This paper investigates factors that impact cropland, labour and income diversification decisions and subsequent impacts on welfare. We use geo-referenced household data collected in 2011 from Malawi. The results show that measures of climate risk generally increase diversification across labour, cropland and income indicating that rainfall riskiness is a “push” factor for these indices. Our results also reveal that “pull” factors such as household wealth and education status of the household generally increase diversification across labour, land and income. Results also show that vulnerability to poverty is lower in environments with greater climate variability. Availability of services and support from rural institutions tends to increase diversification and reduce vulnerability to poverty. Looking at welfare measures as a function of diversification indices, all three measures of diversification increase consumption per capita, but income diversification has the strongest impacts on current consumption per capita and in reducing vulnerability to poverty.