Deutscher Tropentag 2004
Berlin, October 5-7, 2004
Conference on International Agricultural Research for Development

Assessment of Farmer Preferences for Cattle Traits in Smallholder Cattle Production
Systems of Kenya and Ethiopia
Emily Ouma1,2∗ , Awudu Abdulai1, Adam Drucker3 and Gideon Obare4
1

Department of Food Economics and Consumption Studies, University of Kiel, Olshausenstrasse 40, 24098 Kiel,
Germany. Email: aabdula@food-econ.uni-kiel.de
2
Swiss Centre for International Agriculture (ZIL), Scheuchzerstrasse 7, CH-8092, Zürich, Switzerland
3
International Livestock Research Institute, P.O. Box 5689, Addis Ababa, Ethiopia. Email: a.drucker@cgiar.org
4
Department of Agricultural Economics, Egerton University, P.O. Box 536 Postal code 20107 Njoro, Kenya.
Email: gobare@africaonline.co.ke

Abstract
The urgent need to improve livestock productivity in sub–Saharan Africa in order to keep pace
with expected increases in demand for meat and milk is very topical. Breed improvement
provides key entry points for increasing productivity in cattle populations. However, there are
tendencies for genetic improvement programs to focus on single, market driven traits such as
milk or meat production in isolation of environmental constraints and broader livestock system
functions which cattle perform in developing countries. This potentially leads to genotypes that
are not well adapted to the environment and not capable of performing the multiple roles that
cattle assume in smallholder systems. In developing countries, many important functions of
livestock are embedded in non-tradeable traits that are neither captured in economic analysis
nor considered in livestock improvement programs. This study employs Participatory Rural
Appraisal (PRA) ranking techniques and conjoint analysis to evaluate preferences of cattle
keepers in pastoral and agro-pastoral systems of selected sites in Kenya and Ethiopia for various
cattle traits. These systems are characterized by low input management and prevalence of
various cattle diseases. Trypanosomosis is a serious disease constraint in Ghibe valley of
Ethiopia and some of the pastoral areas in Kenya. The results indicate that farmer preferences
for cattle traits are influenced by various factors including production system characteristics,
infrastructural constraints and environmental conditions, especially in relation to disease
prevalence and availability of cattle feeds. In the crop-livestock systems of Ghibe valley in
Ethiopia, preferred cattle traits include trypanotolerance, reproductive potential and fitness to
traction. Milk production is a less important trait. On the other hand, in the pastoral and agropastoral systems of Kenya, important traits include trypanotolerance, reproductive potential,
coat colour and watering needs.
Keywords: Cattle production system, trait preferences, choice experiment, Kenya, Ethiopia

∗

Contact author, Email: a.ouma@cgiar.org

1

2. Background and Aim of the Study
There is an urgent need to improve livestock productivity in sub–Saharan Africa in order to keep
pace with expected increases in demand for livestock products. In sub-Saharan Africa, demand
for meat and milk has almost doubled over the past two decades; In Eastern Africa the same trend
has been observed. For instance, milk consumption in the region increased from 1.5 million
metric tons in 1975 to 3.2 million metric tons in 1995, while meat consumption rose from 0.5
million metric tons to 0.9 million metric tons (Ehui et al., 2002) in the same period. Further
projections indicate that total consumption of meat and milk in eastern Africa will more than
double between 1997 and 2020 to reach 1.9 and 7.3 million metric tons respectively, by 2020.
Unfortunately, livestock productivity in sub-Saharan Africa remains very low compared to other
parts of the world because producers are beset by several technical, institutional and
infrastructural constraints related to feeding, animal health and genotype. The severity of these
constraints varies by the various systems under which cattle production takes place. The
production systems are determined by agro-ecology and commonly differ in exhibiting various
stress factors, such as water shortages, disease and parasites as well as temperature extremes. The
constraints faced by livestock producers would need to be overcome or minimized in order for
improved livestock productivity to be realized.
Animal diseases, especially those caused by parasites, impose severe constraints on animal
production in sub-Saharan Africa. Trypanosomosis is one of the major constraints to livestock
productivity, with forty six million cattle at constant risk of infection (FAO, 1991; Kristjanson et
al, 1999). The annual cost of trypanosomosis in terms of foregone milk and meat production is
estimated at US$1.3 billion (ibid.). These are colossal amounts that could be invested in
alternative development efforts such as improvement of dilapidated physical infrastructures of
sub-Saharan Africa. Control of trypanosomosis currently relies largely on the use of
chemotherapeutic drugs, tsetse vector control or an integrated control approach combining
several strategies. In most cases, such control remains costly and only partially effective. The
control of trypanosomosis using trypanocidal drugs to treat or prevent the disease is limited by
drug costs and availability, and by the development of drug-resistance in target parasites.
Attempts to develop an effective vaccine have so far been unsuccessful and immediate prospects
are not promising.
Breed improvement, through genetic control provides key entry points for increasing productivity
in cattle populations especially those susceptible to trypanosomosis. The advantage of genetic
control over other methods of control is that genetic changes are cumulate and permanent, and
there are no recurring costs to the end users. However, there are tendencies for breed
improvement programs to focus on single, market driven traits such as milk or meat production in
isolation of broader livestock system functions and constraints. This potentially leads to
genotypes not well adapted to the environment and not capable of performing the multiple roles
that livestock assume in livestock systems of developing countries. In developing countries, many
important functions of livestock are embedded in traits that are not traded in the market. These
include functions and products such as traction, manure, form of security (insurance), dowry
payment and use in traditional ceremonies. This study aims to assess farmer preferences for cattle
traits by deriving economic values for cattle traits in selected production systems in Kenya and
Ethiopia, focusing particular attention on farmer preferences for trypanotolerance, relative to
other traits which could be introduced through breeding programs that utilize resistant genotypes.

2

3. Data Sources
The data used in this paper is part of an on-going collaborative livestock project between three
institutions*, ETH, ZIL and ILRI being conducted in Kenya and Ethiopia. The analysis presented is
based on primary data collected through farmer group discussions conducted between February
and March 2004 in selected sites in Kenya and Ethiopia, and a cross-sectional household level
survey conducted on a sample of one hundred cattle keeping households in Kajiado district in
Kenya using choice experiments in August 2004. The sampled households were randomly selected
from seven sub-locations of Magadi division, Kajiado district. The farmer group discussions were held as
part of a pilot study to identify existing cattle production systems to be used in targeting research

areas for the household level survey. In addition, the farmer group discussions were important in
identifying cattle traits preferred by farmers. These traits were then used in the choice experiment
survey.
4. Study Area
Since the study focuses on farmer preferences for cattle traits paying particular attention to
trypanotolerance as a trait, tsetse challenge areas have been identified and contrasted with nontsetse challenge areas. Given that the presence of tsetse flies is a rudimentary indicator of tsetse
challenge, spatial mappings of tsetse fly distribution in Kenya and Ethiopia has been done as an
initial attempt at targeting research areas with tsetse challenge. Data on cattle densities in Kenya
at the division level has also been overlayed to assess areas at risk of trypanosomosis. Two tsetse
challenge districts with different cattle production systems have been randomly selected in Kenya
to capture variations in cattle trait preference structure across three production systems; pastoral,
agro-pastoral and crop-livestock. These systems are characterized by low input management and
prevalence of various cattle diseases. In Ethiopia, various sites in Ghibe valley, where croplivestock system is practised have been selected for the study. Farmer group discussions were
held in various sites in two districts, Narok and Kajiado in Kenya and Ghibe valley in Ethiopia.
Data collection for the household level, choice-experiment survey is still on-going; consequently
the choice experiment survey results presented in this paper is based on findings from only one
division in one district in Kenya, that is, Magadi division of Kajiado district.
4.1 Kajiado District
Kajiado district lies in the south-western part of Kenya (Figure 1). Most of the district lies in the
semi-arid and arid zones, and only 8 percent of its land has some potential for cropping.
Figure 1: Tsetse fly distribution in Kajiado district

*

ETH-Swiss Federal Institute of Technology; ZIL-Swiss Centre for International Agriculture: ILRI-International
Livestock Research Institute

3

Mean annual rainfall ranges from 300 to 800mm, and open grasslands predominate with small
areas of bush and woodland. Livestock and wildlife coexist in much of the area with several
major national parks bordering or falling within the district. Human population in Kajiado has
increased significantly over the last 20 years, from 149,000 in 1979 to 260,000 in 1989 and 406,
054 in 1999 (Government of Kenya, 2001). The area has historically been dominated by the
Maasai pastoralists. Per capita cattle ownership has declined from 3.2 Tropical Livestock Unit
(TLU)† in 1997 to 2.1 TLU in 1998 (ibid). This decline has been attributed to several factors
including increasing human population pressure, several severe droughts, diversification of the
Maasai economy and land tenure changes including sub-division of group ranches (Scarpa et al,
2003). The one hundred sampled households for the choice experiment survey were randomly
selected from seven sub-locations in Magadi division of Kajiado district. In Magadi division,
pastoral systems dominate; however, agricultural production is also practiced in some parts of
Ngurumani sub-location, Magadi division.
5. Methods
A variety of Participatory Rural Appraisal (PRA) tools were used during the farmer group
discussions, including scoring and ranking techniques, timeline and trend analysis, seasonal
calendar analysis as well as community institutional maps. Farmers were asked to indicate their
objectives of cattle keeping and then asked to identify the cattle traits or attributes that they prefer
in cattle, based on their prevailing local and environmental conditions. Pairwise ranking
technique for the attributes was then applied. The identified attributes were then used in the
construction of the choice experiment.
5.1 Conceptual Framework of Choice Experiment
The conceptual framework for choice experiments arises from the consumer theory developed by
Lancaster (1966) which postulates that preferences for goods are a function of the attributes or
characteristics possessed by the good rather than the good per se. An important implication of
this theory is that overall utility of a good can be decomposed into separate utilities for its
constituent characteristics or attributes. In terms of the utility function, this translates into using
the characteristics of goods as the arguments of the function. For cattle breeding, this permits the
analysis of farmer preferences in terms of the benefits they perceive to result from various cattle
traits. Lancaster’s model may be defined more precisely as follows: A consumer maximizes an
ordinal preference function for characteristics, U (z ) where z is a vector of characteristics
1, ........., r , possessed by a single good or combination of goods subject to the budget constraint
px ≤ K , where p is a vector of prices for each of these goods and K is income. Goods, x, are
transformed into characteristics, z, through the relation z = Bx, where B is a r× n matrix which
transforms the n goods into r characteristics.

Choice experiments and hedonic price analysis are alternative empirical applications to the
Lancaster consumer theory. The strength of both hedonic pricing and choice experiment
techniques is the ability to decompose revealed preference data, that is, price of goods in case of
hedonics and choice of goods profiles by individuals in case of choice experiments, into marginal
values or part worth estimates. However, the use of hedonic pricing is not practical when market
transactions data is poor as is the case in Africa. In the rural areas in Africa, most cattle
transactions do not take place in formal markets where transactions are transparent and easily
recorded. Rather, transactions usually take the form of private agreements between buyers and
sellers using cash or barter. Secondly, many cattle are never traded or sold, but stay within the
†

1 TLU equivalent to a 250 Kg animal

4

farm household or are passed on to other households through traditional practices such as dowry
payments. In addition, market prices may be highly distorted due to the presence of
intermediaries. Consequently, price data is likely to be incomplete and can suffer from substantial
measurement errors. In choice experiments, since preferences are measured directly, and then
related to utility, the results are less likely to be adversely affected by traits that are not priced or
transactions that do not occur through organized markets. Consequently, choice experiment
technique was used in this study.
5.2 Experimental Design

Attributes for cows and bulls were identified during the farmer group discussions. For instance, in
the crop-livestock systems of Ghibe valley in Ethiopia, important cattle traits include
trypanotolerance, reproductive potential and fitness to traction. Milk production is a less
important trait (Ouma et al, 2004). A total of eight preferred traits were identified for cows in the
Kenyan sites. These were then used to design the choice experiment, with each trait having two to
three levels. Table 1 presents the various traits and their levels for cows.
Table 1: Traits and Trait Levels for Cows Used in the Choice Experiments
Traits
Trypanotolerance
Milk yield
Reproductive performance
Feeding requirements
Purchase price at 24 months

Watering frequency

Coat colour
Liveweight at 24 months

Levels
1. Tolerant
2. Susceptible
1. 1-2 litres per day
2. 2-4 litres per day
1. 1 calf per year
2. 1 calf every 2 years
1. Need to use supplementary purchased feed rations
2. No need for supplementary purchased feed rations
1. Less than KSh 15,000
2. KSh 15,000 – 20,000
3. Greater than KSh 20,000
1. Once a day
2. Twice a day
3. More than twice per day
1. Light-colored
2. Dark-colored
1. Less than 250 Kg
2. 250-300Kg-cows
3. Greater than 300 Kg

Cards with pictorial representations of the differences in the levels of traits were used to
demonstrate each cattle profile to survey respondents. The administration of the choice
experiment was conducted in the following manner. Each respondent was first introduced to the
type of choice task required and then he/she was presented with twelve sets of pair-wise choices
for cows drawn from an orthogonal design. Each choice task required the respondent to choose
one animal profile he would prefer to buy for rearing from the two profiles presented for each
choice task. If neither of the profiles was found satisfactory, the respondent could choose the
“zero” option and state that he preferred neither. An example of one choice task is presented in
the appendix.
5.3. Analytical Techniques and Data Analysis

In choice experiments, we assume that a sampled individual q (q=1,…,Q) faces a choice amongst I
alternatives in each of T choice situations. Individual q is assumed to consider the full set of offered
5

alternatives in a choice situation t and has to choose the alternative with the highest utility. The
(relative) utility associated with each alternative i as evaluated by each individual q in choice situation
t is represented in a discrete choice model by a utility function of the general form:

Uitq = β q X itq + eitq

(1)

Where; Xitq is a vector of explanatory variables that are observed by the analyst and include attributes
of the alternatives, socio-economic characteristics of the respondent and descriptors of the decision
context and the choice task itself in choice situation t. ßq and eitq are parameters to be estimated and
error terms respectively. In cases such as this, where an economic agent chooses from among a set of
multiple choices, multiple choice models have been employed to model choice behaviour.
Multinomial logit and conditional logit models have been used in this study to model choice
behaviour.
Supposing observed choice Y has values 0, 1, …., m and Xi includes individual q’s characteristics
while Zi are the choice specific characteristics, the multinomial logit model to assess the effect of Xi
on the probability that choice Y has trait j, can be presented thus (Greene, 1997);
Prob (Yi = j ) =

e
m

β ′j x

∑e

β k′ x i

, j = 0, 1, ....... m

(2)

k =0

The independent variable Xi does not vary across choice alternatives but varies across individuals.
On the other hand, independent variables Zi varies across households as well as choice alternatives.
Therefore, to assess the impact of Zi (choice-specific attributes) on Y, the appropriate model to use is
the conditional logit model, presented below for a total of J alternatives:
Prob (Yi | xi = j ) =

e
J

β ′z ij

∑e

β ′z ij

, j = 1, 2, ....... J

(3)

j =1

In conditional logit, we estimate β while in multinomial logit, we estimate βj.
Producers in different production systems may face different trade-offs in cattle attribute
preferences due to varying production activities, and this may have different policy implications.
Therefore, we obtained estimates for the two production systems identified in the area, as well as
pooled estimates for both systems.
6. Results and Discussion

The multinomial and conditional logit models were estimated using STATA 8.0 (2001). The
maximum likelihood parameter estimates from the conditional logit model are presented in Table
2. Since the traits had 2-3 levels each, one level was left out as base during estimation. The
overall explanatory power of the model is good with a pseudo-R2 of 0.46
The trait trypanotolerance is strongly significant (p<0.01) and has the expected positive sign in all
the systems, implying that farmers in the study area prefer trypanotolerant cattle breeds. The
study area is a tsetse challenge area, where cattle are at constant risk of trypanosomosis infection.
Dark coat colour has a positive sign, indicating that respondents prefer cows with dark coat
colour relative to light coat coloured ones. This is despite the dark coated animals having higher
chances of being bitten by the tsetse flies compared to the light coated ones. The study area is
mainly inhabited by the Maasai communities, who have a high preference for dark coat coloured
cattle. Dark red coat colour is considered beautiful and appealing while animals with plain black
coat with white spots around the neck are slaughtered during ceremonial functions.

6

Table 2: Conditional Logit Maximum Likelihood Parameter Estimates of Cow Attributes
All systems
(pooled)
Agro-pastoral system Pastoral system
Cow attributes
***
Trypanotolerant
2.1364 (0.1899)
2.4171*** (0.2864)
2.0489*** (0.2659)
***
***
Dark coat colour
1.8662 (0.1620)
1.9330 (0.2616)
1.9142*** (0.2116)
**
Weight at 2 year (>300Kg)
0.1780 (0.2274)
0.7257 (0.3577)
-0.2355 (0.3057)
Weight at 2 years (250-300Kg)
0.2209 (0.1816)
0.4767 (0.2949)
0.0111 (0.2411)
Needs watering once a day
0.5796*** (0.2220)
-0.4071 (0.3809)
1.2062*** (0.2889)
*
Needs watering twice a day
0.3183 (0.1925)
-0.1451 (0.3110)
0.6025** (0.2543)
Average milk production: 2-4 Lt/day
0.1891 (0.1614)
0.1491 (0.2556)
0.2015 (0.2148)
High reproduction potential
1.2813*** (0.1883)
1.1021*** (0.3093)
1.4936*** (0.2539)
Need to buy purchased feeds
0.0924 (0.1761)
0.2629 (0.2781)
0.0007 (0.2339)
Price at 2 years (> KSh20,000)
-0.3816 (0.2696)
0.4757 (0.4293)
-0.9552 *** (0.3622)
Price at 2 years (KSh15,000-20,000)
-0.1240 (0.2097)
0.3021 (0.3224)
-0.4109 (0.2843)
49%
Pseudo-R2
46%
43%
-226.3
Log likelihood
-591.1
-348.0
427.1
Likelihood ratio
912.49
518.9
0.0000
Significance level
0.0000
0.0000
11
Degrees of freedon
11
11
756
Number of observations
1800
1044
*** ** *
, , indicate that coefficients are statistically significant at the 1, 5 and 10% levels, respectively, using P-values in
maximum likelihood estimation. Robust standard errors are indicated in parentheses.

High relative liveweight at 2 years has the expected positive sign for all the systems but is only
significant for the agropastoral systems (p<0.05) and negative for the pastoral systems. In the
pastoral systems, cattle are frequently moved from place to place in search of pasture and water.
Consequently, farmers would prefer animals that are lighter and can be moved easily and faster.
The trait for cows that need watering only once a day is strongly significant (p<0.01) and has a
positive sign, implying that farmers prefer cows that need to water only once a day relative to
those that need to water more than two times a day. However, when differentiated by production
systems, the sign is negative for the agropastoral system and positive for the pastoral systems
(p<0.01). This is because water is not a constraint in the agropastoral systems, mainly found in
Ngurumani sub-location of Magadi division. Ngurumani is on the escarpment to the western edge
of Magadi with plenty of water and fertile arable land. However, in the pastoral systems water is
a major constraint.
High reproductive potential has a positive impact on herd productivity and herd size, so a positive
sign was expected. This is also highly significant (p<0.01) in all the systems. Daily milk
production of 2-4 litres (representing high milk production as opposed to 1-2 litres per day) is
positive but not statistically significant, suggesting that respondents’ choices are not strongly
influenced by changes in milk yield. This implies that milk production is not a highly ranked
parameter in livestock farmers’ trait preference yet most analyses of cattle systems and producer
decisions focus on marketable yield attributes such as milk production for dairy animals and beef
output for beef animals.
Three different price levels of cows were considered in the choice experiment. Higher prices
yielded negative parameter estimates, implying that farmers are less likely to choose costly
options. However, in the agro-pastoral system, the reverse is observed. Some of the farmers
consider higher purchase prices to be beneficial since it shows that the animal has a higher value
and consequently would fetch high market prices when sold. Another possible explanation would
be due to the multifaceted functions of cattle such that market price is a low rank parameter in
livestock farmers’ trait preference.

7

7. Concluding Remarks

The integration of trypanotolerance traits of cattle into breed improvement programs in tsetse
challenge areas provides a viable option to enhance cattle productivity. Research efforts should
be geared towards this. Empirical results indicate that farmers in both pastoral and agro-pastoral
systems value the adaptation traits, more so trypanotolerance.
Culture and tradition have an influence on the traits preferred by farmers. For instance, farmers in
the study site prefer dark coat-coloured cattle, especially the dark red colour and other dark
colours. The dark coated animals are used for slaughtering during ceremonial functions.
Differences in preferences across production systems are observed due to the varying production
activities and available resources. Producers in pastoral systems have a high preference for hardy
animals that are able to withstand severe environmental conditions. This is in contrast to the
producers in agropastoral systems that prefer animals that are trypanotolerant and heavy, so as to
fetch favourable market prices.
Acknowledgement

The authors gratefully acknowledge the funding support provided by the Swiss Centre for
International Agriculture (ZIL).
References

Ehui, S., Benin, S., Williams, T. and Meijer, S. (2002) Food Security in sub-Saharan Africa to
2020: Socio-economic and Policy Working Paper. International Livestock Research Institute,
Addis Ababa, Ethiopia.
FAO (1991) Programme for the Control of African Animal Trypanosomosis and Related
Development. Second Meeting of the Inter-Secretariat Coordinating Group, Food and
Agriculture Organisation of the United Nations, Rome.
Government of Kenya (2001), "Kajiado District Development Plan, 1997 - 2001," Government
Printer, Nairobi Kenya.
Greene, W. (1997) Econometric Analysis. Macmillan Publishing Co. 3rd edition, Prentice Hall.
Kristjanson, P., Swallow, B., Rowlands, J., Kruska, R. and de Leeuw, P. (1999) Measuring the
costs of African animal trypanosomosis, the potential benefits of control and returns to
research. Agricultural Systems, 59, 79-98.
Lancaster, K. (1966) A New Approach to Consumer Theory. Journal of Political Economy, 74,
132-157.
Ouma, E., Janssen-Tapken, U., Drucker, A., Gibson, J., Obare, G., Ayalew, W., Kadarmideen, H.
and Abdulai, A. (2004). “Developing Optimised Cattle Breeding Schemes with a special focus on
Trypanotolerance based on the demand and opportunities of smallholder farmers in Eastern
Africa” PRA Report
Scarpa, R., Ruto, E., Kristjanson, P., Radeny, M., Drucker, A. and Rege, E. (2003) Valuing
Indigenous Cattle Breeds in Kenya: an empirical comparison of stated and revealed preference
value estimates. Ecological Economics, 45, 409-426.
STATA (2001) User’s Guide Release 8, Stata Press College Station, Texas

8

Appendix
Choice task 1 (Alternative 1)
Tolerant to
Trypanosomosis: Able
to be in good condition
despite infection. No
need for treatment
drugs

Average milk yield per
day (2-4 Lts)

Dark Coat
Colour

Weight at 2 years
(less than 250 Kg)

Watering needs:
Need to be watered
two times a day

Reproduction
Potential: 1 calf
every two years

Does NOT require
purchased
supplementary
feeds

Purchase price:
More than KSh
20,000

Choice task 1 (Alternative 2)
Susceptible to
Trypanosomosis: Poor
condition after
infection. Needs
treatment drugs

Average milk yield per
day (1-2 Lts)

Light Coat
Colour

Weight at 2 years
(300 Kg)

Reproduction
Potential: 1
calf every year

Requires purchased Purchase price: Less
supplementary feeds than KSh 15,000

Watering needs:
Need to be watered
two times a day

Choice task 1 (Alternative 3)
None of the above profiles

9