This makes the exploded logit model a more preferable approach than the weighted rank.

The breeding decisions of smallholder goat producers in Erer and Shinnile conform to producers’ multiple objectives. These factors include why people keep goats, how they choose breeding bucks and does, and the challenges they face in goat production. Understanding these breeding choices is crucial for creating livestock policies that can improve the lives of small-scale farmers and meet the needs of consumers who use goat products. Therefore, any breed improvement strategies that are intended to be implemented in the study area and elsewhere should consider the traditional breeding practices and breeding objectives of the community.

Unlike the findings for the weighted rank results, the exploded logit model was better at helping make decisions about selecting traits. The differences in utility values were supported by significant indicators, indicating clear distinctions between them. With respect to this, the present study has shown that heterogeneity exists among farmers’ trait preferences. Thus, in designing goat breeding programme the methodology used by exploded logit model is a better choice to use and of immense value especially when it comes to selection few traits.

The willingness of the farmers in the study area to participate in the study is gratefully acknowledged.

1. Kosgey IS, Baker RL, Udo HMJ, van Arendonk JAM. Successes and failures of small ruminant breeding programs in the tropics: A review. Small Rumin Res. 2006;61:13-28. doi:10.1016/j.smallrumres.2005.01.003.
2. Baker RL, GD Gray. Appropriate breeds and breeding schemes for sheep and goats in the tropics. In: Sani RA, Gray GD, Baker RL, editors. Worm Control for Small Ruminants in Tropical Asia. Monograph. Australian Centre for International Agricultural Research (ACIAR); 2004. p.63-95.
3. Sölkner J, Grausgruber H, Okeyo AM, Ruckenbauer P, Wurzinger M. Breeding objectives and the relative importance of traits in plant and animal breeding: A comparative review. Euphytica. 2008;161:273-282. doi: 10.1007/s10681-007-9507-2.
4. Ndumu DB, Baumung R, Wurzinger M,  Drucker AG, Okeyo AM,  Semambo D, Sölkner J. Performance and fitness traits versus phenotypic appearance in the African Ankole Longhorn cattle: A novel approach to identify selection criteria for indigenous breeds. Livestock Science. 2008;113:234-242. doi: 10.1016/j.livsci.2007.04.004.
5. Haile A, Wurzinger M, Mueller J, Mirkena T, Duguma G, Mwai O. Guidelines for setting up community-based sheep breeding programs in Ethiopia. ICARDA tools and guidelines. International Center for Agricultural Research in the Dry Areas. 2011.
6. Anderson S. Animal genetic resources and sustainable livelihoods. Ecological economics. 2003;45(3):331-339. doi: 10.1016/S0921-8009(03)00088-0.
7. Mekonnen H, Tadesse G. Goat production and marketing systems in Ethiopia: A review. Global Journal of Animal Scientific Research. 2016;4(2):277-283.
8. Gowane GR, Kumar D. Community-based sheep breeding programs for genetic improvement in developing countries: A review. Indian Journal of Animal Sciences. 2016;86(5):505-509.
9. Tadesse E, Getachew T. Community-based sheep breeding programs in developing countries: A review. World Journal of Agricultural Research. 2019;7(3)83-87.
10. Wurzinger M, Sölkner J, Iñiguez L. Important aspects and limitations in considering community-based breeding programs for low-input smallholder livestock systems. Small Ruminant Research. 2011;98:170-175. doi: 10.1016/j.smallrumres.2011.03.035.
11. Gebreyesus G, Aynalem H, Tadelle D. Breeding scheme based on community-based participatory analysis of local breeding practices, objectives and constraints for goats around Dire Dawa, Ethiopia. Livestock Research for Rural Development. 2013;25(3).
12. Haile A, Mirkena T, Duguma G, Wurzinger M, Rischkowsky B, Tibbo M, Mwai AO, Sölkner J.  Community-based sheep breeding programs: Tapping into indigenous knowledge. Livestock Research for Rural Development. 2013;25(12). doi: 20.500.11766/7250.
13. Tano K, Mamuanga MD, Faminow, Swallow B. Using co-joint analysis to estimate farmers’ preferences for cattle traits in West Africa. Ecological Economics.  2003;45:393-407. doi: 10.1016/S0921-8009(03)00093-4.
14. Jabbar MA, Diedhiou ML. Does breed matter to cattle farmers and buyers? In: Evidence from West Africa. Ecological Economics. 2003;45(3): p.461-472. doi: 10.1016/S0921-8009(03)00097-1.
15. Ejlertsen M, Poole J, Marshall K. Traditional breeding objectives and practices of goat, sheep and cattle smallholders in The Gambia and implications in relation to the design of breeding interventions. Trop Anim Health Prod. 2013 Jan;45(1):219-29. doi: 10.1007/s11250-012-0194-1. Epub 2012 Jun 17. PMID: 22706889.
16. Bennison JJ, Barton D, Jaitner J. The production objectives and feeding strategies of ruminant livestock owners in The Gambia: implications for policy makers. Agricultural Systems, 1997;55(3):425-444. doi:10.1016/S0308-521X(97)00002-4.
17. Megersa B. Markemann A, Angassa A, Zarate AV. The role of livestock diversification in ensuring household food security under a changing climate in Borana. Food Security. 2014;6(1):15-28. doi: 10.1007/s12571-013-0314-4.
18. Traoré SA, Markemann A, Reiber C, Piepho HP, Valle Zárate A. Production objectives, trait and breed preferences of farmers keeping N'Dama, Fulani Zebu and crossbred cattle and implications for breeding programs. Animal. 2017 Apr;11(4):687-695. doi: 10.1017/S1751731116002196. Epub 2016 Oct 27. PMID: 27786142.
19. Assefa E , Adugna T. Analysis of small ruminant production and marketing systems in Erer and Shinile districts of Ethiopia. World Journal of Agricultural Research. 2016;4(3):73-78.
20. Mohammed A, Alemu T. Assessment of the status of agro-pastoral systems in Erer and Shinile districts of Somali region, eastern Ethiopia. Journal of Agricultural Extension and Rural Development. 2018;10(3):43-47.
21. SAS (Statistical Analysis System). SAS for windows, Release 9.4 SAS enterprises, Inc., Cary, NC, USA. 2016.
22. Allison PD, Christakis NA. Logit models for sets of ranked items. Sociological Methodology 1994;24:199-228.
23. Piepho HP. A SAS macro for generating letter displays of pairwise mean comparisons. Communications in Biometry and Crop Science. 2012;7(1):4-13.
24. Jaitner J, Sowe J, Secka-Njie E, Dempfle L. Ownership pattern and management practices of small ruminants in The Gambia - implications for a breeding programme. Small Rumin Res. 2001 May;40(2):101-108. doi: 10.1016/s0921-4488(00)00221-2. PMID: 11295391.
25. Nandolo W, Wurzinger M, Mészáros G, Vantassell C, Sölkner J. Identification of breeding objectives in community-based goat breeding programmes in Malawi. Acta agriculture Slovenica. 2016;5:104.
26. Hulunim G, Halima H, Kefelegn K, Aynalem H, Lobo RNB, Yetayew A, Rischkowsky B. Husbandry practices and phenotypic characteristics of indigenous goat populations in Ethiopia. African J. Agric. Res. 2017;12:2729-2741.
27. Tatek W. Optimizing community-based breeding for indigenous goat breeds in Ethioia. Faculty of Agricultural Sciences at the University of Hohenheim. 2016.
28. Gebre KT, Yfter KA, Teweldemedhn TG, Gebremariam T. Participatory definition of trait preferences and breeding practices of goats in Aba’ala, Afar region: as input for designing genetic improvement program. Tropical Animal Health and Production. 2019. doi:10.1007/s11250-019-01968-1.
29. Abegaz S, Haile A. Characteristics and strategies for improvement. Goat production and marketing systems in Ethiopia. International Journal of Livestock Production. 2014;5(10):150-162.
30. Gebreyesus G. Community-based participatory characterization of the short eared Somali goat population around Dire Dawa. MSc thesis paper presented to Haramaya University. Ethiopia. 2010;219.
31. Tesfaye G, Haile A, Markos T, Solkner J, Wurzinger M. Herd management and breeding practices of sheep owners in a mixed crop-livestock and a pastoral system of Ethiopia. Journal of Agricultural Sciences. 2010;5(8):685-691.
32. Ahmed U. Definition of breeding objective traits for short-eared Somali goat breed in Sitti zone, eastern Ethiopia. MSC Thesis. 2021.
33. Solomon G, Tesfaye G, Zergaw E, Tadelle D, Gemeda D, Markos T, Barbara R, Okeyo M, Maria W,  Johann S, Aynalem H. A basis for designing community-based breeding programs. ICARDA working paper. Syria. 2013;47.