experiment, with each trait having two to three levels. Table 1 presents the traits and their
levels. The choice experiment was administered through a household questionnaire survey, on
a sample of 303 cattle keeping households in Kenya and 204 in Ethiopia using in-person
interviews. Cards with pictorial presentations 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 either
twelve sets of pair-wise choices for cows or eleven for bulls drawn from a main effects only
fractional factorial 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 “none”
option and state that he preferred neither. The household questionnaire also covered other
aspects on the household and farm characteristics as well as market and resource access.
3.2 Estimation Methods
The theoretical foundation of choice experiments derives from Lancasterian consumer
theory (Lancaster, 1966) and the random utility framework developed by Marshak (1960). In
this study, the assumptions supporting the multinomial logit model are applied, the most
prominent being that each error term is independently and identically distributed extreme
value with a cumulative distribution; F(εn) = e-e-εn (Train, 2003). The probability of
individual n, choosing alternative i is specified thus;
αni +λjsn +βn xni
P = ,____________
(1)
(2)
ni ∑ eαnj +λjsn +βnxnj
while the sample log-likelihood function is;
NJ
LL(αnj,βn,λj)=∑∑ynjlnPnj
n =1 i =1