The concept of a livelihood strategy has become central to development practice in recent years. Nonetheless, precise identification of livelihoods in quantitative data has remained methodologically elusive. This paper uses cluster analysis methods to operationalize the concept of livelihood strategies in household data and then uses the resulting strategy-specific income distributions to test whether the hypothesized outcome differences between livelihoods indeed exist. Using data from Kenya’s central and western highlands, we identify five distinct livelihood strategies that exhibit statistically significant differences in mean per capita incomes and stochastic dominance orderings that establish clear welfare rankings among livelihood strategies. Multinomial regression analysis identifies geographic, demographic and financial determinants of livelihood choice. The results should facilitate targeting of interventions designed to improve household livelihoods.