The study applied Multinomial Logit (MNL) model to survey data from a sample of 249 farm households in northern Ghana. The model was used to investigate factors that influence farm families’ decisions to raise a particular small ruminant species (i.e., sheep or goat or both). The MNL analysis indicates that the probability of raising sheep, goat or both animals was influenced by agro-ecological zone, sex and religious background of farmers, risk attitude and income from small ruminant production. In devising strategies (to select farm households) to improve subsistent small ruminant production system, livestock technical staffs must recognize important demographic and farm characteristics as well as risk and income perceptions associated with sheep and goat of the households. In addition, such programs must be supported with improved livestock housing technology that can provide opportunities to control sheep and goat production risks, including theft or predator attacks, disease infestation and exposure to harsh environmental conditions (rains and sun rays).